The Acquisition of Grammatical Gender by Second Language Learners of Dutch: Evidence from Perception MA Thesis English Language and Culture, Utrecht University Programme Language and Development (Linguistics) Marije Takens Student number 0038210 Supervisor Dr. Sharon Unsworth Co-reader Prof. Dr. René Kager August 2008 Acknowledgements Without the support of a considerable number of people inside and outside the linguistics department of Utrecht University this thesis could not have been written. Firstly, I want to thank my supervisors Dr. Sharon Unsworth and Prof. Dr. René Kager. Their expertise has given shape to the research project we carried out and their supervision has resulted in these pages. I also want to thank my internship partner Marlinda Andeweg, with whom I worked on this project for many months. Dr. Hugo Quené, who has assisted in the practical set-up of the experiment by recording the large amount of sentences we eventually presented to our participants, certainly was the 'voice' of the project and I am grateful for this. My gratitude also goes out to Dr. Laura Sabourin, who has allowed us to use stimuli from her PhD study as stimuli for our production task. I also want to thank the members of the Utrecht Institute of Linguistics OTS whose comments and suggestions during the presentation of our research set-up have been of much use. Many thanks go to Sander van der Harst of the UiL-OTS for helping me with the statistics. A vital condition to the current project has been the participation of thirty-four English native speakers who learnt Dutch as a foreign language. Without their support and effort this research project could not have been carried out at all. The same holds for all Dutch native speaker control subjects. The kind participation of volunteers has made this project possible and I want to express my gratitude to all of them. I cannot thank my family and friends enough, especially my parents, my sisters, Saskia, Xiaoli, Camie, Hannah, Erde, Gion and Annemarie, for the necessary support, distraction and useful (statistical) tips they gave me during the process of writing my thesis. Finally, my special thanks go out to Sven for always being kind and supportive and of course for lending me his laptop during the project. 2 Abstract Previous studies into the acquisition of Dutch grammatical gender have shown that second language learners have serious difficulties with the definite determiners, especially with neuter het. Second language learners of Dutch systematically show a delay in the acquisition of the definite determiners as well as signs of fossilisation in a non-targetlike stage of overgeneralisation (Hulk & Cornips 2006a,b; Cornips et al., 2006; Blom et al., in press; Unsworth, in press; Cornips & Hulk, in press; Brouwer et al., in press). Learners mainly overgeneralise de, but make errors in the other direction as well by also using het for common nouns (Cornips et al., 2006; Unsworth, in press). Various researchers argue that the difficulties with the Dutch definite determiners may be due to an underspecification of the gender feature in second language learners, i.e. learners are aware of gender but do not have a complete knowledge and do not know the right gender specification (yet) (Hulk & Cornips, 2006b; Cornips & Hulk, in press; Brouwer et al., in press). Most of the previous research focuses on production. The only study to date which focuses on comprehension is the study by Brouwer et al. (in press). This thesis attempts to gain more insight into the role of comprehension in the second language acquisition of Dutch grammatical gender. It investigates the perception of the definite determiners by second language learners of Dutch. Since het is mostly non-salient in informal speech as a result of its phonological context, while de is mostly salient, second language learners possibly fail to perceive het. A failure to perceive het is likely to lead to a failure to store het in the mental lexicon. A failure to represent het in the mental lexicon may lead to an underspecification of the grammatical gender feature in second language learners of Dutch, causing them to overgeneralise de. This proposal is put forward as the saliency hypothesis. This thesis also investigates the second language learners' production capacities to determine whether they can correctly produce gender. The results of the perception experiment indicate that the second language learners have difficulties with the perception of het. The analysis of the influence of phonological context on the misperception of het and the results of the perception experiment and production task combined suggest that the saliency hypothesis is borne out. However, because the hypothesised hierarchy of difficulty is only largely (and not entirely) borne out in the results, the influence of phonological context on the misperception of het is not entirely clear. The results of the perception experiment and the production task combined suggest that a better perception of het may lead to a better production of het and vice versa. However, the effect of phonological context on this relationship has not been investigated. Further research will have to shed more light on these issues. The results of the production task are in line with the previous research. The vast majority of the second language learners uses de with neuter nouns. This suggests that they overgeneralise by means of a default strategy. Almost one third of the second language learners also uses het with common nouns. The results of the production task also show frequency effects. The second language learners appear to find the medium frequency nouns more difficult than the high frequency nouns. The factors age of acquisition, proficiency and length of exposure appear to have an effect on performance. Earlier acquisition and a higher proficiency level seem to lead to a better perception of het. Furthermore, earlier acquisition and lengthier exposure seem to lead to a better production of het and a higher proficiency level seems to lead to a better production of both de and het. 3 Contents Abstract ...................................................................................................................... 3 1 Introduction .............................................................................................................. 7 2 Theoretical Framework .......................................................................................... 10 2.1 Grammatical Gender ....................................................................................... 10 2.2 The Dutch Grammatical Gender System ........................................................ 11 2.3 The Acquisition of Grammatical Gender ......................................................... 13 2.4 The Acquisition of Dutch Grammatical Gender ............................................... 16 2.5 The Saliency Hypothesis: The Influence of Phonological Context .................. 21 2.5.1 Introduction ............................................................................................... 21 2.5.2 The Cliticisation of Het and De ................................................................. 23 2.5.3 T-deletion in Het ....................................................................................... 26 2.5.4 Summary .................................................................................................. 27 3 The Present Study ................................................................................................. 28 3.1 Introduction ..................................................................................................... 28 3.2 Research Questions........................................................................................ 28 3.3 Hypotheses ..................................................................................................... 29 3.4 Predictions ...................................................................................................... 32 4 The Perception Experiment ................................................................................... 33 4.1 Introduction ..................................................................................................... 33 4.2 The Questionnaire........................................................................................... 34 4.2.1 The Second Language Learners .............................................................. 34 4.2.2 The Dutch Native Speaker Control Group ................................................ 34 4.3 Participants ..................................................................................................... 35 4.4 Experimental Design: Stimuli .......................................................................... 37 4.4.1 Introduction ............................................................................................... 37 4.4.2 The Nonce Nouns ..................................................................................... 38 4.4.3 The Manner Adverbs ................................................................................ 40 4.4.4 The Cliticisation of Het and De in the Experiment .................................... 40 4.4.5 The Fillers ................................................................................................. 42 4.4.6 The Distribution of the Nonce Nouns Across the Contexts ....................... 43 4.5 Procedure: Method.......................................................................................... 45 4.6 The C-Test ...................................................................................................... 47 4.7 Results ............................................................................................................ 47 4.7.1 Introduction ............................................................................................... 47 4.7.2 Factors to Observe ................................................................................... 48 4.7.3 Group Results........................................................................................... 51 4.7.3.1 General Results ................................................................................. 51 4.7.3.2 The Dutch Control Group ................................................................... 52 4.7.3.3 The Second Language Learners........................................................ 52 4.7.3.4 Comparison of the Groups ................................................................. 53 4.7.3.5 Summary............................................................................................ 55 4.7.4 Group Results per Factor ......................................................................... 56 4.7.4.1 Age of First Exposure ........................................................................ 56 4.7.4.2 Length of Exposure ............................................................................ 59 4.7.4.3 Intensity of Exposure ......................................................................... 61 4.7.4.4 Proficiency ......................................................................................... 64 4.7.4.5 Correlation Between the Factors and Performance ........................... 67 4.7.4.6 Summary............................................................................................ 68 4 4.7.5 The Influence of Phonological Context ..................................................... 69 4.7.5.1 The Contexts Before Het.................................................................... 70 4.7.5.1.1 Group Results ............................................................................. 70 4.7.5.1.2 Group Results per Factor: Age of First Exposure ........................ 72 4.7.5.1.3 Group Results per Factor: Proficiency ......................................... 74 4.7.5.1.4 Summary ..................................................................................... 76 4.7.5.2 The Contexts After Het ...................................................................... 77 4.7.5.2.1 Group Results ............................................................................. 77 4.7.5.2.2 Group Results per Factor: Age of First Exposure ........................ 79 4.7.5.2.3 Group Results per Factor: Proficiency ......................................... 81 4.7.5.2.4 Summary ..................................................................................... 83 4.7.5.3 The Het-Stop Contexts ...................................................................... 84 4.7.5.3.1 Group Results ............................................................................. 84 4.7.5.3.2 Group Results per Factor: Age of First Exposure ........................ 86 4.7.5.3.3 Group Results per Factor: Proficiency ......................................... 88 4.7.5.3.4 Summary ..................................................................................... 89 4.7.5.4 The Full Contexts of Het .................................................................... 90 4.7.5.4.1 Group Results ............................................................................. 90 4.7.5.4.2 Group Results per Factor: Age of First Exposure ........................ 93 4.7.5.4.3 Group Results per Factor: Proficiency ......................................... 95 4.7.5.4.4 Summary ..................................................................................... 98 4.7.5.5 The Stop-Het-Stop Contexts .............................................................. 99 4.7.5.5.1 Group Results ............................................................................. 99 4.7.5.5.2 Group Results per Factor: Age of First Exposure ...................... 101 4.7.5.5.3 Group Results per Factor: Proficiency ....................................... 103 4.7.5.5.4 Summary ................................................................................... 104 4.7.6 Individual Results ................................................................................... 105 4.7.7 General Summary .................................................................................. 106 5 The Production Task............................................................................................ 109 5.1 Introduction ................................................................................................... 109 5.2 Participants ................................................................................................... 109 5.3 Experimental Design: Stimuli ........................................................................ 109 5.4 Procedure: Method........................................................................................ 110 5.5 Results .......................................................................................................... 110 5.5.1 Introduction ............................................................................................. 110 5.5.2 General Results ...................................................................................... 111 5.5.3 Results per Factor .................................................................................. 112 5.5.3.1 Age of First Exposure ...................................................................... 112 5.5.3.2 Proficiency ....................................................................................... 114 5.5.3.3 Length of Exposure .......................................................................... 116 5.5.3.4 Intensity of Exposure ....................................................................... 118 5.5.3.5 Correlation Between the Factors and Performance ......................... 120 5.5.3.6 Regression Analysis ........................................................................ 121 5.5.4 Individual Results ................................................................................... 122 5.5.5 General Summary .................................................................................. 123 6 Perception and Production .................................................................................. 125 7 General Discussion ............................................................................................. 129 7.1 Perception ..................................................................................................... 129 7.2 Production ..................................................................................................... 133 5 7.3 Perception and Production ............................................................................ 135 7.4 Theoretical Implications ................................................................................ 136 8 Conclusion ........................................................................................................... 138 References ............................................................................................................. 140 Appendices ............................................................................................................. 148 Appendix 1 .......................................................................................................... 148 Appendix 2 .......................................................................................................... 149 Appendix 3 .......................................................................................................... 151 Appendix 4 .......................................................................................................... 163 Appendix 5 .......................................................................................................... 165 Appendix 6 .......................................................................................................... 172 6 1 Introduction When young children acquire their native language or languages they do this effortlessly and naturally, provided that they possess a normal language capacity and do not suffer from linguistic impairments or general cognitive deficits. Young children are wholly successful in acquiring the language(s) of their environment and appear to become fluent in any language they learn at this stage. Second language acquisition, however, is a different story. Most adolescents and adults who attempt to develop a native mastery of a foreign language do not succeed (e.g. Strozer, 1994). Apparently, learning a second language is much more difficult than first language acquisition. Many theorists have sought to explain the differences between first-and second language acquisition. Some focus on learner’s innate capacities for language acquisition, others emphasize the role of general psychological processes, and still others point to the importance of the sociocultural dimension. The differences between first-and second language acquisition are often related to Universal Grammar, i.e. innateness. It is argued that there are maturational constraints that cause a critical period or multiple critical periods for language acquisition. After such a period (or periods), it becomes increasingly difficult to learn a new language due to the inability of parameter fixing after a certain age (see e.g. Strozer, 1994). The theories on a critical period for language acquisition (see Hyltenstam & Abrahamsson, 2003 for a recent overview) and multiple critical periods (see e.g. Meisel, 2007; Schachter, 1996; Beck, 1998; Eubank & Gregg, 1999) all focus on innateness. Nevertheless, innateness remains a much debated issue. Cognitive and developmental psychologists place the emphasis on general cognitive processes, stating that general theories of learning can account for the development of syntax and other complex aspects of language acquisition. They are of the opinion that there is no need to hypothesise about a specific language learning module in the brain, and that language acquisition and general learning are not distinct processes (see e.g. Favreau & Segalowitz, 1983; Segalowitz, 2003; DeKeyser, 1998, 2001). Connectionist models, computer models imitating the brain, are used to support this view (see e.g. Ellis, 2002, 2003, 2005). The sociocultural perspective states that cognitive development, of which language acquisition is part, takes place because of social interaction (see e.g. Vygotsky, 1978). Connectionist and sociocultural models, however, cannot account for one crucial aspect of language acquisition, namely the fact that learners of a (second) language come to know many things about the language they cannot have derived from the input. Chomsky has called this the poverty of the stimulus (see e.g. Chomsky, 1980; Strozer, 1994). The innatist theory can account for this phenomenon. Therefore, an innatist view is taken in this thesis. Although a more advanced age of acquisition appears to have a negative influence on the ability to acquire a language natively, this does not imply that it is impossible for later learners to achieve a high level of attainment in the new language. There are second language acquirers who reach a level of ultimate attainment (see Singleton & Ryan, 2004 for a recent discussion).1 Relatively few second language acquirers succeed in reaching such a level, but the percentage of them who do (around 10-15%) cannot simply be dismissed (see Birdsong, 1999 and Birdsong & Paik, 2008). 1 The term ultimate is not used here to suggest nativelike. Rather, ultimate attainment is used here to denote the "end state or asymptote of L2A, however close to or far from nativelike that state may be" (Birdsong, 1999: 10). 7 One phenomenon which has proved to be particularly difficult to learn for second language acquirers is grammatical gender. Most learners have great difficulty in mastering a second language grammatical gender system, even after considerable exposure to the target language in question. Grammatical gender is problematic for adult second language acquirers (Carroll, 1989; Dewaele & Veronique, 2001; Bruhn de Garavito & White, 2002; Franceschina, 2005 on Romance languages; Rogers, 1987 on German; Andersson, 1992; Hyltenstam, 1992 on Swedish) and for child second language acquirers (Carroll, 1989; Andersson, 1992; Blom et al., 2006; Hulk & Cornips 2006a). This thesis investigates the second language acquisition of Dutch grammatical gender. Dutch has a two-way gender-system, as nouns have either de (non-neuter, common gender) or het (neuter gender). Studies on the monolingual child acquisition of Dutch grammatical gender show that monolingual children have serious difficulties with the acquisition of neuter gender and that the problems persist until age 6 (see Bol & Kuiken, 1988; Gillis & De Houwer, 1998; Van der Velde, 2003, 2004, Blom et al., under review). It is widely known in linguistics that when first language acquirers have problems acquiring a certain phenomenon, the problems second language acquirers encounter can be expected to be even more serious. The acquisition of Dutch grammatical gender by second language learners, however, is relatively under-researched. Recently, interest in this topic has been increasing. Current research on the second language acquisition of Dutch grammatical gender by children and adults has reported persistent problems with Dutch definite determiners (Hulk & Cornips 2006a,b; Cornips et al., 2006; Brouwer et al., in press; Cornips & Hulk, in press; Blom et al., in press; Unsworth, in press) and problems with adjectival inflection (Weerman et al., 2002; Blom et al., 2005; Blom et al., in press). A pattern that is generally reported in these studies is a delay in the acquisition of Dutch neuter gender, manifesting itself in the overgeneralisation of Dutch common gender. Moreover, the data also suggest fossilisation in this non-targetlike stage of overgeneralisation. Both child and adult second language acquirers of Dutch show signs of fossilisation in the aforementioned studies. It is frequently observed that adult second language acquirers fossilise before reaching a native level of attainment (e.g. Bley-Vroman, 1990), while child second language acquirers are regularly observed to attain native levels of competence (see Hyltenstam & Abrahamsson, 2003 for a discussion). Grammatical gender appears to be exceptionally difficult for child second language acquirers. One could argue that the child second language acquirers of Dutch grammatical gender may not have reached their end state yet at the time of testing. This, however, is not supported by the data. The child acquirers in the study by Cornips et al. (2006) continue to overgeneralise until the age of 10 to 13 years, while the age of first exposure to Dutch (age of onset) ranges from birth to 4 years. Moreover, Unsworth (in press) mentions children who still overgeneralise when they have reached the relatively advanced age of 17 years, with age of first exposure ranging from birth to 7 years.2 The question that arises is why Dutch neuter gender is so difficult to learn for second language acquirers. Monolingual acquirers generally overcome the problems with Dutch neuter gender and become target-like, but this does not seem to be the case for second language acquirers. Instead, they seem to fossilise in the overgeneralisation stage. The problems with Dutch neuter gender second 2 The children in these studies have characteristics of both bilingual and early child second language acquirers. See Chapter 2, section 2.4, for more information. 8 language acquirers experience have many possible causes. The problems may be due to production difficulties, comprehension difficulties, or both. Although the relationship between the perception and production of second language speech sounds is still rather unclear, it is clear that both perception and production are important during the acquisition process (see e.g. Llisterri, 1995; Rochet, 1995). Most of the current research on the second language acquisition of Dutch grammatical gender focuses on production (Hulk & Cornips 2006a,b; Cornips et al., 2006; Blom et al., in press; Unsworth, in press), and a limited amount of data on comprehension is available (Brouwer et al., in press). Therefore, the goal of the present study is to gain more insight into the role of comprehension in the second language acquisition of Dutch grammatical gender. The aim of this thesis is to gain more insight into the perception of Dutch definite determiners by second language learners. As mentioned previously, especially neuter het causes problems for second language learners. The research presented here is motivated by our research team’s observation that the Dutch definite determiner [hεt] is more often than not reduced to the non-salient forms [ət] and [ə] in fast, informal speech as a result of its phonological context (see e.g. Booij, 1995). It is possible that second language learners of Dutch systematically fail to perceive het as a result of its non-saliency. The input second language learners are exposed to is largely informal speech in which het is reduced. If second language learners indeed fail to perceive het on a regular basis, they are also likely to fail to store het in the mental lexicon. The mental lexicon is the speaker’s memory store for words from which information is retrieved (van Berkum, 1997). The failure to represent het in the mental lexicon may lead to an underspecification of the gender feature in second language learners, causing them to overgeneralise. More research into the perception of the Dutch definite determiners, in particular the perception of Dutch neuter het, may yield more insight into the persistent difficulties with grammatical gender second language learners experience. This thesis is organised as follows. Chapter 2 presents the theoretical framework. In this chapter the concept grammatical gender, the Dutch grammatical gender system and the acquisition of (Dutch) grammatical gender are touched on. The phonology of het in informal speech also receives attention in this chapter. In Chapter 3 the main research questions, hypotheses and predictions are put forward. In Chapter 4 the perception experiment is outlined, the results are given and the results are discussed. In Chapter 5 the production task is outlined, the results are given and the results are discussed. Chapter 6 combines the results of the perception experiment and the production task. Chapter 7 discusses the results of both experiments and Chapter 8 provides a conclusion. 9 2 Theoretical Framework 2.1 Grammatical Gender Corbett points out that the word gender etymologically derives from the Latin genus. It has reached us via Old French gendre, and originally meant kind or sort (Corbett, 1991: 1). Grammatical gender is also known as noun class. Linguistically, the term gender is used for a group of nouns and for the category as a whole: “thus we may say that a particular language has, say, three genders, masculine, feminine and neuter, and that the language has the category of gender” (Corbett, 1991: 1). Comrie (1999) defines gender as “a system in which the class to which a noun is assigned is reflected in the forms that are taken by other elements syntactically related to it” (457). In the same passage, he states that “crucial to the concept […] is thus agreement in gender between the noun and other items”. Gender agreement is the way gender is realised in language use, and therefore provides the basis for defining gender. A uniform definition for agreement does not exist and Corbett takes Steele’s (1978) as a working definition: The term agreement commonly refers to some systematic covariance between a semantic or formal property of one element and a formal property of another. For example, adjectives may take some formal indication of the number and gender of the noun they modify (taken from Corbett, 1991: 105, emphasis in text). Gender agreement is especially common in adjectives, i.e. in adjectival inflection (Corbett, 1991: 106). In Dutch, agreement in adjectives occurs when the indefinite article een (a) is used: (1) een groot huis (a big house; neuter); (2) een grote koe (a big cow; non-neuter). When the definite determiners are used there is also agreement, but it is the same for neuter and non-neuter nouns: (1) het grote huis (the big house; neuter); (2) de grote koe (the big cow; non-neuter). The terms concord and agreement can be used interchangeably, but some theorists prefer to distinguish the two (Corbett, 1991: 105). Grammatical gender is both a lexical and a syntactic property of a noun: “the assignment of gender to nouns is considered a lexical property while gender agreement is considered to be part of syntax” (Sabourin, 2003: 15). Because gender agreement is part of syntax, it is constrained by Universal Grammar. Corbett discusses over 200 languages and concludes that grammatical gender is certainly widespread, also among the Indo-European language family which dominates Europe and large parts of Asia. In some languages grammatical gender occurs in practically every sentence that is spoken, while it is entirely absent from other languages. A language can have two or more classes of genders. Some languages have three genders (e.g. German), others two (e.g. Dutch), and some have lost gender altogether (e.g. English) (Corbett, 1991: 1-2). Dutch nouns like boom (tree) and hond (dog) require the definite article de (non-neuter), while nouns like huis (house) and meisje (girl) require the definite article het (neuter). German has a similar gender system with one additional gender form; it has three, namely der (masculine), die (feminine), and das (neuter). The precise make-up of the Dutch grammatical gender system is discussed in section 2.2. Comrie notes that grammatical gender can easily be confused with two related concepts, namely natural gender (sex) and declension(al) class, but that these should be carefully distinguished 10 from grammatical gender.3 Grammatical gender and natural gender sometimes correspond, but they do not always match (the noun meisje (girl), for example, has neuter gender in Dutch and not feminine (non-neuter) gender). The same holds for declensional class. Comrie also discusses the semantic and formal principles by which nouns are allotted to different genders. Predictably, the semantic principle entails that “nouns are assigned to a gender according to their meaning”, and the formal principle that “nouns are assigned to gender according to their form” (Comrie, 1999: 458-59). There are also languages in which gender assignment appears to be quite arbitrary, however. In German, for example, der Löffel (the spoon), die Gabel (the fork) and das Messer (the knife) occur alongside each other (van Berkum, 1997: 119). Moreover, depending on the language, the same noun can have a different gender. Moon, for example, is masculine in German (Mond), feminine in French (lune) and neuter in Greek (Фεууαρι) (Holmes & Segui, 2006: 6). Van Berkum (1996) and Franceschina (2005) have attempted to answer the question of why grammatical gender exists in the first place. After reviewing the relevant literature, they conclude that grammatical gender appears to be a linguistic feature which serves to facilitate and optimise lexical and syntactic processing. Grammatical gender may disambiguate syntactic constructions (van Berkum, 1996). Gender agreement, in its turn, can “help to make functional lexical items such as articles more readily learnable, despite the interference of other factors” (Franceschina, 2005: 84). 2.2 The Dutch Grammatical Gender System Standard Dutch, like German and English, belongs to the West Germanic branch of the IndoEuropean language family. The Dutch grammatical gender system has two genders, neuter and nonneuter (also known as common or uter gender). Dutch grammatical gender is not reflected in the morphology of the noun but in agreeing elements accompanying the noun or referring to it. Grammatical gender in Dutch is clearly visible on singular definite determiners and singular demonstrative determiners, which morphologically vary according to the gender of the accompanying noun. Dutch gender is visible on definite determiners, demonstrative determiners, relative determiners and adjectival inflection with indefinite nouns. Neuter is associated with het-words (het huis, the house) and non-neuter with de-words (de ster, the star), i.e. neuter nouns are preceded by the neuter determiner het and non-neuter nouns are preceded by the non-neuter determiner de. The gender of the noun determines the form of any agreeing elements in the syntactic structure of the Determiner Phrase (DP) (see van Berkum, 1996 for a discussion). Grammatical gender in Dutch is often analysed as an uninterpretable feature on the noun, which checks or values the uninterpretable gender features on agreeing elements, such as definite determiners and adjectives (Carstens, 2000). It has often been argued that Dutch grammatical gender seems to be developing from a state that resembles modern German to one that resembles modern English (e.g. Vandeputte, Vincent & Hermans, 1991; Geerts, 1988; Kooij, 1987). Dutch originally had three genders: masculine (de), feminine (de) and neuter (het). The masculine-feminine distinction, however, has gradually collapsed 3 The terms grammatical gender and gender are used interchangeably in this thesis. When natural gender is discussed, as is the case here, it is explicitly mentioned as natural gender. 11 into a single non-neuter group, probably due to the relative size of the de-and het groups. Van Berkum (1996) has examined the Dutch gender ratio and confirms dictionary-based estimates that Dutch has (much) more de-words (75%) than het-words (25%). Depending on how the words are counted, the ratio varies from 2:1 to 3:1 (van Berkum, 1996: 35). This shows that the distribution of de-and het word occurrences is very uneven. Consequently, de has a default status. De occurs even more frequently because it is also used for plural nouns of both genders (Hulk & Cornips, 2006b). Het is also used in a different part of Dutch speech, namely as expletive pronoun (e.g. het regent, it rains) and personal pronoun (e.g. ik zag het zonet nog, I saw it just now) (van Berkum, 1996: 26). However, because of the default status of de and the fact that de is also used for plural nouns, de still occurs much more frequently than het. In spontaneous Dutch speech, at least twice as many de-words occur as hetwords. The targets for Dutch gender agreement are listed in Table 1. Table 1 Agreement targets in the Dutch gender system (taken from van Berkum, 1996: 25) De-words common gender Het-words neuter gender English equivalent definite article de ster het huis the star, house demonstrative pronoun possessive pronoun deze ster die ster onze ster dit huis dat huis ons huis this star, house that star, house our star, house Interrogative pronoun welke ster? welk huis? which star, house? indefinite pronoun elke ster iedere ster menige ster elk huis ieder huis menig huis each star, house every star, house many a star, house relative pronoun de ster die… the star, house that… adjectives in indefinite NPs (een) kleine ster (een) rode ster de rode ster het huis dat… het huis wat… (een) klein huis (een) rood huis het rode huis adjectives in definite NPs (a) small star, house (a) red star, house the red star, house There is no gender distinction in the morphology of the indefinite determiner in Dutch, which is een for both neuter and non-neuter nouns: de ster, het huis (the star, the house) >> een ster, een huis (a star, a house). There is no evidence for grammatical gender on the plural of the definite determiner either, as this is de for all nouns: de sterren, de huizen (the stars, the houses). Cornips & Hulk (in press) note that this causes a very low saliency of grammatical gender on determiners in the Dutch input. Gender is only clearly visible in the opposition between de and het in the singular definite determiner, and between die/deze and dat/dit in the singular demonstrative determiner. The saliency of the contrast is further weakened by the frequency differences between neuter and common nouns. Consequently, the most salient evidence for determiner features in the Dutch input is the opposition definite-indefinite (Cornips & Hulk, in press: 22). There are hardly any morphological and semantic regularities which could point to the gender of nouns in Dutch, and the regularities that exist have many exceptions (Donaldson, 1987: 27-33; Geerts et al., 1984: 41-49). For example, most words for humans are de-words. There is one unambiguous morpho-phonological cue for neuter gender, however, namely the diminutive suffix -je. All diminutives in Dutch are neuter, even when they are derived from a common noun (e.g. het huis – 12 het huisje, the house – the little house; de muis – het muisje, the mouse – the little mouse). The diminutive suffix, which frequently occurs, is the only regular morphological cue for neuter gender in Dutch. Because such regularities hardly occur, (psycho)linguists see Dutch gender assignment as an essentially random affair (e.g. Jescheniak, 1994; Wijnen & Deutsch, 1987; de Houwer, 1987). 2.3 The Acquisition of Grammatical Gender As stated in the previous sections, the relation between a noun and its gender is in principle arbitrary. For this reason, current psycholinguistic models of gender processing are based on the idea that gender is stored in the mental lexicon as an inherent, syntactic property of nouns and looked up in spontaneous speech (Garrett, 1988, 1992; Levelt, 1989, 1999, 2001; Dell, 1986, 1990; Caramazza, 1997; see Schriefers & Jescheniak, 1999 for a recent discussion). Grammatical gender is one of the earliest properties to emerge in first language acquisition and, depending on the language being learned, also one of the earliest to be mastered (Franceschina, 2005: 107). In order to acquire a grammatical gender system, learners need to distinguish the grammatical categories involved in the gender system from other grammatical categories (Carroll, 1999: 45). Learners need to notice, for instance, that the form of the determiner morphologically varies depending on the noun (or the form of the noun) that accompanies it (Carroll, 1989). When learners notice agreement in gender between the noun and the determiner related to it, this may trigger gender awareness and lead to gender acquisition. Children can rely on different types of gender cues when acquiring the gender system of their native language, namely syntactic, morphophonological and semantic cues. Syntactic cues are the forms articles, adjectives, pronouns and passive participles related to a given noun take, morphophonological cues are the endings of nouns (e.g. -a or -o), and semantic cues are the noun’s (in)animacy and sex (Franceschina, 2005: 107). Very young children seem to rely especially on morphophonological cues and to a lesser extent on semantic and syntactic cues (e.g. MacWhinney, 1978; Böhme & Levelt, 1979; Mills, 1986; Karmiloff-Smith, 1979; Cain, Weber-Olsen & Smith, 1987). As children get older, however, they begin to pay more attention to syntactic information. Researchers think that “the strong effect of phonology at the early stages of development may be explained by assuming that [initially] there are no syntactic gender features in the child’s grammar, and until these are activated children are almost exclusively sensitive to morphophonological patterns” (Franceschina, 2005: 110). It is not clear yet why children do not make more extensive use of semantic gender cues. When morphophonological and semantic gender cues are not available, the gender for each noun has to be acquired individually, i.e. on a case by case basis (Unsworth, in press: 5). Thus, gender acquisition boils down to word learning in the absence of cues (Carroll, 1989: 567). Since grammatical gender acquisition largely has to be done through word learning, learners need to have sufficient input to be able to acquire a targetlike grammatical gender system (Unsworth, in press). How much input is needed exactly is not yet known. According to Holmes & Dejean de la Bâtie, “gender acquisition for native speakers can be plausibly explained by purely associative mechanisms” (1999: 480). Unambiguous gender markers, typically definite and indefinite articles (e.g. French la maison, une maison; le couteau, un couteau), 13 are likely to serve as the learner’s major sources of information in this process. This kind of lexical information is probably most important, while sublexical information such as word endings indicating gender (e.g. French la maison, la baguette, le couteau, le monument) will reinforce the strength of gender knowledge in a postaccess procedure of gender confirmation (Holmes & Dejean de la Bâtie, 1999; Desrochers & Paivio, 1990; Desrochers et al., 1989). According to connectionist models, word endings can also influence the gender decision in an earlier stage, i.e. contribute to the gender decision at the time the article is activated (e.g. Taft & Meunier, 1998). Holmes & Segui (2004) found that people paid attention to lexical and sublexical gender cues in parallel. The associative model as proposed by Homes & Dejean de la Bâtie (1999) assumes that all first language acquirers can acquire the lexical and syntactic properties of gender and become targetlike. Indeed, all normal first language learners of grammatical gender become target-like, even though the way in which this happens is dependent on the language being learned. The second language acquisition of grammatical gender, however, shows a very different pattern. Adult second language learners have persistent difficulties with grammatical gender and many studies observe that they fail to become target-like, even after considerable exposure to the target language in question (Carroll, 1989; Dewaele & Veronique, 2001; Bruhn de Garavito & White, 2002; Franceschina, 2005 on Romance languages; Rogers, 1987 on German; Andersson, 1992; Hyltenstam, 1992 on Swedish). Child second language acquirers also have difficulties with grammatical gender (Carroll, 1989; Andersson, 1992; Blom et al., 2006; Hulk & Cornips, 2006a; Unsworth, in press). Importantly, second language learners show signs of fossilisation, as they often seem unable to develop past the overgeneralisation stage (e.g. Franceschina, 2005; Hulk & Cornips, 2006a,b; Cornips et al., 2006). First language learners also overgeneralise (e.g. Clark, 1985; Gillis & De Houwer, 1998), but generally overcome this stage and become target-like. Second language learners experience difficulties with using gender syntactically, i.e. with gender agreement, but also with the relatively ‘simple’ procedure of assigning the correct gender to nouns (Sabourin, 2003: 15). Consequently, both the syntactic and lexical properties of gender are problematic for second language learners. Various proposals have been put forward to explain the difficulties with grammatical gender adult second language learners experience. One of these proposals is the Failed Functional Features Hypothesis (FFFH) (Hawkins & Chan, 1997). The FFFH advances the idea that parameterised functional features cannot be acquired after childhood (post-puberty) unless they are represented in the first language. Parameterised functional features are features that do not substantially contribute to meaning (e.g. Tense), but that do induce morphosyntactic operations (e.g. Agreement) (Blom et al., in press: 3). According to the FFFH, the first language of a second language learner has to instantiate grammatical gender; if this is not the case the learner will experience serious difficulties with gender (see also Carroll, 1989; Franceschina, 2005; Hawkins & Francescina, 2004; Unsworth, in press, Blom et al., in press). Thus, the FFFH states that uninterpretable functional features are subject to critical period effects (Blom et al., in press: 3). Other researchers do not seek an explanation in the presence or absence of the grammatical gender feature in the first language. Bruhn de Garavito & White (2002) point out that problems with 14 gender can also be found in adult second language learners whose first language does not instantiate gender. Moreover, there are adult second language learners with a first language without gender, who are able to acquire gender-marked determiners and adjectives (White et al., 2004). The explanation Bruhn de Garavito & White (2002) propose, following Lardiere (2000) and Prévost & White (2000), is the hypothesis that second language learners of grammatical gender experience a mapping problem between syntax and morphology. Bruhn de Garavito & White state that second language learners of gender may experience “difficulties in relating underlying abstract features to appropriately inflected surface forms” (2002: 170). Brouwer et al. (in press) have, to some extent, found evidence supporting this hypothesis. They report that child second language learners of Dutch grammatical gender show awareness of gender specification in Dutch, but that they seem unable to produce the appropriate morphological form. In line with Van der Velde’s (2004) proposal on the monolingual first language acquisition of Dutch gender, Hulk & Cornips (2006b), Cornips & Hulk (in press) and Brouwer et al. (in press) speculate that the difficulties with gender second language learners experience may be due to an underspecification of the gender feature, i.e. second language learners are aware of gender but do not have a complete knowledge and do not know the right gender specification (yet). This may cause second language learners to overgeneralise. According to Blom et al. (in press), the reason for the difficulties with grammatical gender may be that adult second language acquirers use different learning strategies than children learning their first language. Blom et al. propose that lexicon-driven strategies (learning based on associative memory) remain available for adult learners, whereas grammar is subject to critical period effects. This thesis further investigates the possibility of an underspecification of the Dutch gender feature in second language learners as proposed by Hulk & Cornips (2006b), Cornips & Hulk (in press) and Brouwer et al. (in press). More specifically, a possible cause for such an underspecification is explored. This thesis focuses on the acquisition of the Dutch definite determiners, in particular the comprehension (perception) of Dutch neuter het. As mentioned previously, the Dutch determiner [hεt] varies depending on its phonological context and is more often than not reduced to [ət] and [ə] in informal speech. This makes het non-salient, which may cause perception difficulties in second language learners of Dutch. If second language learners indeed systematically fail to perceive het, they are likely to fail to store het in the mental lexicon. The failure to represent het in the mental lexicon may lead to an underspecification of the gender feature in second language learners, causing them to overgeneralise. The next section gives an overview of previous research on the acquisition of the Dutch definite determiners, culminating in the study by Brouwer et al. (in press). Section 2.5 discusses the phonology of het and de in Dutch informal, connected speech. The way het manifests itself in Dutch informal speech is linked to the underspecification hypothesis as proposed by Hulk & Cornips (2006b), Cornips & Hulk (in press) and Brouwer et al. (in press). 15 2.4 The Acquisition of Dutch Grammatical Gender The monolingual first language acquisition of Dutch grammatical gender is a gradual, long drawn-out process (e.g. Verrips & Wijnen, 1998). Globally speaking, four stages can be discerned in the monolingual acquisition of determiners in Dutch. Like children acquiring other languages, monolingual Dutch children initially produce bare nouns (Chierchia et al., 2001). Subsequently, before the age of two years, children begin to produce a schwa-element /ə/ which resembles the indefinite article een (a/an) (Bol & Kuiken, 1988). Examples of child utterances in the literature show that combinations of definite article and noun are certainly present in the first half of the third year, which is the third stage. Also around this time, children overgeneralise the non-neuter definite determiner de instead of using neuter het (Van Zonneveld, 1992). The Dutch gender system appears to be difficult to learn in comparison with, for example, the gender systems of the Romance languages and German. German and French gender are acquired more quickly, i.e. by the age of three years (Mϋller, 1990; Clark, 1985). By the age of three years gender distinctions in Dutch are not fully acquired yet, and de is still overgeneralised (Gillis & De Houwer, 1998). Van der Velde (2003, 2004) has found that Dutch children between three and six years still overgeneralise de where neuter het is expected, and the target grammar only seems to be in place around age six. This is the final stage four. L1 Dutch children overgeneralise in one direction only, i.e. they use a default (Van Zonneveld, 1992; Gillis & De Houwer, 1998; Van der Velde, 2002, 2003). This also seems to hold for Romance languages (Clark, 1985). For Dutch, this means that de is consistently used instead of het, and not the other way round. Van der Velde (2004) explains the overgeneralisation of de in terms of a dissociation or mapping problem between syntax and morphology. Monolingual first language acquirers of Dutch may already have acquired the uninterpretable gender feature, but it is still underspecified. Van der Velde (2004) hypothesises that children initially adopt the unmarked default value (non-neuter) because the gender feature is still underspecified when they begin to produce definite determiners. As a consequence, they initially overgeneralise de. Later, children gain complete knowledge of the marked value (neuter) and also begin to produce neuter het. Johnson (2004) examined the comprehension of Dutch definite determiners by toddlers aged 26 to 30 months and found an asymmetry on het vs. de trials at this very early age as well. The performance on de-words versus het-words varied; toddlers were insensitive to het-world trials, in contrast to de-word trials. Apparently, Dutch-learning toddlers are only sensitive to common gender. According to Johnson, the asymmetry can partly be explained by the fact that het has more than one grammatical function. As mentioned previously, het is also used as expletive pronoun (e.g. het regent, it rains) and as personal pronoun (e.g. ik zag het zonet nog, I saw it just now). The second language acquisition of Dutch grammatical gender shows both similarities and differences with the first language acquisition of gender in Dutch. As is the case in monolingual acquisition, second language learners have difficulties in acquiring the Dutch definite determiners and they, too, overgeneralise (Hulk & Cornips 2006a,b; Cornips et al., 2006; Blom et al., in press; Unsworth, in press; Cornips & Hulk, in press; Brouwer et al., in press). However, unlike in monolingual first language acquisition, second language learners of Dutch appear to fossilise in such a non- 16 targetlike stage of overgeneralisation (Hulk & Cornips 2006a,b; Cornips et al., 2006; Brouwer et. al, in press; Unsworth, in press). Another difference with monolingual acquisition is that second language learners of Dutch gender make errors in both directions, i.e. they do not exclusively use de as the default (Cornips et al., 2006; Unsworth, in press). In a series of studies, Cornips and Hulk investigate the acquisition of grammatical gender in ethnic community children in The Netherlands (Hulk & Cornips, 2005; Hulk & Cornips 2006a, 2006b; see also Cornips & Hulk, in press). The children in these studies have characteristics of both bilingual and early child second language learners, because the input they receive is often non-targetlike and because it is unclear what the quantity and quality is of the Dutch that these children are exposed to in their earliest years. Bilingual children are children acquiring two languages from birth (2L1 acquisition), and the two grammars emerge simultaneously. In second language acquisition (L2 acquisition) the two grammars emerge successively. Sometimes children acquiring a second language between age 4 and 7 are also called bilinguals, but a clear distinction is made between the two groups in the (generative) literature on acquisition. Hulk & Cornips (2006a) compare bilingual children in three different age groups with age matched monolingual controls. All bilingual children are between 3 and 10 years old. In the study, both monolingual and bilingual children show a development in the correct use of de. Still, the bilingual children show a delay in their development, i.e. there is a quantitative difference between the two emerging grammars. When it comes to the production of het, the bilingual children have severe acquisition difficulties. They seem to have fossilized in a developmental stage where they overgeneralise the non-neuter definite article de. Thus, the data suggest that the bilinguals fossilise in the non-targetlike stage of overgeneralisation. Hulk & Cornips note that the monolinguals also go through a stage of overgeneralisation of de, but that contrary to the bilinguals, they progress beyond this stage. Consequently, there also appears to be a qualitative difference between the two emerging grammars. Hulk & Cornips note that this is a striking result, as previous literature has not reported qualitative differences between mono-and bilinguals (Hulk & Cornips, 2006a). A delay or quantitative difference is not unexpected, since under certain conditions an acceleration or delay is predicted in bilingual acquisition (e.g. Paradis & Genesee, 1995; Hulk & Müller, 2000; Müller & Hulk, 2001). Hulk & Cornips (2006b) examine possible explanations for the qualitative difference between monolingual and bilingual children. They speculate that deficient input may be the cause of the differences. As mentioned previously, the quantitative difference between mono-and bilinguals is not unexpected. Bilingual children are exposed to quantitatively less input than monolingual children, which may be the cause of a delay in the acquisition of het. Anderson (1999) and Umbel & Oller (1995) also note that a reduced input has serious consequences for the lexical development of bilinguals. The qualitative difference, however, is unexpected. In order to explain this difference, Hulk & Cornips argue that bilingual children are not only exposed to a quantitatively different input, but also to a qualitatively different input (e.g. overgeneralisation of de by parents, siblings, etc.) than monolinguals, and that this may explain the difference in attainment between the two groups. Hulk & Cornips (2006b) also propose a purely linguistic analysis. This proposal is in line with Van der Velde’s (2004) theory of underspecification in monolingual first language acquirers of Dutch. 17 Hulk & Cornips state that when monolinguals begin to produce definite determiners, they “have already acquired the [uninterpretable] gender feature, but it is still underspecified. It therefore takes the default value [non-neuter] and is spelled out as de on all definite determiners” (Hulk & Cornips, 2006b: 14). According to Hulk & Cornips, there is no reason to assume that monolinguals and bilinguals differ in this respect. Therefore, it is possible that the bilingual children have acquired the [neuter] gender feature, but continue to use the default form de in spelling it out on the definite determiner because the gender feature is still underspecified, i.e. it is possible that the bilinguals show a dissociation or mapping problem between syntax and morphology (Hulk & Cornips, 2006b). Monolinguals progress beyond this stage, whereas bilinguals possibly remain in this stage and fossilise. Cornips et al. (2006) investigate the acquisition of Dutch grammatical gender (among other things the definite determiners) in older monolingual and bilingual ethnic community children of 10-13 years old and their results show a similar pattern as the previous studies. The age of first exposure to Dutch for these children ranges from birth to 4 years. Interestingly, the children make errors in both directions, i.e. they do not exclusively use de as the default, but het as well. This is different from what is observed in the monolingual first language acquisition of Dutch gender (Van Zonneveld, 1992; Gillis & De Houwer, 1998; Van der Velde, 2002, 2003). Blom et al. (in press) also investigate the acquisition of Dutch grammatical gender by ethnic community children and adults. The children in this study are between 4 and 8 years old and the adults are between 25 and 39 years old. The age of first (substantial) exposure to Dutch for the children is approximately 4 years, while the age of first exposure for the adults is 15 years or older. Blom et al. find a pattern of overgeneralisation of non-neuter articles to neuter contexts, while the reverse infrequently occurs. As mentioned in the previous section, Blom et al. (in press) argue that different learning strategies may account for the difficulties with grammatical gender second language learners experience. Following Ullmann (2001a,b; 2004) and Hawkins & Franceschina (2004), Blom et al. hypothesise that biological age has an influence on learner’s more abstract grammar-driven representations, but not on lexical gender representations (in press: 5). Thus, lexicon-driven strategies remain available for older learners, whereas grammar becomes inaccessible. Unsworth (in press) extends Hulk & Cornips’ (2006a,b) proposal to a different bilingual population. She investigates English-speaking children (5,5 to 17,5 years and 9,5 to 18,5 years) and adults (22 to 50 years) acquiring Dutch as a second language. For the L2 children, age of first exposure ranges from birth to 7 years, for the L2 preteens age of first exposure ranges from 8 to 12 years and for the L2 adults, age of first exposure ranges from 21 to 43 years. Unlike in Hulk & Cornips’ study, the English-speaking children in Unsworth’s study have not been exposed to a qualitatively different input by the community (parents, siblings, etc.), and therefore, Unsworth hypothesises, should not fossilise in the overgeneralisation stage. Unsworth's results show that the acquisition of het is indeed delayed, and that the children overgeneralise de. She notes that this may also be an effect of transfer from the first language, since English has the. The is phonologically similar to de rather than to het. Unsworth also finds that some of the bilinguals use het with common nouns. Unsworth notes that this may be a marker of bilingualism in Dutch. She cannot say yet whether her bilinguals have fossilised in a non-targetlike state of overgeneralisation. There are learners who are targetlike in 18 Unsworth’s study, suggesting that “targetlike acquisition of grammatical gender as marked by the definite determiner is in principle possible by English-speaking children and adults” (Unsworth, in press: 29), but there are many more non-targetlike learners. Overall, Unsworth’s results suggest that lengthy and intensive exposure may lead to more targetlike responses in bilinguals. Although Unsworth cannot say yet how much input is needed exactly, she tentatively suggests that second language learners need to have had at least 12 years of intensive exposure. Sabourin et al. (2006) investigate the adult second language acquisition of Dutch grammatical gender. Native speakers of English, German and a Romance language (French, Italian or Spanish) who learnt Dutch as a second language are investigated. The results show a clear transfer effect, with the following hierarchy in performance: German > Romance > English. This suggests that languages which are similar are easier to learn than languages which differ, and that transfer plays an important role in second language acquisition. Thus, Sabourin et al. claim that the morphological similarity between first-and second language is crucial for the successful acquisition of grammatical gender. In addition, Sabourin et al. find a clear frequency effect. All of the bilinguals produce the targetlike determiner more often for high frequency nouns than for low frequency nouns. Cornips & Hulk (in press) review the factors proposed in the literature that may explain the success or failure in the child second language acquisition of grammatical gender in Dutch definite determiners. These factors are (i) early age of onset (Blom et al., in press); (ii) lengthy and intensive input (Unsworth, in press); (iii) the quality of the input (Hulk & Cornips, 2006b) and (iv) the role of the other language (Sabourin et al., 2006). Cornips & Hulk claim that the first two factors may indeed contribute to an explanation of the differences in success between less and more successful bilingual children. With respect to the other two factors, Hulk & Cornips claim that “the influence of the quality of the input in (standard) Dutch appears to be inconclusive, whereas the (structural) similarity of the gender systems in the two languages may reinforce the children’s awareness of the grammatical gender category” (in press: 2). Hulk & Cornips also claim that “it appears that individual bilingualism versus societal bilingualism, that is the sociolinguistic context in which Dutch is acquired, is not a factor for failure or success with respect to the acquisition of grammatical gender” (in press: 2). Finally, Cornips & Hulk argue that the important role of the input is related to a language-internal factor. The specific difficulties second language acquirers of Dutch grammatical gender experience may be due to a specific factor in the Dutch gender system of the definite determiner which distinguishes it from that of other languages and results in different acquisition paths. This factor is the way Dutch grammatical gender on the determiner is dependent on definiteness, in contrast to the gender systems of Romance languages or other Germanic languages. As mentioned in section 2.2, the only clear evidence for gender on the determiner is the opposition between de and het in the singular definite determiner, and between die/deze and dit/dat in the singular demonstrative determiner. However, the saliency of the contrast is further weakened by the frequency differences between neuter and common nouns. Consequently, the most salient evidence for determiner features in the Dutch input is the opposition definite-indefinite (Cornips & Hulk, in press: 22). Following Hawkins & Franceschina (2004), Cornips & Hulk assume that young monolingual and bilingual children have no gender specification in their Dutch grammar. At a certain point, children become aware of gender, and they add a grammatical gender 19 feature to their grammar. This gender feature remains underspecified, however, and in the stage that follows children may overgeneralise de, but also het. Monolingual children progress beyond this stage, whereas bilinguals possibly remain in this stage and fossilise (Cornips & Hulk, in press). As mentioned previously, the study by Brouwer et al. (in press) is directly relevant to the perception experiment discussed in this thesis. In contrast to the studies mentioned above, which focus on production, Brouwer et al. focus on comprehension in second language learners of Dutch grammatical gender. Brouwer et al. investigate whether the acquisition problems with neuter gender are due to a misrepresentation of the gender feature specification in the grammar of bilingual children (misrepresentation hypothesis), or due to a mapping problem, i.e. difficulties in relating the appropriate surface form to the correct underlying abstract feature (production hypothesis). Brouwer et al. try to find evidence for either one of the hypotheses by carrying out a determiner comprehension task with older bilingual ethnic community children (11-13 years). They do not find evidence for the first hypothesis, but do find some evidence for the second. The children show knowledge of gender specification in Dutch, but this is certainly not good knowledge. In addition, the children seem unable to correctly produce gender and de seems to be used as a default production strategy. Similar to previous studies, Brouwer et al.’s results suggest fossilisation in a non-targetlike stage of acquisition. When combining their results with previous production data, Brouwer et al. hypothesise (in line with Van der Velde, 2004 and Hulk & Cornips, 2006b, also see Cornips & Hulk, in press) that the reason for the overgeneralisation of de in young bilinguals is that they initially do not have a gender specification in their grammar. They are only aware of the definite feature and choose de as a default. At some point they acquire a gender specification for definite determiners, but it remains underspecified. As a result, they also use het for common nouns. Monolinguals progress beyond this stage, in contrast to bilinguals, who fossilise (Brouwer et al., in press). The proposal advanced by Brouwer et al. (in press) is in line with previous research by Sabourin & Haverkort (2003). Sabourin & Haverkort investigate the representation of grammatical knowledge and language processing in adult second language learners of Dutch with German as their native language. Sabourin & Haverkort hypothesise that advanced second language learners may have native-like knowledge, but process this knowledge in a non-native like manner. Sabourin & Haverkort make use of ERPs, and find a qualitative difference between native speakers and second language learners of Dutch in terms of language processing. Sabourin & Haverkort suggest a dissociation between the representation and processing of grammatical knowledge. As the previous research shows, second language learners of Dutch systematically show a delay in the acquisition of the Dutch definite determiners as well as signs of fossilisation in a nontargetlike stage of overgeneralisation. Especially Dutch neuter gender (het) causes problems for second language learners. The question is what causes these problems with the Dutch definite determiners. Various researchers (Hulk & Cornips, 2006b; Cornips & Hulk, in press; Brouwer et al., in press) argue that the difficulties may be due to an underspecification of the gender feature in second language learners. Van der Velde (2004) also argues for an underspecification of the gender feature, 20 in order to explain the difficulties with Dutch definite determiners monolingual first language acquirers experience. In this thesis, the theory of an underspecification of the gender feature in second language learners is further explored. Most of the previous research on the second language acquisition of Dutch definite determiners focuses on production; only the study by Brouwer et al. (in press) is on comprehension. Because the study by Brouwer et al. is the only study to date which focuses on comprehension, the goal of the present study is to gain more insight into the role of comprehension in the second language acquisition of Dutch gender. More specifically, this thesis investigates the perception of the Dutch definite determiners by second language learners. Because especially neuter gender (het) causes problems for second language learners, this thesis focuses on their perception of het. It is possible that second language learners fail to perceive het. If this is the case, this may lead to an underspecification of the gender feature in second language learners, which may cause them to overgeneralise de. As the studies by Cornips et al. (2006) and Unsworth (in press) show, second language learners also overgeneralise het. However, because the overgeneralisation of de to neuter nouns far outweighs the other way round, this thesis focuses on this type of overgeneralisation only. The next section outlines why especially het may easily be missed in perception. This section also discusses the consequences of a systematic misperception of het by second language learners of Dutch. 2.5 The Saliency Hypothesis: The Influence of Phonological Context 2.5.1 Introduction As mentioned in the previous section, Hulk & Cornips (2006b), Cornips & Hulk (in press) and Brouwer et al. (in press) suggest that the cause of the problems with neuter gender in second language learners of Dutch may be an underspecification of the grammatical gender feature. As Van der Velde (2004) points out, monolinguals also seem to go through this stage of underspecification, but progress beyond it and become targetlike. Second language learners, however, possibly remain in this stage and fossilise. In this section, a possible cause of underspecification in second language learners of Dutch gender is proposed. It is argued here, that the cause of underspecification in second language learners of Dutch may be the non-saliency of het as induced by its phonological context. This proposal is put forward as the saliency hypothesis. Het, phonologically [hεt], varies depending on its phonological context in informal speech. As a result of its phonological context, het is more often than not reduced to [ət] and occasionally to [ə] in informal speech. De, on the other hand, mostly remains unreduced and is phonologically salient (see Table 2 below). Moreover, de is the default and de is used for all plural nouns in Dutch. Our research team has checked the way het and de are realised in Dutch speech in the CGN (Corpus Gesproken Nederlands, Spoken Dutch Corpus). The CGN is a collection of spoken standard Dutch comprising of about 9 million words. The results of the procedure are given in Table 2. 21 Table 2 The occurrence of de and het in the CGN Het De Total frequency: 11.395 Total frequency: 31.319 Pronunciation Frequency Percentage Pronunciation Frequency Percentage [hεt] 576 5% [də] 25.668 80% [hət] 556 5% [tə] 3.972 12% [ət] 7.675 67% [ə] 607 2% [əd] 1.395 12% [d] 387 1,2% [t] 325 3% Table 2 shows that the frequency difference between common and neuter nouns as mentioned in section 2.2 is clearly visible in the CGN: de occurs 31.319 times and het occurs only 11.395 times. Importantly, Table 2 reveals that het is hardly ever produced as [hεt]. In only 5 to 10% of the cases is [hεt] produced in its full form, i.e. as [hεt] and as [hət]. In the rest of the cases [hεt] is reduced, mostly to [ət] and [əd]. De, on the other hand, features as [də] in as much as 80% of the cases and as [tə] in 12% of the cases. In the rest of the cases [də] is mostly reduced to [ə]. This means that phonological context has an influence on the pronunciation of both het and de, and that both are subject to reduction. However, het mostly occurs in its reduced (non-salient) form, whereas de mostly occurs in its unreduced (salient) form. The input second language learners are exposed to is largely informal, fast speech in which het is reduced. It is possible that second language learners of Dutch systematically fail to perceive het in everyday speech as a result of the non-saliency of het. Second language learners may hear [də] instead of [hεt] because of the many occurrences of the reduced forms [ət] and [ə] in combination with the default status of [də], the use of [də] for all plural nouns in Dutch and the phonological saliency of [də]. If second language learners indeed fail to perceive het, they are likely to fail to store het in the mental lexicon and only de may be represented in the mental lexicon. The failure to represent het in the mental lexicon may lead to an underspecification of the gender feature in second language learners, causing them to overgeneralise de. A failure to perceive het would be a plausible explanation for an underspecification of the grammatical gender feature in second language learners of Dutch. As mentioned in section 2.2, the saliency of grammatical gender on determiners in the Dutch input is very low (Cornips & Hulk, in press). Gender is only clearly visible (and perceivable) in the opposition between de and het in the singular definite determiner and between die/deze and dat/dit in the singular demonstrative determiner. The saliency of the contrast is further weakened by the frequency differences between neuter and common nouns (Cornips & Hulk, in press: 22). In addition, there are hardly any morphological and semantic regularities which could point to the gender of nouns in Dutch and the regularities that exist have many exceptions (Donaldson, 1987: 27-33; Geerts et al., 1984: 41-49). These circumstances, combined with the phonological non-saliency of het and the phonological saliency of de, are disadvantageous enough to lead to perception difficulties with regard to het. 22 The influence of phonological context on the perception and acquisition of the Dutch definite determiners and the saliency hypothesis as proposed in this thesis have not met with experimental research yet. Therefore, the aim of the present study is to lay a foundation for further research. Sections 2.5.2, 2.5.3 and 2.5.4 serve to illustrate the reduction–and assimilation processes which cause the reduction and assimilation of het and de in Dutch informal, connected speech. The main work of reference here is Booij’s The Phonology of Dutch (1995). In connected speech, phonological rules are often optional, as opposed to the rules of word phonology, which are obligatory (Booij, 1995: 125). The rules of connected speech are to a large extent dependent on the linguistic properties and environments of the segments involved. Booij notes that “it is characteristic of casual speech that ease of production gets priority over ease of perception, because speakers can only afford to give priority to ease of production in more informal situations” (125). As the articulatory efforts of the speaker are reduced, reduction–and assimilation processes occur that apply across (prosodic) word boundaries (Booij, 125-26). 2.5.2 The Cliticisation of Het and De Cliticisation is an integral part of Dutch connected speech. Clitics can be defined as “function words such as pronouns, determiners, auxiliaries, particles, conjunctions, and prepositions which are phonologically dependent on a host word to which they attach, and with which they form a prosodic constituent. In addition, they may also have special syntactic distributional properties” (Booij, 1995: 165). Cliticisation applies to, among other things, the Dutch determiners de, het and een and a number of /d/-initial function words, the latter mostly being demonstratives (die, dit, dat, deze). The strong form of het, [hεt], is used in stressed position, whereas ‘t, [ət], is weak and unstressed.4 De, [də], and een, [ən], have only one, weak form, because their only vowel is a schwa (Booij, 1995: 176). It follows from their phonological form that the weak forms [ət], [də] and [ən] do not form prosodic words of their own but syllables at most, because in Dutch a prosodic word requires at least one syllable with a full vowel (Booij, 1995: 169). The weak form [ət] is phonologically dependent on the preceding or following word in the sentence, its host. Because [ət] does not form a prosodic word of its own, resyllabification takes place according to the universal CV-rule. Example (1) illustrates this process. (1) Jan kocht het boek (Jan bought the book) Syntactically: [Jan]NP [kocht]V [‘t boek]NP Phonologically: [jɑn] [[kɔx] [tət]] [buk] (Booij, 1995: 165) In example (1) the clitic [ət] induces obligatory resyllabification of the preceding word (here kocht). Thus, the coda consonant /t/ of kocht becomes the onset of the next syllable [tət]. 4 Syllables with schwa never bear word stress, and word-stress rules only apply to the prosodic word (Booij, 1995: 170). 23 The syntactic distribution of [ət] in [Jan]NP [kocht]V [‘t boek]NP, is determined by the rules for the construction of Dutch noun phrases, i.e. [ət] syntactically depends on the following noun. Phonologically, however, [ət] is dependent on its host, i.e. the preceding verb in [jɑn] [[kɔx] [tət]] [buk]. Thus, the syntactic structure is non-isomorphic to the prosodic structure (Booij, 1995: 165). The effect of resyllabification is “that after incorporation of the clitic, the prosodic word fulfils all conditions on wellformed prosodic words” (Booij, 1995: 166). The clitic [ət] is also subject to voice assimilation. The (regressive) assimilation of [ət] occurs before voiced plosives. Example (2) illustrates this. (2) Ik zie het beest (I see the beast) Syntactically: [Ik]NP [zie]V [het beest]NP Phonologically: [Ik] [zi] [ədbest] Example (2) shows that the clitic [ət] assimilates into [əd] before the voiced plosive [b] of the following noun. The clitic [əd] is adjoined to the noun. This process facilitates production. Contraction processes also facilitate production. Booij gives the example of /h/-deletion. The /h/ disappears in certain combinations, e.g. in clitic + verb combinations. If /h/-deletion would not take place in such combinations, this would lead to ill-formed consonant clusters. For example, Ik heb (I have) becomes [Ikεp] and Ze heeft (She has) becomes [zeft], because /kh/ and /zh/ are ill-formed.5 Mere deletion of the schwa in [zə] of Ze heeft is not possible, because this would yield the consonant cluster /zh/. This consonant cluster causes severe production difficulties. Not all /h/-initial words are contracted, however: Je houdt (You hold, You keep) cannot be pronounced as *[jɔut] (Booij, 1995: 180). The disappearance of /h/ in contraction is important for the present study, since /h/-deletion can be witnessed in the cliticisation of het. Het changes from [hεt] to [ət] in consonant clusters. Example (3) illustrates this. (3) Ik heb het hem gezegd (lit. I have it him told, i.e. I have told it to him) Syntactically: [Ik]NP [heb]V [het]NP [hem]NP [gezegd]V Phonologically: [Ik] [[[hεpətəm]]] [xəzεxt]; [[[[Ikεpətəm]]]] [xəzεxt] (Booij, 1995: 180). Example (3) shows that the /h/ of [hət] is deleted in order to prevent the consonant cluster /bh/. The /h/s of heb, [hεp], and hem, [həm] are also deleted, to prevent the clusters /kh/ and /th/. The weak forms of the Dutch determiners may be encliticised, provided that a potential host word on the left is present. Encliticisation is “the prosodic integration of function words into a left host” (Booij, 1995: 176). An example is the enclisis of [ət] as shown by example (1) above. When there is no potential host word on the left, however, the weak forms may be adjoined to the following prosodic word. In this case, the word on the right serves as a host and procliticisation takes place. This is also possible when a clitic has more than one potential host word. Booij notes that ”for clitics with more 5 Many Dutch personal pronouns, e.g. ik (I) and zij (she, they) can occur as pronominal clitics (Booij, 1995). 24 than one potential host word, encliticization is optional, and they may also procliticize to the following word” (1995: 177). Booij also notes that Dutch schwa-containing clitic words may be either enclitics or proclitics (1995: 174). Example (2) above illustrates the procliticisation of [ət]. Example (4) illustrates the procliticisation of [də]. (4) Geef me de pen even (Hand me the pen, will you?) Syntactically: [Geef]VP [me]NP [de pεn]NP [even]Adv Phonologically: [xef] [mə] [dəpεn] [evən] Booij observes that “proclitics have the same prosodic status as schwa-containing prefixes, and enclitics have the same prosodic status as suffixes” (1995: 170-71). Dutch appears to prefer enclisis to proclisis, however (Gussenhoven, 1985). Many d-initial function words (such as the demonstratives die, dit, etc.) contain a full vowel, i.e. they are strong forms, and as a consequence do not necessarily require a host word. The weak form [də], however, does need a host word. Example (5) illustrates the encliticisation of [də]. (5) Hij leest de krant (He reads the newspaper) Syntactically: [Hij]NP [leest]V [de krant]NP Phonologically: [hei] [[lestə]] [krɑnt] (Booij, 1995: 177). Example (5) shows that [də] becomes [tə] due to (progressive) voice-assimilation after voiceless obstruents. Another case in which [də] procliticises is when the schwa of [də] is deleted before vowels and the [d] is adjoined to the following word. Example (6) illustrates this. (6) De aarde is een bol (The earth is a globe) Syntactically: [De aarde]NP [is]V [een bol]NP Phonologically: [dardə] [Izən] [bɔl] The phenomenon of schwa-deletion before vowels as illustrated in example (6) is relatively rare in Dutch, however. 25 2.5.3 T-deletion in Het Another phenomenon in Dutch connected speech of relevance to the present study is /t/deletion. T-deletion in Dutch typically occurs in fast speech, but may also occur in more formal speech. The rule of t-deletion is rather complicated and here we can limit ourselves to t-deletion across phrasal boundaries. Booij states: ”Across phrasal boundaries, /t/-deletion is possible if both the preceding and the following consonants are obstruents, less probable after or before a nasal consonant, and even less probable before liquids and glides” (1995: 153). Consequently, /t/-deletion takes place in Wint Piet? (Does Pete win?): [טInpit], but not in Zakt Ria? (Does Ria fail?): *[zɑkrija] (Booij, 1995: 154).6 T-deletion is very important for the present study, because [hεt] ends in /t/. Following Booij (1995), the /t/ of [hεt] may be deleted in informal speech when the preceding and following consonants are obstruents or, in some cases, nasals. In the case of het, the preceding and following consonants belong to the coda of the preceding word and the onset of the following word, since the /h/ of [hεt] disappears in consonant clusters to prevent ill-formedness, thereby yielding the clitic [ət] (see section 2.5.2 above). As a consequence, after /t/-deletion only the schwa of [ət] remains: (7) Zij eet verrukt het taartje op. (lit. She eats delightedly the cake up. i.e. She delightedly eats up the cake). Phonology: [vərʏktətartʃə] (8) Oma rent aangedaan het tehuis uit. (lit. Grandmother runs emotionally the home out of. i.e. Grandmother emotionally runs out of the home). Phonology: [anxədanəʔtəhœys]; [anxədanətəhœys] 7 (9) Hij springt gemakkelijk het perron af. (lit. He jumps easily the platform off. i.e. He easily jumps off the platform) Phonology: [xəmɑkələkətpərɔn]; [xəmɑkələkəʔpərɔn]; [xəmɑkələkəpərɔn] Our research team has concluded that het is especially difficult to perceive in fast speech in contexts such as shown in examples (7) and (8), i.e. in contexts where the following word begins in /t/. In such contexts, the /t/ of het is always deleted as a result of degemination. In addition, /h/-deletion takes place to prevent consonant clusters. Thus, only the schwa [ə] of het remains. Of these contexts, especially the contexts as shown in example (7) render het difficult to perceive. Due to the contraction of [hεt] to mere [ə], het may sound as [tə] after a word-final /t/. As pointed out in section 2.5.2, [tə] 6 For more information on /t/-deletion in Dutch, see Goeman (1999) and Mitterer & Ernestus (2006). Globally, the same claims are made here as in Booij (1995). 7 The glottal stop [ʔ] is not a separate phoneme in Dutch, but is inserted before vowel-initial syllables within words after /a/ and /ə/ and often also at the beginning of a word (Jongenburger & van Heuven, 1991). 26 occurs as an encliticised form of [də]. Because de is the default and de is used for all Dutch plural nouns, this makes the perception of het extra difficult for second language learners. 2.5.4 Summary Table 2 in section 2.5.1 shows that phonological context influences the pronunciation of both het and de, and that both undergo reduction in informal speech. However, het mostly occurs in its reduced (non-salient) form, whereas de mostly occurs in its unreduced (salient) form. It was stated that the phonological non-saliency of het, the phonological saliency of de and the frequency difference between common and neuter nouns (de is the default and de is used for all plural nouns in Dutch) may cause a misperception of het in second language learners. A misperception of het may lead to a failure to represent het in the mental lexicon, which may lead to an underspecification of the grammatical gender feature in second language learners, causing them to overgeneralise de. This proposal is put forward as the saliency hypothesis in section 2.5.1. Sections 2.5.2 and 2.5.3 outline the reduction–and assimilation processes which cause the reduction and assimilation of het and de in Dutch informal, connected speech. The phonological processes which affect the clitics de and het are summarised in Table 3. Table 3 The phonological processes affecting the clitics [də] and [ət] Phon. processes De Het 1. resyllabification - + 2. voice-assimilation + + 3. schwa-deletion before vowels + - 4. /h/-deletion - + 5. /t/-deletion - + Table 3 illustrates that het is affected by all of the processes except 3 (schwa-deletion before vowels). De, on the other hand, is only affected by processes 2 and 3. Dutch native speaker listeners who have knowledge of the phonological processes affecting the clitics [də] and [ət] may supplement the missing information when they are confronted with ambiguous sentences such as presented in section 2.5.2 and 2.5.3. Moreover, their lexical knowledge helps them in this respect. Second language learners, however, generally have fewer lexical knowledge than native speakers and are often unfamiliar with the phonology of Dutch. Consequently, ambiguous sentences are difficult for them. In a perception experiment with nonce nouns such as the current experiment, the lexical knowledge of native speakers and second language learners is of no use, and they can only fall back on their phonological knowledge. 27 3 The Present Study 3.1 Introduction In this chapter the present study is introduced. The present study investigates the role of the non-saliency of het in the acquisition of grammatical gender by second language learners of Dutch. As discussed in section 2.5, the way in which het and de are realised in spontaneous speech is for a large part dependent on the phonological context, which renders het non-salient on a regular basis. These circumstances may cause het to be missed in perception, thus leading to persistent acquisition difficulties. In order to test whether het is perceived despite its non-saliency, our research team has designed a perception experiment. In the experiment, second language learners of Dutch and native speakers of Dutch have to indicate whether they hear de or het. See Chapter 4 for a full outline of the perception experiment. In an additional gender assignment task, our research team also tests the gender production capacities of the second language learners. In this task, the participants have to assign the correct gender (de or het) to given nouns of high and medium frequency. See Chapter 5 for a full outline of the production task. The perception experiment is the main experiment of this study. The production task serves to gain an additional insight into the way the Dutch definite determiners are produced. This chapter is organised as follows. Section 3.2 addresses the research questions and section 3.3 formulates the hypotheses. Section 3.4 states the predictions to the hypotheses. 3.2 Research Questions The research questions follow from the general question of why Dutch neuter gender is so difficult to learn for second language learners (see section 2.4). The main research questions are the following. Research questions (1) Can second language learners of Dutch perceive the difference between het and de in the perception experiment or do they fail to perceive het? Does the performance of the second language learners significantly differ from that of the Dutch native speaker control group? Does phonological context have an influence on perception? (2) Can second language learners of Dutch correctly produce gender (de and het)? That is, do the second language learners of Dutch produce de and het correctly or do they overgeneralise? If they overgeneralise, in what way do they overgeneralise? Do they use de as the default? 28 The study by Unsworth (in press) suggests that lengthy and intensive exposure should lead to more targetlike responses in second language learners of Dutch grammatical gender. Thus, sufficient input is necessary. Unsworth focuses on production. It is possible, however, to extend her proposal to perception. Following Unsworth, a third research question can be posed. (3) Is there evidence for an effect of length and intensity of exposure in the second language learners' perception of het? That is, do second language learners who have had lengthy and intensive exposure to Dutch hear het better than second language learners who have had little exposure to Dutch? 3.3 Hypotheses The research questions in section 3.2 lead to hypotheses. The hypotheses can be formulated as follows. Hypotheses (1) With respect to the first and main research question, I hypothesise that second language learners of Dutch will fail to perceive het on a regular basis. The second language learners will regularly perceive de for het. The performance of the second language learners will significantly differ from that of the Dutch native speaker control group, as the natives will perceive het significantly better than the second language learners. Het will more often be missed in perception than de. Both the second language learners and the Dutch native speakers will perceive de significantly better than het. With respect to the influence of phonological context, the following hierarchy of difficulty can be hypothesised: het will be more difficult to perceive when followed by consonants than when followed by vowels, and more difficult to perceive when followed by /t/ than when followed by other consonants. Both the second language learners and the Dutch control group will especially have difficulties with the t-het-t contexts. This hypothesis is based on the following facts. As discussed in section 2.5, het systematically features as a non-salient form in Dutch spontaneous speech as a result of its phonological context, whereas de mainly occurs in its salient form. As noted in section 4.4.4 below, het systematically features as non-salient [ət] and [ə] in the perception experiment. De occurs in the same phonological contexts as het and may also undergo reduction, e.g. voice assimilation, but de mainly features as salient [də]. The second language learners are expected to be affected by the non-saliency of het, failing to perceive het on a regular basis. Because de is mostly salient, the second language learners are expected to be able to perceive de. The second language learners will perceive de instead of het because of the phonological saliency of de and their knowledge of the frequency difference between common and neuter nouns in 29 Dutch. The Dutch control group will also have difficulties with the perception of het, but they have phonological knowledge of the processes listed in Table 3. Because the second language learners do not have this knowledge, they will not be able to detect the effects of /t/-deletion, etc, and thus perception will be more difficult for them. As pointed out in section 2.5.3, when the consonants preceding and following het are obstruents or nasals, both /h/-and /t/-deletion may occur, only leaving [ə] for het. Moreover, the consonant immediately preceding this [ə] is subject to resyllabification. Taking into account Booij's observation that /t/-deletion is improbable before liquids and glides, our research team assumes that /t/-deletion will certainly not occur before vowels, and that the phonological contrast between [də] and [ət] is maximally clear in this case. When het is followed by a /t/, the /t/ of het is always deleted as a result of degemination. When also /h/-deletion occurs (i.e. in about 90% of the cases, see Table 2, section 2.5.1), only the schwa [ə] of het remains. Of these contexts, especially the t-het-t contexts are difficult to perceive. Due to the contraction of [hεt] to mere [ə], het may sound as [tə] after a word-final /t/. As pointed out in section 2.5.2, [tə] occurs as an encliticised form of [də]. Not only the second language learners, but also the Dutch native speakers will find this extremely confusing. (2) With respect to the second research question, I hypothesise that many of the second language learners of Dutch will show non-targetlike behaviour in the production task. They will mainly overgeneralise de, but they may also use het with common nouns. Learners who have had lengthy and intensive exposure to Dutch may show targetlike behaviour on de and het, however. With lengthy and intensive exposure to Dutch, 12 years or more intensive exposure is meant here. The importance of length of exposure will also become visible in a frequency effect, i.e. the participants will perform better on the high-frequency nouns than on the medium-frequency nouns. This hypothesis is based on the following facts. As discussed in section 2.4, second language learners of Dutch have serious problems with the acquisition of the definite determiners. In production studies, they show a delay as compared to native speakers of Dutch, as well as signs of fossilisation in a non-targetlike stage of acquisition (Hulk & Cornips 2006a,b; Cornips et al., 2006; Cornips & Hulk, in press; Blom et al., in press; Unsworth, in press). Second language learners of Dutch mainly overgeneralise de but they do not exclusively use de as the default, they rather make errors in both directions (Cornips et al., 2006; Unsworth, in press). As noted in section 2.3, the acquisition of gender is for a large part word learning (Carroll, 1989: 567). In the absence of morphophonological and semantic gender cues, the gender for each noun has to be acquired individually (Unsworth, in press). Gender assignment in Dutch is largely random, and as noted in section 2.2, there are hardly any morphological and semantic regularities which could point to the gender of Dutch nouns. Consequently, 30 learners need to have sufficient input to be able to acquire a targetlike grammatical gender system (Unsworth, in press). Anderson (1999) and Umbel & Oller (1995) also note that a reduced input can greatly affect the lexical development of bilinguals. How much input is needed exactly is still unknown. Unsworth (in press) tentatively suggests that second language learners of Dutch need at least 12 years of intensive exposure in order to acquire a targetlike grammatical gender system. The more exposure a learner has had, however, the more targetlike he/she can be expected to perform. Sabourin (2003) has found that second language learners of Dutch produce the targetlike determiner more often for high frequency nouns than for low frequency nouns. (3) With respect to the third, additional research question, I hypothesise that the second language learners who have had lengthy and intensive exposure to Dutch will perceive het better than second language learners who have had little exposure to Dutch. Again, with lengthy and intensive exposure to Dutch, 12 years or more intensive exposure is meant here. This hypothesis is based on the following facts. The study by Unsworth (in press) suggests that lengthy and intensive exposure may lead to more targetlike responses in second language learners of Dutch grammatical gender. Unsworth (in press) suggests that a minimum of 12 years of intensive exposure is needed. Thus, sufficient input appears to be crucial for the acquisition of a targetlike grammatical gender system (Unsworth, in press). Unsworth’s study focuses on lexical development and production. It is possible (and likely) that a lengthy and intensive exposure to Dutch also leads to a better perception of Dutch neuter gender (het), since in order to be able to produce (neuter) gender one first ought to perceive (neuter) gender. The following should also be considered. The previous studies show that second language learners of Dutch systematically overgeneralise de and have problems with the acquisition of het (Hulk & Cornips 2006a,b; Cornips et al., 2006; Blom et al., in press; Unsworth, in press; Cornips & Hulk, in press; Brouwer et al., in press). As pointed out above, het systematically features as a non-salient form in Dutch. De, on the other hand, has a default status, is used for plural nouns and is phonologically salient. As a consequence, second language learners of Dutch have very little exposure to neuter het in everyday speech, while exposure to de is frequent. The overgeneralisation of de and the problems with het could be the result of frequent exposure to de as opposed to little exposure to het, i.e. exposure could be the crucial factor. This would also imply that if second language learners hear enough exemplars of het in non-salient contexts (however 'enough' may be), they should, in principle, be able to acquire this aspect of Dutch. 31 3.4 Predictions Predictions to the hypotheses in section 3.3 can be made. The predictions can be formulated as follows. Predictions (1) With respect to the first hypothesis, I predict that the second language learners of Dutch will systematically fail to perceive het in the perception experiment, perceiving de instead. Because the second language learners will systematically choose de when in fact het occurs, in a sense, they can be said to ‘overgeneneralise’ de in perception. The Dutch native speaker control group will perceive het significantly better than the second language learners. Het will more often be missed in perception than de, and both the second language learners and the Dutch control group will perceive de significantly better than het. With respect to the influence of phonological context, the following hierarchy of difficulty can be predicted: het will be more difficult to perceive when followed by consonants than when followed by vowels, and more difficult to perceive when followed by /t/ than when followed by other consonants. Both the second language learners and the Dutch control group will especially have difficulties with the t-het-t contexts. (2) With respect to the second hypothesis, I predict that the second language learners of Dutch will mainly overgeneralise de. They will show signs of using het with common nouns as well, however. The learners who have had long and intensive exposure (i.e. 12 years or more intensive exposure to Dutch) will produce the targetlike determiner significantly more often than learners who have had little exposure. This will be the case for both de and het. The participants will also produce the high frequency nouns significantly better than the low frequency nouns. (3) With respect to the third hypothesis, I predict that the second language learners who have had lengthy and intensive exposure (i.e. 12 years or more intensive exposure to Dutch) will perceive het significantly better than those second language learners who have had little exposure to Dutch. Thus, the learners who have had lengthy and intensive exposure will choose de less often when in fact het occurs. There will be a significant difference in performance between the two groups. The hypotheses and predictions as stated in this chapter will be tested by means of a perception experiment and a production task. Chapter 4 describes the perception experiment and Chapter 5 describes the production task. 32 4 The Perception Experiment 4.1 Introduction As discussed in Chapter 2, the acquisition of grammatical gender has proven to be seriously problematic for second language learners (Carroll, 1989; Dewaele & Veronique, 2001; Bruhn de Garavito & White, 2002; Franceschina, 2005 on Romance languages; Rogers, 1987 on German; Andersson, 1992; Hyltenstam, 1992 on Swedish). This is also the case for the acquisition of Dutch grammatical gender. Second language learners of Dutch show a delay as compared to monolingual first language acquirers of Dutch, and even show signs of fossilisation in the non-targetlike stage of overgeneralisation (Hulk & Cornips 2006a,b; Cornips et al., 2006; Brouwer et al., in press; Cornips & Hulk, in press; Blom et al., in press; Unsworth, in press). In line with Van der Velde (2004), who focuses on monolingual first language acquisition, Hulk & Cornips (2006b), Cornips & Hulk (in press) and Brouwer et al. (in press) propose that the reason for second language learners' difficulties with Dutch grammatical gender may be an underspecification of the gender feature in second language learners, i.e. they are aware of gender but do not have a complete knowledge and do not know the right gender specification (yet). In this thesis, it is proposed that an underspecification of the gender feature may be caused by the non-saliency of het (saliency hypothesis). As discussed in section 2.5, het, phonologically [hεt], varies depending on its phonological context in informal, connected speech (see Booij, 1995). As a result of its phonological context, het is more often than not reduced to [ət] and occasionally to [ə] in informal speech. These circumstances render het difficult to perceive. In addition, de, phonologically [də], has a default status, [də] is used for all plural nouns in Dutch and [də] is phonologically salient. A possible failure to perceive het is likely to lead to a failure to store het in the mental lexicon, which may lead to an underspecification of the gender feature in second language learners. In this chapter, the saliency hypothesis as proposed in Chapter 2 is tested by means of a perception experiment. Our research team has designed an experiment which tests whether het is perceived despite its non-saliency. This chapter is organised as follows. Section 4.2 discusses the questionnaires presented to the second language learners and the Dutch native speaker control group. Section 4.3 introduces the participants taking part in the experiment. Section 4.4 describes the experimental design and section 4.5 outlines the method. Section 4.6 outlines the goal of the C-test. Section 4.7 presents the results. 33 4.2 The Questionnaire 4.2.1 The Second Language Learners Using a questionnaire is standardised procedure in linguistic experimenting. In his handbook Questionnaires in Second Language Research: Construction, Administration and Processing (2003) Dörnyei observes that The popularity of questionnaires in data-collection for second language research is due to their easy construction, versatility, and quick processing by means of either quantitative or qualitative methods. Indeed, the frequency of use of self-report questionnaires in the L2 field is surpassed only by language proficiency tests (back cover). Our research team has constructed a questionnaire for the second language learners of Dutch gender. This questionnaire is used to collect various data. Many personal characteristics of the second language learners are of importance for the current study, for example their ages, other second languages they have learnt, their lengths of stay in The Netherlands, the amount of exposure to Dutch they have received, etc. The questionnaire is provided in Appendix 5. In section A of the questionnaire general information is elicited such as name, gender, date of birth, country of birth, the year in which the participants have moved to The Netherlands and their highest education. Section B is on language proficiency. The participants are asked about their mother tongue, other languages they speak beside English and Dutch and the age at which they have started to learn these other languages. They are also asked whether they have followed Dutch language lessons and, if they have, at what level. In section C the participants are asked about their daily language use (Dutch, English) in various situations (at home, at school, at work, etc.). Section D is the production task, and section E features a language proficiency test (C-test). The production task is fully discussed in Chapter 5. The participants were asked to fill in sections A, B and C of the questionnaire before the perception experiment was started. The production task and the C-test, however, were given afterwards. Immediately after the participants had filled in the questionnaire, the perception experiment was started. The questionnaire took the subjects approximately 5 minutes. The time spent depended on individual speed. 4.2.2 The Dutch Native Speaker Control Group Our research team has also constructed a questionnaire for the Dutch native speaker control group. This questionnaire asks about the age, gender, data of birth, mother tongue and education of the Dutch native speakers and is provided in Appendix 6. The Dutch native speaker participants were asked to fill in the questionnaire before the perception experiment was started. 34 4.3 Participants Thirty-four second language learners of Dutch with English as their native language and fifteen native Dutch speaker control subjects participated in the perception experiment.8 The reasons why our research team decided to test native speakers of English is that English does not have grammatical gender and that this will yield data which can be compared to earlier obtained production data (see e.g. Unsworth, in press). The group of second language learners consists of males (N = 15) and females (N = 19). The Dutch control group also consists of males (N = 7) and females (N = 8). The second language learners are all adults, i.e. all ages are above 18 years. The bulk of the Dutch control group also consists of adults, but one teenager of 16 years is included. The ages of the participants are given in Table 4. Table 4 Ages of the participants (in years and months) Second language learners (N = 34) Dutch control group (N = 15) Minimum 19;5 15;11 Maximum 62;10 57;0 Mean 39;0 32;6 Std. Deviation 11;4 13;3 All participants, the English native speakers as well as the Dutch native speakers, have been raised monolingually. Bilingual subjects have been excluded.9 All English native speakers have learnt Dutch as a second language, i.e. after the age of about 10 years, which is generally considered to be after the critical period for language acquisition. This idea is based on Lenneberg (1964, 1967), who states that learning a language natively becomes difficult after a certain age, certainly after the onset of puberty (see also Hyltenstam & Abrahamsson, 2003 for a recent discussion). Age of first exposure to Dutch is given in Table 5. Table 5 The second language learners' ages of first exposure to Dutch (in years and months) Age of first exposure to Dutch Minimum 8,10 Maximum 41;6 Mean 24;3 Std. Deviation 9;0 Many of the English native speakers can speak several other languages besides Dutch, mostly French and German. However, all participants have learnt these other languages as a second language, i.e. after the age of about 10 years (the bottom line of age of first exposure to French and German is 11;0 years for both languages). Moreover, most of the second language learners who have 8 Initially, 35 second language learners of Dutch would be tested instead of 34, but it turned out that one participant had been raised bilingually, i.e. partly in German. This participant was excluded from the experiment. 9 When testing bilinguals all of the languages they have been raised in should be taken into account. A person raised both in English and German has knowledge of a gender system similar to the Dutch gender system. This could have an influence on their perception and production of Dutch grammatical gender. 35 knowledge of other languages besides Dutch have not learnt these languages extensively, often only in secondary school. The second language learners’ lengths of stay in The Netherlands are provided in Table 6. Table 6 The second language learners’ lengths of stay in The Netherlands (in years and months) Length of stay Minimum 0;4 Maximum 31;0 Mean 12;8 Std. Deviation 9;3 Table 7 shows the second language learners spread over groups with respect to their lengths of stay in The Netherlands. Table 7 The second language learners' lengths of stay in The Netherlands (in years and months): groups Groups Length of stay Number of second language learners Group 1 Less than 12 years 18 (Range: 0;4 years - 8;10 years) Group 2 More than 12 years 16 (Range: 12;0 years - 31;0 years) This division is based on Unsworth (in press), who suggests that second language learners of Dutch grammatical gender need a minimum of 12 years of intensive exposure in order to become targetlike. For some of the second language learners, their lengths of exposure to Dutch differ considerably from their lengths of stay in The Netherlands. These participants have married a Dutch person in their country of birth and/or have followed Dutch language lessons prior to their coming to The Netherlands. The second language learners’ lengths of exposure to Dutch are given in Table 8. Table 8 The second language learners’ lengths of exposure to Dutch (in years and months) Length of exposure Minimum 3;6 Maximum 30;6 Mean 14;9 Std. Deviation 9;3 Table 9 below shows the second language learners spread over groups with respect to their lengths of exposure to Dutch. Again, this division is based on Unsworth (in press), who suggests that second language learners of Dutch grammatical gender need to have had at least 12 years of intensive exposure in order to become targetlike. 36 Table 9 The second language learner's lengths of exposure to Dutch (in years and months): groups Groups Group 1 Length of exposure Number of second language learners Less than 12 years 15 (Range: 3;6 years - 10;5 years) Group 2 More than 12 years 19 (Range: 12;0 years - 30;6 years) It is important to make the distinction between length of stay in The Netherlands and length of exposure to Dutch, because three individuals who have only resided in The Netherlands for a short time (i.e. 0;4 years, 0;9 years and 3;4 years respectively) turn out to have had exposure to Dutch for as long as 17;10 years, 30;0 years and 18;2 years respectively. These individuals are included in Group 2 in Table 9. As the ages, lengths of stay in The Netherlands and lengths of exposure to Dutch of the second language learners show, there is a great variability among the participants. There is also a great variability in their daily occupations, but nearly all participants are highly educated: they are university teachers, PhD students, students in university, consulate and embassy employees, etc. All participants reside in the urbanised west of The Netherlands called the Randstad in Dutch. They have all come to The Netherlands for a permanent or prolonged stay. The data collected in Section C of the questionnaire reveal that all participants use Dutch on a daily basis, often alongside English. The monolingual Dutch control subjects also reside in the Randstad. They are mostly family and friends of the researchers, and they, too, for the large part, are highly educated. All subjects, both monolinguals and second language learners, are of normal mental capacity and do not show signs of linguistic impairments. They also have normal hearing capacity. 4.4 Experimental Design: Stimuli 4.4.1 Introduction The chosen task is a computerised audio-visual identification task. The participants hear 108 Dutch test sentences in which nonce nouns occur. For each sentence, the nonce noun in that sentence appears on the screen. The nonce nouns in the sentences are linked to determiners, and manner adverbs are placed immediately before the determiners. Thus, the phonological contexts of the determiners are formed by the final phonemes of the manner adverbs and the initial phonemes of the nonce nouns. The participants have to indicate which determiner they hear and how certain they are of their answer. Examples 10 to 13 show some of the sentences used in the experiment. (10) Zij verlaat ingezakt het tef en kijkt niet meer om. (lit. She leaves bent the tef and does look not anymore back. i.e. She leaves the tef with her back bent and does not look back anymore.) 37 (11) Hij test nieuwsgierig het ies uit op zijn werk. (lit. He tests curiously the ies out at work. i.e. He curiously tests out the ies at work.) (12) Bang voor infecties, riep zij angstig de tieg terug. (lit. Afraid of infections, called she frightened the tieg back. i.e. She was afraid of infections, and called back the tieg.) (13) Suzanne kuste stiekem de ump tijdens wiskundeles. (lit. Suzanne kissed stealthily the ump during math class. i.e. Suzanne stealthily kissed the ump during math class.) In Chapter 3, section 3.3, it is hypothesised that het will be more difficult to perceive when followed by consonants than when followed by vowels, and more difficult to perceive when followed by /t/ than when followed by other consonants. In order to test this hypothesis, the consonants and vowels as mentioned in sections 4.3.2 and 4.3.3 below are used as phonological contexts. Note that de occurs in the same contexts as het and may also undergo reduction, e.g. voice assimilation.The phonological contexts used are a representative imitation of the natural contexts in which the determiners may occur. Section 4.3.2 describes the conditions on the nonce nouns and section 4.3.3 outlines the conditions on the manner adverbs. Section 4.3.4 provides examples of the cliticisation of het and de in the experiment. Section 4.3.5 describes the fillers and section 4.3.6 gives a schematic and a numerical distribution of the nonce nouns across the contexts. 4.4.2 The Nonce Nouns Nonce nouns are used in the experiment because the focus of the experiment is on the perception of the definite determiners only. Using existing nouns might influence the perception of the second language learners because of their lexical knowledge. As previously noted in section 2.5.3, Booij states that /t/-deletion across phrasal boundaries is possible if both the preceding and the following consonants are obstruents. T-deletion is also possible, but less probable, after or before a nasal consonant. T-deletion is even less probable before liquids and glides (Booij, 1995: 153). Following Booij, it can be assumed that the /t/ of [hεt] is deleted when the preceding and the following consonants are obstruents or, to a lesser extent, nasal consonants. Therefore, our research team has chosen to use nonce nouns with the initial stops /p/, /t/ and /k/. In addition, our research team has chosen to use nonce nouns with initial vowels, in order to create contexts in which the phonological difference between de and het should be maximally clear. Taking into account Booij's observation that /t/-deletion is improbable before liquids and glides, our research team assumes that t-deletion will certainly not occur before vowels. The initial phonemes of the nonce 38 nouns have been chosen in cooperation with Prof. Dr. René Kager of the Utrecht Institute of Linguistics OTS (Utrecht University), who is specialised in phonology. Thus, the plosives /p/, /t/ and /k/ are used as the initial phonemes of the nonce nouns, as well as the vowels /a/, /ɑ/, /e/, /ε/, /ø/, /I/, /i/, /εi/, /ɔ/, /u/, /o/, /ɔu/, /œy/, /ʏ/ and /y/. This is the complete inventory of Dutch vowels, monophthongs as well as diphthongs, except for /ə/ (schwa) because this is not a full vowel. For a complete inventory see Booij (1995). Because het is especially difficult to perceive in fast speech when the following word begins in /t/ (e.g. in t-het-t contexts, see section 2.5.3), the bulk of the nonce nouns with initial plosives in the perception experiment has initial /t/. The number of nonce nouns used for each context (initial /t/, /k/, /p/ and vowel) is given in Table 10. Table 10 The distribution of the nonce nouns across the sentences initial /t/ initial /k/ initial /p/ initial vowel total 35 18 18 37 108 Initially, a more logical distribution of the nonce nouns was planned, namely 36 nonce nouns with initial /t/, 18 nonce nouns with initial /k/, 18 nonce nouns with initial /p/ and 36 nonce nouns with initial vowel. Due to a typing error, however, while programming the experiment in PRAAT, tirf unfortunately lost its /t/ and became irf. As a consequence, one additional nonce noun with an initial vowel was created, and one less with initial /t/. A full overview of the constructed nonce nouns is given in Appendix 1. The nonce nouns designed for the experiment are all monosyllabic nouns with CVC and CVCC (plosive-initial, e.g. taaf, kacht, pank) and VC and VCC (vowel-initial, e.g. aag, eft) syllable structures. All nonce nouns were checked with a reliable online dictionary (van Dale) to make sure that they are indeed nonce nouns and not, accidentally, existent nouns. In addition, the lexical neighbourhoods of the nonce nouns were checked in the Dutch CELEX lexicon (Baayen, Piepenbrock & Van Rijn, 1993) to see whether they (closely) resemble existent nouns, thereby possibly causing gender bias. Certain nonce nouns indeed resemble existing nouns: taaf, for example, resembles taak (task) and taal (language). Most lexical neighbours are de-words, simply because there are more dethan het-words in Dutch. Because it is not certain that participants will think of existing words when they hear, for example, taaf, the nonce nouns have not been changed. Instead, our research team has decided to run a post hoc test if the results prove to be conspicuously odd. As an additional measure to test the reliability of the nonce nouns, the list of nouns was given to a group of 33 native speakers of Dutch who do not take part in the experiment as controls. These native speakers were asked to indicate their gender bias for each nonce noun. The results of this procedure show a strong preference for de (87,04 % de vs. 12,96% het). This was to be expected, because de is the default in Dutch. The findings clearly point to a default strategy and reflect the distribution of de vs. het words in the language as a whole. If the results had been very different, showing, for example, a strong preference for het in the native speakers, this might also be the case for the second language learners of Dutch. Such a gender bias could influence the perception of the 39 second language learners. However, the results of this small gender elicitation task show nothing extraordinary, which also suggests that the nonce nouns can be used in the experiment. 4.4.3 The Manner Adverbs As pointed out in section 2.5.2, the weak form [ət] is phonologically dependent on the preceding word in the sentence, its host. As its phonology shows, [ət] cannot form a prosodic word of its own, because [ət] does not contain a full vowel. The clitic [ət] is phonologically dependent on its host, and induces obligatory resyllabification of the preceding or following word. To further the cliticisation and resyllabification of het in the perception experiment, suitable phonemes have been placed at the end of the manner adverbs before the determiners. As noted in sections 2.5.3 and 4.3.2 above, the /t/ of [hεt] may be deleted when the preceding and the following consonants are obstruents or, in some cases, nasal consonants (Booij, 1995: 153). Therefore, the phonemes used for the endings of the manner adverbs are /t/, /k/, /m/, /n/ and /x/. These phonemes further the enclisis of [ət], which results in resyllabification: [-tət], [-kət], [-mət], [-nət] and [-xət]. Thus, the manner adverbs serve as hosts for the clitics. As with the choice of the initial phonemes of the nonce nouns, the final phonemes of the manner adverbs have been chosen in cooperation with Prof. Dr. René Kager of the Utrecht Institute of Linguistics OTS (Utrecht University). As noted in section 2.5.3 above, het is especially difficult to perceive in fast speech when the preceding word ends in /t/ and the following word begins in /t/ (t-het-t context). Therefore, relatively many of the manner adverbs used in the experiment have final /t/. The number of manner adverbs used for each context (final /t/, /k/, /m/, /n/ and /x/) is represented in Table 11. Table 11 The distribution of the manner adverbs across the sentences final /t/ final /k/ final /m/ final /n/ final /x/ total 36 18 18 18 18 108 A full overview of the constructed manner adverbs and their English translation is given in Appendix 2. 4.4.4 The Cliticisation of Het and De in the Experiment Examples of the test sentences with het and de are given below. All of the sentences feature in the experiment. The examples show how the final phonemes of the manner adverbs and the initial phonemes of the nonce nouns elicit the cliticisation of het and de in the experiment. (14) Zij verlaat ingezakt het tef en kijkt niet meer om. (lit. She leaves bent the tef and does look not anymore back. i.e. She leaves the tef with her back bent and does not look back anymore.) Phonology: [Inxəzɑktətεf] 40 (15) Hij ving behendig het tiemp op, zodat er geen schade was. (lit. He caught dexterously the tiemp, so there no damage was. i.e. He dexterously caught the tiemp, thereby preventing damage.) Phonology: [bəhεndəxətimp] (16) Veel leerlingen bestuderen moeizaam het purt voor vrijdag. (lit. Many students study laboriously the purt for Friday. i.e. Many students laboriously study the purt for Friday.) Phonology: [mujzamətpʏrt]; [mujzaməʔpʏrt]; [mujzaməpʏrt] (17) Bang voor infecties, riep zij angstig de tieg terug. (lit. Afraid of infections, called she frightened the tieg back. i.e. She was afraid of infections, and called back the tieg.) Phonology: [ɑŋstəx] [dətix] (18) Zij lachte zeer aanstekelijk de tilk toe. (lit. She laughed very infectiously the tilk at. i.e. She laughed very infectiously at the tilk.) Phonology: [anstekələk] [dətIlk]; [anstekələk] [tətIlk] (19) Hij begon verhit de kacht te slaan. (lit. He began heatedly the kacht to hit. i.e. He heatedly began to hit the kacht.) Phonology: [vərhItəkɑxt] The examples show that the consonant immediately preceding [ət] is often subject to resyllabification, and that [də] sometimes assimilates into [tə]. When constructing the test sentences, our research team made sure that only one determiner occurs in every sentence. Relative clauses (het meisje dat; the girl that), expletive pronouns (het regent; it rains) and personal pronouns (ik zag het zonet nog; I saw it just now) were avoided. A full overview of the test sentences, which all have the same overall syntax with the nonce nouns placed in object position, is given in Appendix 3. English translations of the test sentences are also provided. As a final measure, to check whether [hεt] is indeed reduced in the test sentences, the sentences were read out loud by our research team. It was found that in natural speech [hεt] is indeed reduced to [ət] or [ə] in all of the test sentences featuring het. De, however, systematically features as [də] and occasionally as [tə] due to assimilation. 41 4.4.5 The Fillers As fillers, our research team has used the gender-specific demonstratives deze, die, dit and dat (see section 2.2. on the Dutch grammatical gender system) in the perception experiment. The demonstratives are also used to check whether the second language learners of Dutch have a problem with Dutch gender as such or not. It is possible that second language learners have a general problem with Dutch gender. We do not expect this to be so, however, because a test with the forms [ət] and [ə] causes more ambiguity than a test with e.g. [d It] and [dɑt]. Example sentences with the fillers are provided below. All of the example sentences feature in the experiment. (20) Op Kerstavond zingen zij vreedzaam dat uuk toe. (lit. On Christmas Eve sing they peacefully that uuk for. i.e. On Christmas Eve they peacefully sing for that uuk.) Phonology: [vretzam] [dɑt] [ʔyk] (21) Een visagist bracht geleidelijk deze polt aan op haar gezicht. (lit. A make-up artist applied gradually this polt to her face. i.e. A make-up artist gradually applied this polt to her face.) Phonology: [xəlɛidəlʏk] [dezə] [pɔlt]; [xəlɛidəlʏk] [tezə] [pɔlt] (22) Wij proberen bewust die arf te negeren. (lit. We try deliberately that arf to ignore. i.e. We deliberately try to ignore that arf.) Phonology: [bəwʏsti] [ʔɑrf] (23) Moeder komt verrukt dit tarp binnen en gaat alles bekijken. (lit. Mother comes enrapturedly this tarp into and begins to everything examine. i.e. Mother enters this tarp in an enraptured state and begins to examine everything.) Phonology: [vərʏktIt] [ʔtɑrp]; [vərʏktItɑrp] (24) Voor zijn afstuderen heeft Michel enthousiast dit ift ontworpen. (lit. For his graduation has Michel enthusiastically this ift designed. i.e. Michel has enthusiastically designed this ift for his graduation.) Phonology: [ɛntusjɑstIt] [ʔIft]; [ɛntusjɑstItIft] The examples above show that, in some cases, the demonstratives undergo assimilation. It is rather assimilation than cliticisation that is witnessed here, because (as pointed out in section 2.5.2) the demonstratives die, deze, dit and dat all contain a full vowel, i.e. they are strong forms, and as a 42 consequence do not necessarily require a host word. As with the sentences with het and de, our research team has read out loud the sentences with the demonstratives. It was found that, despite occasional assimilation, the demonstratives remain clearly audible. The following section outlines the distribution of the demonstratives across the sentences in the experiment. 4.4.6 The Distribution of the Nonce Nouns Across the Contexts The determiners are randomly linked to the nonce nouns in the experiment. A schematic overview of the distribution of the nonce nouns across the contexts is given in Table 12. A numerical overview of the distribution of the nonce nouns across the contexts is given in Table 13 below. Table 12 Schematic distribution of the nonce nouns across the contexts Adverb -x Nonce noun De Voc iens Het ties Dat ijn Deze ocht Die ief Dit ilg K koem P T kijn kecht pir tieg tiemp pirk toft teuk perg tuul tilp -k -m Voc irf ilk oem oerk olk irk K kuft kurp P pelk puig polt T tilk tirs tir tils Voc oup unk uuk upt oos kolm oes kump toem ump K kuig kuun P porf purt purg T turp tulg tuup tuuf tung turs -n Voc olp omp ulp org ulg uim K kuum P pijt T tolg koft kolg puuk pung tolm tups tomp tolp aag ak euk eep arf eft eps eus icht kacht kag pank tuip -t Voc K ers ieg ift kilg keup kijp picht peef penk taar terk kieg P paaf parp T taat taaf tauk tacht techt tift taus tef tarp taam 43 Table 13 Numerical distribution of the nonce nouns across the contexts Adverb Nonce noun De Het -x Voc 2 1 Def. Det. 3 K 1 1 2 P T Total -k Die 1 1 1 1 1 2 3 1 1 2 Dit Demonstr. Total 1 3 6 1 1 3 1 2 3 1 3 6 3 2 9 18 1 1 3 7 1 2 3 1 3 1 4 5 9 2 2 2 4 1 K 1 1 1 P 1 1 2 T 1 1 2 1 1 1 3 5 5 4 9 2 2 2 3 9 18 2 1 3 1 1 1 3 6 1 1 1 1 2 3 Voc K 1 1 3 1 1 3 6 2 1 9 18 1 3 6 1 3 2 3 3 6 P 1 1 2 1 T 1 2 3 1 4 5 9 4 2 Voc 1 2 3 1 1 K 1 1 2 P 1 T 2 1 3 1 1 1 Total -n Deze Voc Total -m Dat Total 1 1 1 1 5 4 9 2 3 2 2 9 18 Voc 3 3 6 1 2 1 2 6 12 K 2 1 3 1 1 1 3 6 P 1 2 3 1 1 1 3 6 T 3 3 6 2 1 2 1 6 12 Total 9 9 18 5 5 4 4 18 36 TOTAL 27 27 54 15 14 13 12 54 108 -t As Table 13 shows, each context is used equally frequently for de and het: 27 nonce nouns are linked to de and 27 to het. The distribution of the nonce nouns across the demonstratives dat, deze, die and dit is 15, 14, 13 and 12 respectively. Initially, the distribution was planned as 14 nonce nouns with dat, 14 with deze, 13 with die and 13 with dit. However, a typing error while programming the experiment in PRAAT ensued in one extra nonce noun with dat and one nonce noun fewer with dit. 44 4.5 Procedure: Method The experiment was run in PRAAT (Boersma & Weenink, 2007). Two screenshots of the experiment are provided below. Screenshot 1 Screenshot 2 45 First the test sentences were recorded with a native speaker of Dutch. The speaker was Dr. Hugo Quené, linguist and phonetician at Utrecht University. He was chosen as the speaker because of his voice (easy to listen to), his speech (clear, standard Dutch) and his experience as a speaker for similar linguistic experiments. Dr. Quené was informed beforehand about the way in which the sentences and the determiners in the sentences should be pronounced, i.e. naturally and not formally. Dr. Quené was not informed of the goal of the experiment. After the recording, our research team double checked the sentences on completeness, correctness and pronunciation. Subsequently, they were linked to PRAAT and the experiment was constructed. The experiment was run on a laptop (hp pavilion ze4500) and all subjects received written instructions from the experimenter immediately before they started the experiment.10 This instruction sheet is provided in Appendix 4. The experiment took place in a quiet (where possible soundproof) room and the participants wore headphones.11 Trial order was randomised in the experiment, i.e. every subject heard the sentences in a different order. This was done to prevent order effects. Per trial, subjects saw a nonce noun appear at the top of the screen (e.g. taaf, koem), which occurred in the test sentence they simultaneously heard through their headphones. The subjects then had to indicate whether they had heard de taaf or het taaf, deze koem or dat koem, etc. In the case of de/het taaf they could indicate their choice by clicking on a yellow bar featuring de or a yellow bar featuring het. This is shown in Screenshot 1 above. Immediately after they had made their choice, their answer became red (instead of yellow). The subjects then had to indicate how certain they were of their answer. They could indicate this at the bottom of the screen, where a large yellow bar featured the numbers 1 to 5. Number 1 stood for uncertain and 5 for certain, which was made explicit on the screen. Numbers 2, 3 and 4 stood for ‘quite uncertain’, ‘not certain, not uncertain’ and ‘quite certain’ respectively. This was assumed to be clear, especially after the subjects had read the instructions, and was therefore not made explicit on the screen. The certainty measure is shown in Screenshot 2 above.12 Subjects could click as soon as they heard the noun and its determiner; they did not have to wait until the sentence was over. They could not leave sentences unanswered, however, because an answer had to be given (gender choice as well as certainty measure) before the next sentence started. Two times during the experiment, after trial 36 and 72 of 108, subjects could take a short break if they wished to. This option was programmed into the experiment and was made explicit on the screen. Subjects could also choose to proceed if they did not wish a break. Immediately before the experiment was started, all subjects did a small ‘practice experiment’ consisting of three sentences. This was done to make sure that all subjects knew exactly what to do. The practice experiment was like the real experiment in every way, but the three test sentences did 10 It is standardised procedure that the voice of an experimenter is not used in a experiment with bilinguals, as sociolinguistic reasons prevent this. 11 The experiment was largely run in the cabins of the Utrecht Institute of Linguistics OTS (Utrecht University) which are specially designed for phonetic experiments. These cabins are soundproof. However, sometimes testing on location was necessary because subjects could not come to the UiL-OTS. In these cases, a quiet room in which the subjects were not disturbed was always provided for. 12 Due to a lack of time and means the results of the certainty measure are not analysed and discussed in this thesis. This thesis would become too long when such an analysis would be included. 46 not occur in the real experiment. The real experiment immediately followed the practice experiment and took the subjects approximately 15-20 minutes. The time spent depended on individual speed. 4.6 The C-Test In addition, in order to quickly test the Dutch proficiency of the second language learners, they are given a language proficiency test called C-test. The C-test is a tool that serves to measure proficiency in a quick way. It is a short text in which parts of words are deleted. The first sentence is left intact in order to provide context. Individuals taking the C-test are asked to fill in the gaps. The subjects’ (in)capability to fill in the gaps gives an insight into their overall proficiency. The C-test used for this experiment consists of three parts (i.e. three short texts) provided by Prof. Dr. René Kager of the Utrecht Institute of Linguistics OTS (Utrecht University). The C-test is provided in Appendix 5 in the questionnaire under Vul het woorddeel in (Fill in the gap), section E. Brief written instructions were included immediately before the C-test. The participants were asked to fill in the gaps by deriving the answers from the contexts of the words. The participants filled in the C-test at the very end of the testing process, i.e. after they had filled in the production task. The C-test took the participants approximately 5 minutes. Again, the time spent depended on individual speed. 4.7 Results 4.7.1 Introduction The results of the perception experiment were analysed with SPSS, version 14.0. The significance level α used is the standard significance level of 0.05 (5%). This means that the reliability of the tests is 95% (100% - 5%). This section is organised as follows. Section 4.7.2 outlines the various factors that should be observed with respect to the second language learners' performance on the perception task. Sections 4.7.3 and 4.7.4 discuss the participants' performances on de, het and the demonstratives (fillers) and investigate the effect of the various factors. Section 4.7.3 discusses and compares the general results per group and gives a summary of the results. Section 4.7.4 investigates the effect of the various factors on the group results and gives a summary of the results. Section 4.7.5 investigates the influence of phonological context on the perception of het. Section 4.7.5.1 discusses the contexts before het, section 4.7.5.2 discusses the contexts after het, section 4.7.5.3 discusses the het-stop contexts, section 4.7.5.4 discusses the full contexts of het and section 4.7.5.5 discusses the stop-hetstop contexts. In every section, the factors age of first exposure and proficiency are controlled for. Every section gives a summary of the results. Section 4.7.6 gives an analysis of the individual results. Section 4.7.7 gives a general summary of the results for performance on the perception task. 47 4.7.2 Factors to Observe There are a number of factors which interact with the performance of the second language learners and which should therefore be controlled for when analysing performance on the perception task. These factors are length of exposure to Dutch, intensity of exposure to Dutch, age of first exposure to Dutch and general proficiency in Dutch. The factors length of exposure and intensity of exposure are mentioned in Chapter 3 and are observed in the hypotheses and predictions in sections 3.3. and 3.4. The factor age of first exposure should also be controlled for. As Table 5 in section 4.3 shows, the second language learners do not form a homogeneous sample in this respect: age of first exposure ranges from 8;10 years to 41;6 years. The factor proficiency should also be controlled for, because it is very unlikely that all second language learners have the same proficiency level. It is widely known in second language acquisition research that proficiency is a good predictor of performance. In order to be able to control for these factors, the second language learners are subdivided into two groups for each factor. The groups are outlined below. Length of Exposure The 34 second language learners are subdivided into two groups for length of exposure to Dutch. As the cut-off point 12;0 years of exposure has been chosen. This is based on Unsworth (in press), who suggests that second language learners of Dutch need to have had at least 12 years of exposure in order to acquire a targetlike grammatical gender system. It should be noted that Unsworth states that 12 years of intensive exposure may be needed. Nevertheless, the cut-off point of 12 years is used here to subdivide the second language learners into groups. Group LE1 has had less than 12 years of exposure (medium exposure), whereas group LE2 has had more than 12 years of exposure (lengthy exposure). The groups are represented in Table 14. Table 14 Subdivision into groups for length of exposure (in years and months) Group N Minimum Maximum Mean Std. Deviation LE1 15 3;6 10;5 6;2 2;2 LE2 19 12;0 30;6 21;6 6;6 Because group LE2 has had more exposure to Dutch than group LE1, group LE2 is expected to perform better on the test than group LE1. 48 Intensity of Exposure The 34 second language learners are subdivided into two groups for intensity of exposure. This was done in the following way. In section C of the questionnaire the participants were asked about their daily language use (Dutch, English) in various situations (at home, at school, at work, etc.). For all of these situations, the participants could choose from the options Always English (0% Dutch), Mostly English/Sometimes Dutch (25% Dutch), English and Dutch (50% Dutch), Mostly Dutch/Sometimes English (75% Dutch) and Always Dutch (100% Dutch). This information has been used to determine the second language learners' intensity of exposure to Dutch. Table 15 below shows the various situations in which the second language learners use Dutch and the degree to which this happens. Table 15 Use of Dutch in various situations: intensity of exposure (Mean in percentages) N Minimum Maximum Mean Std. Deviation Reading books Various Situations 34 0 75 29.4 21.9 At home 33 0 100 38.6 34.3 Watching TV 32 0 75 39.1 21.0 At work 33 0 100 56.1 35.4 With friends 33 0 100 62.1 23.5 Listening to the radio 30 0 100 74.2 31.8 Reading newspapers 34 25 100 75.7 25.7 First encounter 34 0 100 82.4 25.8 At school 10 50 100 82.5 20.6 In a shop 34 75 100 97.8 7.2 Total 34 33 86 62.3 13.0 Table 15 shows that the second language learners especially use Dutch in a shop, at school, when meeting someone for the first time, when reading newspapers and when listening to the radio. They predominantly use English in more personal situations, i.e. when reading books, when at home, when watching TV, when at work and when meeting with friends. Two groups are created on the basis of the total mean use of Dutch (62.3 %), which serves as the cut-off point. Group IE1 is the group which has had average exposure to Dutch and group IE2 is the group which has had more than average or intensive exposure to Dutch. The groups are represented in Table 16. Table 16 Subdivision into groups for intensity of exposure (use of Dutch in percentages) Group N Minimum Maximum Mean Std. Deviation IE1 IE2 18 16 33 64 60 86 52.1 73.7 6.9 7.5 Because group IE2 has had more than average or intensive exposure, while group IE1 has had average exposure, group IE2 is expected to perform better on the test than group IE1. 49 Age of First Exposure The 34 second language learners are subdivided into two groups for age of first exposure to Dutch. The cut-off point has been set at 18;0 years. The reason for this, is that an age of 18;0 years marks the onset of adulthood. Thus, the second language learners are subdivided into a group for whom age of first exposure lies before adulthood (puberty learners), and a group for whom age of first exposure lies in adulthood (adult learners). This is done because of the claims of various researchers, who state that the end of the purported critical period for language acquisition can be extended to the onset of adulthood, i.e. to the end of puberty (see Hyltenstam & Abrahamsson, 2003 for a recent discussion). Group AE1 are the puberty learners and group AE2 are the adult learners. The groups are represented in Table 17. Table 17 Subdivision into groups for age of first exposure (in years and months) Group N Minimum Maximum Mean Std. Deviation AE1 7 8;10 17;11 12;4 3;10 AE2 27 18;0 41;6 27;3 7;2 Following those who claim that the critical period does not end until puberty has ended, i.e. until the onset of adulthood, group AE1 should perform better on the test than group AE2 because they have not reached adulthood yet. Proficiency The second language learners are subdivided into two groups for their general Dutch proficiency level. This was done on the basis of their scores on the C-test. Table 18 shows the second language learners' mean scores for performance on the three separate parts of the test as well as the mean score for performance on the test as a whole. Table 18 The second language learners' scores on the C-test (Mean in percentages correct) C-Test N Minimum Maximum Mean Std. Deviation Part 1 34 30 100 84.9 16.6 Part 2 34 0 100 76.6 26.7 Part 3 34 0 100 67.7 29.7 Whole C-Test 34 20 100 76.4 22.0 Two groups are created on the basis of the total mean score on the C-test (76.4% correct), which serves as the cut-off point. Group P1 is the group with an average proficiency level and group P2 is the group with a more than average or high proficiency level. The groups are represented in Table 19. 50 Table 19 Subdivision into groups for proficiency (Mean C-test scores in percentages correct) Group N Minimum Maximum Mean Std. Deviation P1 14 P2 20 20 75 55.3 18.9 82 100 91.3 5.7 Because group P2 has a higher proficiency level than group P1, group P2 is expected to perform better on the test than group P1. 4.7.3 Group Results 4.7.3.1 General Results Tables 20, 21 and 22 show the performance of the Dutch control group and the second language learner group on the perception task. Table 20 Perception results per group (Mean in percentages correct) Dutch control group (N = 15) Second language learners (N = 34) De 99.5 (SD = 1.4) 97.4 (SD = 6.5) Het 94.6 (SD = 5.3) 83.7 (SD = 16.9) All Determiners (de, het ) 96.9 (SD = 2.8) 90.5 (SD = 10.5) Non-neuter demonstratives (deze, die) 99.8 (SD = 0.9) 98.3 (SD = 9.4) Neuter demonstratives (dit, dat) 99.7 (SD = 1.0) 96.4 (SD = 8.8) All Demonstratives (deze, die, dit, dat) 99.7 (SD = 0.7) 97.3 (SD = 8.7) Whole Test (all test items) 98.4 (SD = 1.5) 93.9 (SD = 9.1) Table 21 Range of performance (Mean in percentages correct) of the Dutch control group (N = 15) Minimum Maximum Range Median De 96 Het 85 100 4 100 100 15 96 All Determiners (de and het) Non-neuter demonstratives 93 100 7 98 97 100 4 100 Neuter demonstratives 96 100 4 100 All Demonstratives 98 100 4 100 Whole Test (all test items) 95 100 5 98.3 Table 22 Range of performance (Mean in percentages correct) of the second language learners (N = 34) Minimum Maximum Range Median De 63 100 37 100 Het 26 100 74 89 All Determiners (de and het ) 54 100 46 93.5 Non-neuter demonstratives 45 100 55 100 Neuter demonstratives 55 100 46 100 All Demonstratives 50 100 50 100 Whole Test (all test items) 52 100 48 96.8 51 In order to determine whether there are significant differences in performance within and between the two groups, (performance within) the groups can be compared statistically. Before this can be done, however, it should be determined whether test performance shows an even or uneven distribution. It was found that the second language learner group shows an uneven distribution for performance on the separate test categories and performance on the test as a whole (KolmogorovSmirnov and Shapiro-Wilk: p < 0.05). The Dutch control group shows an even distribution for performance on the test as a whole (Kolmogorov-Smirnov: p > 0.05, Shapiro-Wilk: p < 0.05), but uneven distributions for performance on the separate test categories (Kolmogorov-Smirnov and Shapiro-Wilk: p < 0.05). Because the two groups show uneven distributions for performance and the number of Dutch control subjects is < 30, non-parametric tests are used to compare (performance within) the groups. The Wilcoxon signed-ranks test is used for comparisons within groups and the Mann-Whitney test is used for comparisons between groups. The Kruskal-Wallis test is used for comparisons between more than two groups. 4.7.3.2 The Dutch Control Group Table 23 shows that no significant differences are found within the Dutch control group with respect to performance on de vs. het, the determiners vs. the demonstratives and the non-neuter demonstratives vs. the neuter demonstratives. Table 23 Perception results of the Dutch control group (N = 15) De vs. Het Z = -.343; p = .732 Determiners vs. Demonstratives Z = -.531; p = .596 Non-neuter Demonstratives vs. Neuter Demonstratives Z = .000; p = 1.000 4.7.3.3 The Second Language Learners Table 24 shows that no significant differences are found within the group of second language learners with respect to performance on de vs. het, the determiners vs. the demonstratives and the non-neuter demonstratives vs. the neuter demonstratives. Table 24 Perception results of the second language learners (N = 34) De vs. Het Z = -.099; p = .922 Determiners vs. Demonstratives Z = -.196; p = .844 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -.508; p = .612 52 4.7.3.4 Comparison of the Groups This section compares the performance of the Dutch control group and the second language learners. It should be observed that the results of only 31 second language learner participants (34 3) are used for this analysis. In order to obtain a homogeneous sample which can be compared to the Dutch control group, three outliers are excluded from the group of second language learners. When exploring the scores on the perception test as a whole for all 34 second language learner participants, two extremes and one outlier are found. The two extremes have total test scores of 52% and 74% correct respectively, and the outlier has a total test score of 85% correct. All other second language learners have a total test score of more than 85% correct. The extremes and the outlier are represented in Figure 1 below. The extremes (participants 14 and 31) are represented by a star ( * ) and the outlier (participant 34) is represented by a circle ( o ). Figure 1 Boxplot of the second language learners' total test scores (1.00 = 100% correct) Two extremes ( * ) and one outlier ( o ) N = 34 1,00 0,90 34 TotalTest 0,80 31 0,70 0,60 14 0,50 English MotherTongue In order to determine whether the 3 outliers in Figure 1 are indeed outliers, a statistical test can be done. For this thesis, use has been made of a Kolmogorov-Smirnov test via a web-interface (Kirkman, 1996: Kolmogorov-Smirnov test). According to this test, the outliers in Figure 1 are indeed outliers (Kirkman, 1996: outcome Kolmogorov-Smirnov test). This means that the second language learners who have total test scores of 52%, 74% and 85% correct respectively are outliers. 53 Figure 2 below represents the total test scores of the Dutch native speaker control group. Figure 2 Boxplot of the Dutch native speaker control group's total test scores (1.00 = 100% correct) No extremes, no outliers N = 15 1,00 0,90 TotalTest 0,80 0,70 0,60 0,50 Dutch MotherTongue Figure 2 does not show any extremes or outliers. An additional Kolmogorov-Smirnov test via the webinterface (Kirkman, 1996: Kolmogorov-Smirnov test) confirms this: no outliers are found. Because the Dutch native speaker control group does not show any extremes or outliers, the results of all 15 Dutch native speaker controls are used. All Dutch controls have a total test score of 95% correct or higher. Table 25 shows the results of the comparison of the groups analysis. The second language learners (N = 31) are compared to the Dutch control group (N = 15). In sections 3.3 and 3.4 above it was hypothesised and predicted that the second language learners will especially have difficulties with het when het occurs in the t-het-t contexts (t-het-tef, t-hettaaf, t-het-techt). In order to check whether the differences in performance between the groups are significant when the t-het-t contexts are excluded from the analyses, performance on het without the thet-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts have also been investigated. 54 Table 25 Comparison of the groups: the Dutch control group vs. the second language learners Mean in percentages correct for each group and outcome of the Mann-Whitney test Significant differences are shaded Dutch control group (N = 15) Second language learners (N = 31) Dutch control group vs. second language learners De 99.5 (SD = 1.4) 98.7 (SD = 2.3) Z = -1.214; p = .225 Het 94.6 (SD = 5.3) 87.8 (SD = 10.0) Z = -2.315; p = .021 Het without t-het-t 98.4 (SD = 2.0) 91.6 (SD = 9.2) Z = -2.981; p = .003 Determiners (de, het ) 96.9 (SD = 2.8) 93.2 (SD = 5.4) Z = -2.368; p = .018 Determiners without t-het-t 98.9 (SD = 1.3) 95.1 (SD = 5.2) Z = -2.933; p = .003 Deze 99.5 (SD = 1.8) 100 (SD = 0.0) Z = -1.438; p = .151 Die 100 (SD = 0.0) 99.7 (SD = 1.4) Z = -.696; p = .487 Dit 99.5 (SD = 2.1) 99.0 (SD = 2.7) Z = -.630; p = .529 Dat 100 (SD = 0.0) 98.0 (SD = 3.6) Z = -2.137; p = .033 Non-neuter demonstratives 99.8 (SD = 0.9) 99.9 (SD = 0.7) Z = -.497; p = .619 Neuter demonstratives 99.7 (SD = 1.0) 98.5 (SD = 2.4) Z = -1.835; p = .067 Demonstratives 99.7 (SD = 0.7) 99.1 (SD = 1.5) Z = -1.435; p = .151 Whole Test (all test items) 98.4 (SD = 1.5) 96.2 (SD = 3.2) Z = -2.195; p = .028 Whole Test without t-het-t 99.3 (SD = 0.7) 97.1 (SD = 3.1) Z = -2.878; p = .004 Table 25 shows that the Dutch control group performs better on all test categories except deze and the non-neuter demonstratives taken together. There are significant differences between the groups for performance on het (with and without the t-het-t contexts), the determiners (with and without the thet-t contexts), dat and the test as a whole (with and without the t-het-t contexts). In all cases, the Dutch control group performs significantly better than the second language learner group. The fact that the differences remain significant when the t-het-t contexts are excluded indicates that there are also other contexts of het which cause difficulties for the second language learners. 4.7.3.5 Summary The results of the comparison of the groups analysis tell us a number of things. According to the test results, the second language learners have more difficulties than the Dutch control group with the perception of het, the determiners and the test as a whole. This is also the case when the t-het-t contexts are excluded from the analysis. The fact that the differences remain significant when the thet-t contexts are excluded from the analysis indicates that there are also other contexts of het which cause difficulties for the second language learners. Since our focus is on the perception of het, the second language learners' difficulties with het are very important. The second language learners also have more difficulties than the Dutch control group with the perception of dat. This is striking, and a possible explanation for this is discussed in Chapter 7. The second language learners perform at native level on de, deze, die, dit, the non-neuter demonstratives, the neuter demonstratives and the demonstratives taken together. 55 No significant differences have been found within the Dutch control group and the second language learner group with respect to performance on de vs. het, performance on the determiners vs. the demonstratives and performance on the non-neuter demonstratives vs. the neuter demonstratives. 4.7.4 Group Results per Factor This section analyses the group results per factor. Again, the Wilcoxon signed-ranks test is used for comparisons within groups and the Mann Whitney test is used for comparisons between groups. The Kruskal-Wallis test is used for comparisons between more than two groups. This section is organised as follows. Section 4.7.4.1 discusses the results for the factor age of first exposure, section 4.7.4.2 discusses the results for the factor length of exposure, section 4.7.4.3 discusses the results for the factor intensity of exposure and section 4.7.4.4 discusses the results for the factor proficiency. Section 4.7.4.5 discusses the correlation between the factors and performance, and section 4.7.4.6 gives a summary of the group results per factor. 4.7.4.1 Age of First Exposure General Results As noted in section 4.7.2, the second language learners have been subdivided into two groups for age of first exposure: groups AE1 and AE2. Group AE1 are the early acquirers of Dutch (before 18 years of age; puberty learners) and group AE2 are the later acquirers (after 18 years of age; adult learners). Table 26 shows the performance of the two groups. Table 26 Group results for AE1 and AE2 (Mean in percentages correct) AE1 (N = 7) AE2 (N = 27) De 99.4 (SD = 1.5) 96.8 (SD = 7.2) Het 94.7 (SD = 3.5) 80.8 (SD = 17.9) All Determiners (de, het ) 97.0 (SD = 2.2) 88.8 (SD = 11.2) Non-neuter demonstratives (deze, die) 100 (SD = 0.0) 97.8 (SD = 10.6) Neuter demonstratives (dit, dat) 99.5 (SD = 1.3) 95.6 (SD = 9.8) All Demonstratives (deze, die, dit, dat) 99.7 (SD = 0.8) 96.6 (SD = 9.7) Whole Test (all test items) 98.4 (SD = 1.2) 92.8 (SD = 9.9) Group AE1 Table 27 shows no significant differences within group AE1 with respect to performance on the various test categories. Table 27 Results of group AE1 (N = 7) De vs. Het Z = -1.876; p = .061 Determiners vs. Demonstratives Z = -1.018; p = .309 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -1.265; p = .206 56 Group AE2 Table 28 shows no significant differences within group AE2 with respect to performance on the various test categories. Table 28 Results of group AE2 (N = 27) De vs. Het Z = -.598; p = .550 Determiners vs. Demonstratives Z = -.830; p = .406 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -1.028; p = .304 Comparison of the Groups This section compares the performance of groups AE1, AE2 and the Dutch control group. It should be observed that the results of only 31 second language learner participants (34 - 3) are used for this analysis. The 3 outliers mentioned in section 4.7.3 are excluded from the analysis. They are all part of group AE2 and, consequently, group AE2 consists of only 24 second language learners (27 3). Again, performance on het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts are also looked at. Table 29 shows the results of the comparison of the groups analysis. Table 29 Comparison of the groups: the Dutch control group vs. group AE1 vs. group AE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Dutch control group (N = 15) Group AE1 (N = 7) Group AE2 (N = 24) De 99.5 (SD = 1.4) 99.4 (SD = 1.5) 98.4 (SD = 2.4) Comparison of the groups (df = 2) χ2 = 2.651; p = .266 Het 94.6 (SD = 5.3) 94.7 (SD = 3.5) 85.8 (SD = 10.5) χ2 = 10.440; p = .005 Het without t-het-t 98.4 (SD = 2.0) 97.1 (SD = 3.0) 90.0 (SD = 9.8) χ2 = 12.211; p = .002 Determiners (de, het ) 96.9 (SD = 2.8) 97.0 (SD = 2.2) 92.1 (SD = 5.6) χ2 = 11.121; p = .004 Determiners without t-het-t 98.9 (SD = 1.3) 98.3 (SD = 2.1) 94.2 (SD = 5.5) χ2 = 12.994; p = .002 Deze 99.5 (SD = 1.8) 100 (SD = 0.0) 100 (SD = 0.0) χ2 = 2.067; p = .356 Die 100 (SD = 0.0) 100 (SD = 0.0) 99.7 (SD = 1.6) χ2 = .917; p = .632 Dit 99.5 (SD = 2.1) 100 (SD = 0.0) 98.7 (SD = 3.0) χ2 = 1.917; p = .383 Dat 100 (SD = 0.0) 99.0 (SD = 2.6) 97.7 (SD = 3.8) χ2 = 5.424; p = .066 Non-neuter demonstratives 99.8 (SD = 0.9) 100 (SD = 0.0) 99.8 (SD = 0.8) χ2 = .478; p = .787 Neuter demonstratives 99.7 (SD = 1.0) 99.5 (SD = 1.3) 98.2 (SD = 2.6) χ2 = 5.212; p = .074 Demonstratives 99.7 (SD = 0.7) 99.7 (SD = 0.8) 98.9 (SD = 1.7) χ2 = 3.743; p = .154 Whole Test (all test items) 98.4 (SD = 1.5) 98.4 (SD = 1.2) 95.6 (SD = 3.4) χ2 = 10.054; p = .007 Whole Test without t-het-t 99.3 (SD = 0.7) 99.0 (SD = 1.0) 96.5 (SD = 3.3) χ2 = 12.179; p = .002 57 Table 29 shows significant differences between the groups for performance on het (with and without the t-het-t contexts), the determiners (with and without the t-het-t contexts) and the test as a whole (with and without the t-het-t contexts). This only tells us that there is variation among the groups for performance on these test categories. Table 30 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). The standard significance level α used for the Mann Whitney tests needs to be adjusted here because of the multiple comparisons made in the Kruskal-Wallis test. When adjusting the standard significance level by means of a Bonferroni adjustment, the new significance level becomes 0.0167 (0.05 : 3). When the Bonferroni adjustment is taken into account, the shaded differences in Table 30 are significant. Table 30 Comparison of the groups (Dutch control group, group AE1 and group AE2) in pairs Significant differences are shaded (Bonferroni adjustment) AE1 vs. AE2 AE1 vs. Dutch control group AE2 vs. Dutch control group Het Z = -2.345; p = .019 Z = -.109; p = .913 Z = -2.805; p = .005 Het without t-het-t Z = -1.891; p = .059 Z = -.964; p = .335 Z = -3.292; p = .001 Determiners (de, het ) Z = -2.438; p = .015 Z = -.072; p = .942 Z = -2.886; p = .004 Determiners without t-het-t Z = -2.091; p = .037 Z = -.588; p = .556 Z = -3.371; p = .001 Whole Test (all test items) Z = -2.351; p = .019 Z = -.036; p = .971 Z = -2.719; p = .007 Whole Test without t-het-t Z = -2.038; p = .042 Z = -.731; p = .465 Z = -3.247; p = .001 Table 30 shows that there is a significant difference between groups AE1 and AE2 for performance on the determiners as a whole. Group AE1 perceives the determiners significantly better than group AE2. The difference in performance on the determiners without the t-het-t contexts is not significant, however. For this reason, the significant difference in performance on the determiners appears to be solely due to the inclusion of the t-het-t contexts. There are no significant differences in performance between group AE1 and the Dutch control group. This means that group AE1 performs at native level, perceiving the test categories equally well as the native speakers. There are significant differences between group AE2 and the Dutch control group for performance on het (with and without the t-het-t contexts), the determiners (with and without the t-het-t contexts) and the test as a whole (with and without the t-het-t contexts). In all cases, the Dutch control group performs significantly better than group AE2. The fact that the differences remain significant when the t-het-t contexts are excluded indicates that there are also other contexts of het which cause difficulties for group AE2. When observing the difference in performance between groups AE1 and AE2 when they are compared to the Dutch control group, it can be concluded that age of first exposure appears to have an effect on performance on het, the determiners and the test as a whole (all with and without the thet-t contexts). The puberty learners (group AE1) perform better on these categories than the adult learners (group AE2), because they perform at native level and the adult learners do not. Thus, earlier acquisition seems to lead to a better perception of het, the determiners and the test as a whole (all with and without the t-het-t contexts). This is also suggested by the fact that group AE1 perceives the 58 determiners significantly better than group AE2. The difference between groups AE1 and AE2 could also be an effect of length of exposure, however. This factor is discussed in the next section. 4.7.4.2 Length of Exposure General Results As noted in section 4.7.2, the second language learners have been subdivided into two groups for length of exposure: groups LE1 and LE2. Group LE1 are the learners who have had exposure to Dutch for less than 12 years (medium exposure) and group LE2 are the learners who have had exposure to Dutch for 12 years or more (lengthy exposure). Table 31 shows the performance of the two groups. Table 31 Group results for LE1 and LE2 (Mean in percentages correct) LE1 (N = 15) LE2 (N = 19) De 98.2 (SD = 2.4) 96.7 (SD = 8.5) Het 83.5 (SD = 18.5) 83.8 (SD = 16.1) All Determiners (de, het ) 90.8 (SD = 9.6) 90.2 (SD = 11.5) Non-neuter demonstratives (deze, die) 99.7 (SD = 1.0) 97.1 (SD = 12.6) Neuter demonstratives (dit, dat) 96.6 (SD = 6.7) 96.3 (SD = 10.4) All Demonstratives (deze, die, dit, dat) 98.1 (SD = 3.5) 96.6 (SD = 11.4) Whole Test (all test items) 94.5 (SD = 6.4) 93.5 (SD = 10.9) Group LE1 Table 32 shows no significant differences within group LE1 with respect to performance on the various test categories. Table 32 Results of group LE1 (N = 15) De vs. Het Z = -.227; p = .820 Determiners vs. Demonstratives Z = -.245; p = .807 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -.870; p = .384 Group LE2 Table 33 shows no significant differences within group LE2 with respect to performance on the various test categories. Table 33 Results of group LE2 (N = 19) De vs. Het Z = -.110; p = .913 Determiners vs. Demonstratives Z = -.592; p = .554 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -.251; p = .802 59 Comparison of the Groups This section compares the performance of groups LE1, LE2 and the Dutch control group. The results of only 31 second language learners (34 - 3) are used for this analysis, as the 3 outliers are excluded. One of the outliers is part of group LE1 and two are part of group LE2. Consequently, group LE1 consists of only 14 second language learners (15 - 1) and group LE2 of only 17 second language learners (19 - 2). Again, performance on het without the t-het-t contexts, the determiners without the thet-t contexts and the test as a whole without the t-het-t contexts are also looked at. Table 34 shows the results of the comparison of the groups analysis. Table 34 Comparison of the groups: the Dutch control group vs. group LE1 vs. group LE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Dutch control group (N = 15) Group LE1 (N = 14) Group LE2 (N = 17) De 99.5 (SD = 1.4) 98.4 (SD = 2.4) 98.9 (SD = 2.2) Comparison of the groups (df = 2) χ2 = 2.057; p = .357 Het 94.6 (SD = 5.3) 87.6 (SD = 9.9) 88.0 (SD = 10.5) χ2 = 5.456; p = .065 Het without t-het-t 98.4 (SD = 2.0) 92.1 (SD = 8.4) 91.2 (SD = 10.1) χ2 = 8.914; p = .012 Determiners (de, het ) 96.9 (SD = 2.8) 92.9 (SD = 5.1) 93.4 (SD = 5.8) χ2 = 5.871; p = .053 Determiners without t-het-t 98.9 (SD = 1.3) 95.4 (SD = 4.2) 94.8 (SD = 6.1) χ2 = 8.619; p = .013 Deze 99.5 (SD = 1.8) 100 (SD = 0.0) 100 (SD = 0.0) χ2 = 2.067; p = .356 Die 100 (SD = 0.0) 99.4 (SD = 2.1) 100 (SD = 0.0) χ2 = 2.286; p = .319 Dit 99.5 (SD = 2.1) 99.4 (SD = 2.1) 98.6 (SD = 3.1) χ2 = 1.252; p = .535 Dat 100 (SD = 0.0) 97.0 (SD = 3.6) 98.8 (SD = 3.5) χ2 = 9.064; p = .011 Non-neuter demonstratives 99.8 (SD = 0.9) 99.7 (SD = 1.1) 100 (SD = 0.0) χ2 = 1.209; p = .546 Neuter demonstratives 99.7 (SD = 1.0) 98.2 (SD = 2.4) 98.7 (SD = 2.5) χ2 = 4.118; p = .128 Demonstratives 99.7 (SD = 0.7) 98.9 (SD = 1.7) 99.1 (SD = 1.4) χ2 = 3.222; p = .200 Whole Test (all test items) 98.4 (SD = 1.5) 96.0 (SD = 3.1) 96.4 (SD = 3.4) χ2 = 5.434; p = .066 Whole Test without t-het-t 99.3 (SD = 0.7) 97.1 (SD = 2.6) 97.1 (SD = 3.6) χ2 = 8.494; p = .014 Table 34 shows significant differences between the groups for performance on het without the t-het-t contexts, the determiners without the t-het-t contexts, dat and the test as a whole without the t-het-t contexts. Table 35 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 35 Comparison of the groups (Dutch control group, group LE1 and group LE2) in pairs Significant differences are shaded (Bonferroni adjustment) LE1 vs. LE2 LE1 vs. Dutch control group LE2 vs. Dutch control group Het without t-het-t Z = -.163; p = .871 Z = -2.645; p = .008 Z = -2.618; p = .009 Determiners without t-het-t Z = -.162; p = .872 Z = -2.536; p = .011 Z = -2.619; p = .009 Dat Z = -1.771; p = .077 Z = -2.798; p = .005 Z = -1.350; p = .177 Whole Test without t-het-t Z = -.443; p = .658 Z = -2.686; p = .007 Z = -2.370; p = .018 60 Table 35 shows no significant differences between groups LE1 and LE2 for performance on the separate test categories and the test as a whole. There are significant differences between group LE1 and the Dutch control group for performance on all of the test categories. On all of the test categories, the Dutch control group performs significantly better than group LE1. There are also significant differences between group LE2 and the Dutch control group for performance on het without t-het-t and the determiners without t-het-t. On both categories, the Dutch control group performs significantly better than group LE2. When observing the performance of groups LE1 and LE2 as compared to the Dutch control group, it can be concluded that length of exposure only appears to have an effect on the perception of dat and the test as a whole without the t-het-t contexts. Table 35 shows that group LE2 (the second language learners who have had lengthy exposure) performs better on these categories than group LE1 (the second language learners who have had average exposure), because they perform at native level and group LE1 does not. Tables 34 and 35 show significant differences between group LE1, LE2 and the Dutch control group for performance on het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts, but there are no significant differences between the groups for performance on het, the determiners and the test as a whole (with the t-het-t contexts). Thus, it appears that the significant differences between the groups for these contexts are solely due to the exclusion of the t-het-t contexts. This indicates that it is not the t-het-t contexts which cause the significant differences in performance, but other contexts of het. 4.7.4.3 Intensity of Exposure General Results As noted in section 4.7.2, the second language learners have been subdivided into two groups for intensity of exposure: groups IE1 and IE2. Group IE1 are the learners who have had average exposure to Dutch and group IE2 are the learners who have had more than average or intensive exposure to Dutch. Table 36 shows the performance of the two groups. Table 36 Group results for IE1 and IE2 (Mean in percentages correct) IE1 (N = 18) IE2 (N = 16) De 95.8 (SD = 8.6) 99.1 (SD = 2.1) Het 80.2 (SD = 20.3) 87.5 (SD = 11.5) All Determiners (de, het ) 88.1 (SD = 13.0) 93.4 (SD = 6.2) Non-neuter demonstratives (deze, die) 96.9 (SD = 13.0) 99.8 (SD = 1.0) Neuter demonstratives (dit, dat) 94.8 (SD = 11.7) 98.2 (SD = 2.9) All Demonstratives (deze, die, dit, dat) 95.8 (SD = 11.8) 98.9 (SD = 1.9) Whole Test (all test items) 92.0 (SD = 11.8) 96.1 (SD = 3.8) 61 Group IE1 Table 37 shows no significant differences within group IE1 with respect to performance on the various test categories. Table 37 Results of group IE1 (N = 18) De vs. Het Z = -.569; p = .569 Determiners vs. Demonstratives Z = -.740; p = .459 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -1.044; p = .297 Group IE2 Table 38 shows no significant differences within group IE2 with respect to performance on the various test categories. Table 38 Results of group IE2 (N = 16) De vs. Het Z = -.415; p = .678 Determiners vs. Demonstratives Z = -.311; p = .756 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -.537; p = .591 Comparison of the Groups This section compares the performance of groups IE1, IE2 and the Dutch control group. The results of only 31 second language learner participants (34 - 3) are used for this analysis, as the 3 outliers are excluded. The outliers are all part of group IE1. Consequently, group IE1 consists of only 15 second language learners (18 - 3). Again, performance on het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts are also looked at. Table 39 shows the results of the comparison of the groups analysis. 62 Table 39 Comparison of the groups: the Dutch control group vs. group IE1 vs. group IE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Dutch control group (N = 15) Group IE1 (N = 15) Group IE2 (N = 16) De 99.5 (SD = 1.4) 98.2 (SD = 2.4) 99.1 (SD = 2.1) Comparison of the groups (df = 2) χ2 = 3.191; p = .203 Het 94.6 (SD = 5.3) 88.1 (SD = 8.6) 87.5 (SD = 11.5) χ2 = 5.430; p = .066 Het without t-het-t 98.4 (SD = 2.0) 92.9 (SD = 8.4) 90.4 (SD = 10.0) χ2 = 9.292; p = .010 Determiners (de, het ) 96.9 (SD = 2.8) 93.2 (SD = 4.7) 93.2 (SD = 5.4) χ2 = 5.799; p = .055 Determiners without t-het-t 98.9 (SD = 1.3) 95.3 (SD = 5.3) 94.9 (SD = 5.4) χ2 = 8.703; p = .013 Deze 99.5 (SD = 1.8) 100 (SD = 0.0) 100 (SD = 0.0) χ2 = 2.067; p = .356 Die 100 (SD = 0.0) 100 (SD = 0.0) 99.5 (SD = 2.0) χ2 = 1.875; p = .392 Dit 99.5 (SD = 2.1) 98.9 (SD = 2.8) 99.0 (SD = 2.7) χ2 = .402; p = .818 Dat 100 (SD = 0.0) 98.6 (SD = 2.9) 97.4 (SD = 4.2) χ2 = 5.334; p = .069 Non-neuter demonstratives 99.8 (SD = 0.9) 100 (SD = 0.0) 99.8 (SD = 1.0) χ2 = .990; p = .610 Neuter demonstratives 99.7 (SD = 1.0) 98.8 (SD = 1.8) 98.2 (SD = 2.9) χ2 = 3.377; p = .185 Demonstratives 99.7 (SD = 0.7) 99.3 (SD = 1.0) 98.9 (SD = 1.9) χ2 = 2.093; p = .351 Whole Test (all test items) 98.4 (SD = 1.5) 96.3 (SD = 2.7) 96.1 (SD = 3.8) χ2 = 4.890; p = .087 Whole Test without t-het-t 99.3 (SD = 0.7) 97.3 (SD = 2.9) 96.9 (SD = 3.4) χ2 = 8.400; p = .015 Table 39 shows significant differences between the groups for performance on het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts. Table 40 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 40 Comparison of the groups (Dutch control group, group IE1 and group IE2) in pairs Significant differences are shaded (Bonferroni adjustment) IE1 vs. IE2 IE1 vs. Dutch control group IE2 vs. Dutch control group Het without t-het-t Z = -.608; p = .543 Z = -2.268; p = .023 Z = -2.965; p = .003 Determiners without t-het-t Z = -.322; p = .748 Z = -2.395; p = .017 Z = -2.735; p = .006 Whole Test without t-het-t Z = -.321; p = .748 Z = -2.304; p = .021 Z = -2.724; p = .006 Table 40 shows no significant differences between groups IE1 and IE2 for performance on the test categories and the test as a whole. There are also no significant differences in performance between group IE1 and the Dutch control group. This means that there are no significant differences in perception between the two groups and that group IE1 performs at native level. There are significant differences between group IE2 and the Dutch control group for performance on all of the test categories. In all cases, the Dutch control group performs significantly better than group IE2. When observing the performance of groups IE1 and IE2 as compared to the Dutch control group, it can be concluded that, for the test categories het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t context, intensity of exposure does not appear to have an effect on performance. Table 40 shows that group IE1 (the group which 63 has had average exposure) performs better on these categories than group IE2 (the group which has had intensive exposure), because they perform at native level and group IE2 does not. This is striking, because rather than group IE1, group IE2 is expected to perform at (near-)native level. Tables 39 and 40 show significant differences between group IE2 and the Dutch control group for performance on het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts, but there are no significant differences between the groups for performance on het, the determiners and the test as a whole (with the t-het-t contexts). Thus, it appears that the significant differences between the groups for these contexts are solely due to the exclusion of the t-het-t contexts. This indicates that it is not the t-het-t contexts which cause the significant differences in performance, but rather other contexts of het. 4.7.4.4 Proficiency General Results As noted in section 4.7.2, the second language learners have been subdivided into two groups for intensity of exposure: groups P1 and P2. Group P1 are the learners who have an average level of proficiency and group P2 are the learners who have a more than average or high level of proficiency. Table 41 shows the performance of the two groups. Table 41 Group results for P1 and P2 (Mean in percentages correct) P1 (N = 14) P2 (N = 20) De 96.8 (SD = 2.8) 97.8 (SD = 8.3) Het 75.3 (SD = 19.0) 89.5 (SD = 12.8) All Determiners (de, het ) 86.1 (SD = 10.0) 93.5 (SD = 10.0) Non-neuter demonstratives (deze, die) 99.7 (SD = 1.1) 97.3 (SD = 12.3) Neuter demonstratives (dit, dat) 95.3 (SD = 6.7) 97.2 (SD = 10.1) All Demonstratives (deze, die, dit, dat) 97.4 (SD = 3.5) 97.2 (SD = 11.1) Whole Test (all test items) 91.8 (SD = 6.4) 95.4 (SD = 10.5) Group P1 Table 42 shows that Group P1 perceives the non-neuter demonstratives significantly better than the neuter demonstratives. Table 42 Results of group P1 (N = 14) Significant differences are shaded De vs. Het Z = -.245; p = .807 Determiners vs. Demonstratives Z = -1.099; p = .272 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -1.992; p = .046 64 Group P2 Table 43 shows no significant differences within group P2 with respect to performance on the various test categories. Table 43 Results of group P2 (N = 20) De vs. Het Z =-.471; p = .638 Determiners vs. Demonstratives Z = -.465; p = .642 Non-neuter Demonstratives vs. Neuter Demonstratives Z = -1.777; p = .076 Comparison of the Groups This section compares the performance of groups P1, P2 and the Dutch control group. The results of only 31 second language learner participants (34 - 3) are used for this analysis, as the 3 outliers are excluded. Two of the outliers are part of group P1 and one is part of group P2. Consequently, group P1 consists of only 12 second language learners (14 - 2) and group P2 of only 19 second language learners (20 - 1). Again, performance on het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts are also looked at. Table 44 shows the results of the comparison of the groups analysis. Table 44 Comparison of the groups: the Dutch control group vs. group P1 vs. group P2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Dutch control group (N = 15) Group P1 (N = 12) Group P2 (N = 19) De 99.5 (SD = 1.4) 97.2 (SD = 2.3) 99.6 (SD = 1.3) Comparison of the groups (df = 2) χ2 = 11.139; p = .004 Het 94.6 (SD = 5.3) 81.3 (SD = 10.9) 91.9 (SD = 7.2) χ2 = 13.700; p = .001 Het without t-het-t 98.4 (SD = 2.0) 86.9 (SD = 10.7) 94.6 (SD = 6.9) χ2 = 15.140; p = .001 Determiners (de, het ) 96.9 (SD = 2.8) 89.4 (SD = 5.6) 95.6 (SD = 3.8) χ2 = 16.950; p = .000 Determiners without t-het-t 98.9 (SD = 1.3) 92.3 (SD = 5.2) 96.8 (SD = 4.6) χ2 = 18.578; p = .000 Deze 99.5 (SD = 1.8) 100 (SD = 0.0) 100 (SD = 0.0) χ2 = 2.067; p = .356 Die 100 (SD = 0.0) 99.3 (SD = 2.3) 100 (SD = 0.0) χ2 = 2.833; p = .243 Dit 99.5 (SD = 2.1) 98.0 (SD = 3.6) 99.6 (SD = 1.8) χ2 = 3.290; p = .193 Dat 100 (SD = 0.0) 96.0 (SD = 4.5) 99.3 (SD = 2.2) χ2 = 12.573; p = .002 Non-neuter demonstratives 99.8 (SD = 0.9) 99.7 (SD = 1.2) 100 (SD = 0.0) χ2 = 1.501; p = .472 Neuter demonstratives 99.7 (SD = 1.0) 97.0 (SD = 3.0) 99.4 (SD = 1.4) χ2 = 11.250 ; p = .004 Demonstratives 99.7 (SD = 0.7) 98.2 (SD = 2.) 99.7 (SD = 0.7) χ2 = 9.918; p = .007 Whole Test (all test items) 98.4 (SD = 1.5) 93.8 (SD = 3.4) 97.7 (SD = 2.1) χ2 = 17.654; p = .000 Whole Test without t-het-t 99.3 (SD = 0.7) 95.3 (SD = 3.3) 98.3 (SD = 2.5) χ2 = 19.339 ; p = .000 65 Table 44 shows significant differences between the groups for performance on all test categories (with and without the t-het-t contexts) except deze, die, dit and the non-neuter demonstratives. Table 45 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 45 Comparison of the groups (Dutch control group, group P1 and group P2) in pairs Significant differences are shaded (Bonferroni adjustment) P1 vs. P2 P1 vs. Dutch control group P2 vs. Dutch control group De Z = -2.884; p = .004 Z = -2.502; p = .012 Z = -.249; p = .804 Het Z = -2.896; p = .004 Z = -3.510; p = .000 Z = -.942; p = .346 Het without t-het-t Z = -2.518; p = .012 Z = -3.738; p = .000 Z = -1.816; p = .069 Determiners (de, het ) Z = -3.344; p = .001 Z = -3.873; p = .000 Z = -.761; p = .447 Determiners without t-het-t Z = -3.054; p = .002 Z = -4.204; p = .000 Z = -1.385; p = .166 Dat Z = -2.448; p = .014 Z = -3.034; p = .002 Z = -1.276; p = .202 Neuter demonstratives Z = -2.572; p = .010 Z = -2.814; p = .005 Z = -.744; p = .457 Demonstratives Z = -2.610; p = .009 Z = -2.549; p = .011 Z = -.198; p = .843 Whole Test (all test items) Z = -3.587; p = .000 Z = -3.817; p = .000 Z = -.546; p = .585 Whole Test without t-het-t Z = -3.210; p = .001 Z = -4.282; p = .000 Z = -1.242; p = .214 Table 45 shows significant differences between groups P1 and P2 for performance on all of the test categories. In all cases, group P2 performs significantly better than group P1. There are also significant differences between group P1 and the Dutch control group for performance on all of the test categories. In all cases, the Dutch control group performs significantly better than group P1. There are no significant differences in performance between group P2 and the Dutch control group. When observing the differences in performance between group P1 and P2, as well as the differences in performance when groups P1 and P2 are compared to the Dutch control group, proficiency level appears to have an effect on performance on de, het (with and without the t-het-t contexts), the determiners (with and without the t-het-t contexts), dat, the neuter demonstratives, the demonstratives as a whole and the test as a whole (with and without the t-het-t contexts). Both group P2 and the Dutch control group perform better on the test categories than group P1. Moreover, group P2 performs at native level while group P1 does not. Thus, a higher proficiency level seems to lead to a better perception of the test categories. The fact that the differences between the groups remain significant when the t-het-t contexts are excluded from the analysis indicates that there are also other contexts of het which cause difficulties for the second language learners. 66 4.7.4.5 Correlation Between the Factors and Performance Table 46 shows the correlation between the factors and the perception of the test categories. The total number of second language learners is 34 here. Table 46 Correlation coëfficients for the factors and the second language learners' performance on the perception task Significant correlations are shaded Length of Exposure Intensity of Exposure Age of First Exposure Proficiency De r = -.259; p = .139 r = .161; p = .362 r = -.028; p = .877 r = .009; p = .959 Het r = -.085; p = .633 r = .207; p = .240 r = -.262; p = .134 r = .375; p = .029 Het without t-het-t r = -.149; p = .400 r = .194; p = .271 r = -.252; p = .150 r = .309; p = .075 Determiners (de, het ) r = -.149; p = .400 r = .212; p = .228 r = -.215; p = .222 r = .298; p = .087 Determiners without t-het-t r = -.330; p = .057 r = .186; p = .292 r = -.182; p = .302 r = .201; p = .255 Deze r = -.302; p = .082 r = .091; p = .607 r = .044; p = .807 r = -.173; p = .327 Die r = -.262; p = .134 r = .067; p = .708 r = .037; p = .837 r = -.170; p = .337 Dit r = -.250 ; p = .154 r = .127; p = 472 r = -.073; p = .682 r = -.106; p = .550 Dat r = -.038; p = .833 r = .142; p = .424 r = -.085; p = .634 r = .167; p = .345 Non-neuter demonstratives r = -.287; p = .100 r = .081; p = .647 r = .041; p = .819 r = -.173; p = .329 Neuter demonstratives r = -.169; p = .339 r = .139; p = .434 r = -.081; p = .649 r = .007; p = .970 Demonstratives r = -.237; p = .177 r = .113; p = .524 r = -.020; p = .911 r = -.082; p = .645 Whole Test (all test items) r = -.201; p = .254 r = .180; p = .308 r = -.136; p = .443 r = .133; p = .454 Whole Test without t-het-t r = -.299; p = .086 r = .158; p = .372 r = -.109; p = .539 r = .068; p = .703 Table 46 shows that there is a significant (positive) correlation between the second language learners' performance on het (with the t-het-t contexts included) and the factor proficiency. Thus, according to the correlation analysis, of all the factors, only proficiency plays a role and only for performance on het. When proficiency increases, performance on het improves, i.e. the higher the proficiency level, the better het is perceived. The correlation is not very strong, however, as only 14.0% of the variance is explained (r² = .140). It is not possible to perform a regression analysis in order to see which factor best predicts the perception of het, de, etc. because the multiple regression models do not show a normal distribution of standardised residuals. Moreover, in some cases, the variance of the residuals is non-homoscedastic and the relationship between the predictor and criterion variables is non-linear. 67 4.7.4.6 Summary In this section various factors which interact with the second language learners' performance on the perception task have been controlled for. First the factor age of first exposure has been discussed. The second language learners were subdivided into group AE1 (early acquirers) and group AE2 (late acquirers). The comparison of the groups analysis shows that age of first exposure appears to have an effect on performance on the test. There are significant differences between groups AE1 and AE2, and earlier acquisition seems to lead to a better perception of het, the determiners and the test as a whole. Within groups AE1 and AE2, no significant differences have been found for performance on the various test categories. Secondly, the factor length of exposure has been discussed. The second language learners were subdivided into group LE1 (medium exposure) and group LE2 (lengthy exposure). The comparison of the groups analysis shows that length of exposure only seems to have an effect on performance on dat and the test as a whole without the t-het-t contexts. Group LE2 performs at native level on these test categories while group LE1 does not. Therefore, lengthier exposure seems to lead to a better perception of these test categories. Within groups LE1 and LE2, no significant differences have been found for performance on the various test categories. Thirdly, the factor intensity of exposure has been discussed. The second language learners were subdivided into group IE1 (average exposure) and group IE2 (intensive exposure). The comparison of the groups analysis shows that intensity of exposure does not seem to have an effect on performance on the test. Since group IE1 performs at native level on het without the t-het-t contexts, the determiners without the t-het-t contexts and the test as a whole without the t-het-t contexts while group IE2 does not, intensity of exposure does not seem to have an effect on performance on these test categories. Rather than group IE1, group IE2 is expected to perform at (near-)native level on these test categories. Within groups IE1 and IE2, no significant differences have been found for performance on the various test categories. Fourthly, the factor proficiency has been discussed. The second language learners were subdivided into group P1 (average proficiency) and group P2 (high proficiency). The comparison of the groups analysis shows that proficiency appears to have an effect on performance on the test. There are significant differences between groups P1 and P2 and a higher proficiency level seems to lead to a better perception of de, het, the determiners, dat, the neuter demonstratives, the demonstratives and the test as a whole. Group P1 perceives the non-neuter demonstratives significantly better than the neuter demonstratives. Within group P2, no significant differences have been found for performance on the various test categories. The correlation between the factors and performance on the test categories has also been investigated. A significant positive correlation has been found between the second language learners' performance on het and proficiency. When proficiency increases, performance on het improves, i.e. the higher the proficiency level, the better het is perceived. The correlation is not very strong, however, as only 14.1% of the variance is explained. Thus, according to the correlation analysis, only the factor proficiency plays a role, and only for performance on het. 68 When the various factors are controlled for it becomes clear that the second language learners mainly have difficulties with het. It has been found that only groups LE1 and P1 have difficulties with the perception of dat, whereas most groups have difficulties with het. Therefore, in the next section, the focus is on the perception of het. 4.7.5 The Influence of Phonological Context The previous sections show that there are significant differences between the Dutch control group and the second language learners for performance on the perception task. It has been found that the second language learners especially have difficulties with the perception of het (with and without the t-het-t contexts). In this section, the second language learners' misperception of het is further analysed by investigating the influence of phonological context. As the saliency hypothesis proposed in this thesis states, the non-saliency of het as induced by its phonological context may be the cause of the second language learners' difficulties with het, causing het to be misperceived. The following hierarchy of difficulty was hypothesised: het will be more difficult to perceive when followed by consonants than when followed by vowels, and more difficult to perceive when followed by /t/ than when followed by other consonants. It was also hypothesised that both the second language learners and the Dutch control group will especially have difficulties with the t-het-t contexts. This section investigates whether there are contexts in which het is more difficult to perceive than in others. This section is organised as follows. Section 4.7.5.1 discusses the contexts before het and compares the results per group. Section 4.7.5.2 discusses the contexts after het and compares the results per group. Section 4.7.5.3 discusses the het-stop contexts and compares the results per group. Section 4.7.5.4 discusses the full contexts of het and compares the results per group. Section 4.7.5.5 discusses the stop-het-stop contexts and compares the results per group. In every section, two factors are controlled for, namely age of first exposure and proficiency. The previous section shows that these two factors have an effect on performance on the test in general and performance on het in particular. The sections which investigate whether the hypothesised hierarchy of difficulty is borne out in the results, are especially sections 4.7.5.2, 4.7.5.3 and 4.7.5.5. As in the previous sections, the Wilcoxon signed-ranks test is used for comparisons within groups and the Mann-Whitney test is used for comparisons between groups. The Kruskal-Wallis test is used for comparisons between more than two groups. 69 4.7.5.1 The Contexts Before Het As pointed out in sections 4.4.1 and 4.4.3, the contexts before het are the final phonemes of the manner adverbs. As noted in section 4.4.3, five different phonemes are used for the endings of the manner adverbs, namely /k/, /t/, /m/, /n/ and /x/. These phonemes can be subdivided into three phonetic categories, namely stops (/k/, /t/), nasals (/m/, /n/) and fricatives (/x/). All of these phonetic categories occur before het in the experiment. Table 47 shows the phonetic categories that occur before het. Table 47 The phonetic categories of the contexts before het stops nasals fricatives /k/, /t/ /m/, /n/ /x/ 4.7.5.1.1 Group Results General Results Tables 48, 49 and 50 show the performance of the Dutch control group and the second language learners on the contexts before het. Table 48 Group results for performance on the contexts before het (Mean in percentages correct) Dutch control group (N = 15) Second language learners (N = 34) Stop (/k/, /t/) 92.9 (SD = 7.0) 84.7 (SD = 17.6) Nasal (/m/, /n/) 98.1 (SD = 5.0) 88.0 (SD = 19.9) Fricative (/x/) 99.2 (SD = 3.2) 84.2 (SD = 19.8) Context before het Table 49 Range of performance on the contexts before het (Mean in percentages correct): the Dutch control group (N = 15) Minimum Maximum Range Median Stop (/k/, /t/) 81 100 19 95.3 Nasal (/m/, /n/) 86 100 14 100 Fricative (/x/) 88 100 13 100 Context before het Table 50 Range of performance on the contexts before het (Mean in percentages correct): the second language learners (N = 34) Minimum Maximum Range Median Stop (/k/, /t/) 29 100 71 90.4 Nasal (/m/, /n/) 7 100 93 100 Fricative (/x/) 13 100 88 87.5 Context before het 70 The Dutch Control Group Table 51 shows a significant difference within the Dutch control group for performance on nasal-het vs. the fricative-het contexts. The Dutch control group perceives the fricative-het contexts significantly better than the nasal-het contexts. Table 51 Results of the Dutch control group (N = 15) for the contexts before het Significant differences are shaded Stop-het vs. Nasal-het Z = -.029; p = .977 Stop-het vs. Fricative-het Z = -.381; p = .703 Nasal-het vs. Fricative-het Z = -1.993; p = .046 The Second Language Learners Table 52 shows no significant differences within the second language learner group with respect to performance on the contexts before het. Table 52 Results of the second language learner group (N = 34) for the contexts before het Stop-het vs. Nasal-het Z = -.368; p = .713 Stop-het vs. Fricative-het Z = -.143; p = .886 Nasal-het vs. Fricative-het Z = -.086; p = .932 Comparison of the Groups This section compares the performance of the Dutch control group and the second language learner group on the contexts before het. As in the previous comparison of the groups analyses, the results of only 31 second language learner participants are used (34 - 3). The 3 outliers are excluded. Table 53 shows the results of the comparison of the groups analysis. The second language learner group (N = 31) is compared to the Dutch control group (N = 15). Performance on the stops before het is given both with and without the t-het-t contexts. Table 53 Comparison of the groups for performance on the contexts before het The Dutch control group vs. the second language learners Mean in percentages correct for each group and outcome of the Mann Whitney test Significant differences are shaded Dutch control group (N = 15) Second language learners (N = 31) Dutch control group vs. second language learners Stop (/k/, /t/) 92.9 (SD = 7.0) 88.6 (SD = 11.3) Z = -1.053; p = .292 Stop (/k/, /t/) without t-het-t 97.2 (SD = 6.0) 93.6 (SD = 10.5) Z = -1.151; p = .250 Nasal (/m/, /n/) 98.1 (SD = 5.2) 92.9 (SD = 10.3) Z = -1.820; p = .069 Fricative (/x/) 99.2 (SD = 3.2) 88.7 (SD = 12.6) Z = -3.261; p = .001 Context before het 71 Table 53 shows that the Dutch control group performs better than the second language learners on all of the contexts before het. There is a significant difference between the groups for performance on the fricative-het contexts. The Dutch controls perceive these contexts significantly better than the second language learners. 4.7.5.1.2 Group Results per Factor: Age of First Exposure General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for age of first exposure: groups AE1 and AE2. Group AE1 are the early acquirers of Dutch (before 18 years of age; puberty learners) and group AE2 are the later acquirers (after 18 years of age; adult learners). Table 54 shows the performance of the two groups on the contexts before het. Table 54 Group results for performance on the contexts before het: group AE1 and AE2 (Mean in percentages correct) AE1 (N = 7) AE2 (N = 27) Stop (/k/, /t/) 96.6 (SD = 2.3) 81.7 (SD = 18.5) Nasal (/m/, /n/) 96.9 (SD = 8.1) 85.7 (SD = 21.4) Fricative (/x/) 94.6 (SD = 6.7) 81.5 (SD = 21.2) Context before het Group AE1 Table 55 shows no significant differences within group AE1 with respect to performance on the contexts before het. Table 55 Results of group AE1 (N = 7) for the contexts before het Stop-het vs. Nasal-het Z = -.694; p = .488 Stop-het vs. Fricative-het Z = -.171; p = .864 Nasal-het vs. Fricative-het Z = -.171; p = .864 Group AE2 Table 56 shows no significant differences within group AE2 with respect to performance on the contexts before het. Table 56 Results of group AE2 (N = 27) for the contexts before het Stop-het vs. Nasal-het Z = -.374; p = .709 Stop-het vs. Fricative-het Z = -.067; p = .946 Nasal-het vs. Fricative-het Z = -.036; p = .971 72 Comparison of the Groups This section compares the performance of groups AE1, AE2 and the Dutch control group on the contexts before het. The results of only 31 second language learner participants (34 - 3) are used in this analysis, as the 3 outliers are excluded. The outliers are all part of group AE2. Consequently, group AE2 consists of only 24 second language learners (27 - 3). Table 57 shows the results of the comparison of the groups analysis. Groups AE1 (N = 7), AE2 (N = 24) and the Dutch control group (N = 15) are compared. Performance on the stops before het is given both with and without the t-het-t contexts. Table 57 Comparison of the groups: the Dutch control group vs. group AE1 vs. group AE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Context before het Dutch control group (N = 15) Group AE1 (N = 7) Group AE2 (N = 24) Stop (/k/, /t/) 92.9 (SD = 7.0) 96.6 (SD = 2.3) 86.2 (SD = 11.8) Comparison of the groups (df = 2) χ2 = 6.710; p = .035 Stop (/k/, /t/) without t-het-t 97.2 (SD = 6.0) 100 (SD = 0.0) 91.8 (SD = 11.3) χ2 = 6.529; p = .038 Nasal (/m/, /n/) 98.1 (SD = 5.0) 96.9 (SD = 8.1) 91.7 (SD = 10.7) χ2 = 5.411; p = .067 Fricative (/x/) 99.2 (SD = 3.2) 94.6 (SD = 6.7) 87.0 (SD = 13.5) χ2 = 12.176; p = .002 Table 57 shows significant differences between the groups for performance on the stop-het contexts (with and without t-het-t) and the fricative-het contexts. Table 58 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 58 Comparison of the groups (Dutch control group, group AE1 and group AE2) in pairs Significant differences are shaded (Bonferroni adjustment) Context before het AE1 vs. AE2 AE1 vs. Dutch control group AE2 vs. Dutch control group Stop (/k/, /t/) Z = -2.474; p = .013 Z = -.847; p = .397 Z = -1.625; p = .104 Stop (/k/, /t/) without t-het-t Z = -2.128; p = .033 Z = -1.241; p = .215 Z = -1.700; p = .089 Fricative (/x/) Z = -1.294; p = .196 Z = -2.003; p = .045 Z = -3.389; p = .001 Table 58 shows that there is a significant difference between groups AE1 and AE2 for performance on the stop-het contexts. Group AE1 perceives these contexts significantly better than group AE2. The difference in performance on the stop-het contexts without t-het-t, however, is not significant. For this reason, the significant difference in performance on the stop-het contexts appears to be solely due to the inclusion of the t-het-t contexts. There are no significant differences in performance between group AE1 and the Dutch control group. Thus, group AE1 performs at native level. There is a significant difference between group AE2 and the Dutch control group for performance on the fricative-het contexts. The Dutch control group perceives these contexts significantly better than group AE2. It should be observed that the outcome of the Kruskal-Wallis test gives a significant difference between the groups for performance on the stop-het contexts without t-het-t (Table 57). However, there are no 73 significant differences between the groups for these contexts according to the Mann-Whitney test (Table 58). 4.7.5.1.3 Group Results per Factor: Proficiency General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for proficiency: groups P1 and P2. Group P1 are the second language learners with an average proficiency level and group P2 are the second language learners with a high proficiency level. Table 59 shows the performance of the two groups on the contexts before het. Table 59 Group results for performance on the contexts before het: group P1 and P2 (Mean in percentages correct) P1 (N = 14) P2 (N = 20) Stop (/k/, /t/) 77.2 (SD = 16.7) 90.0 (SD = 16.6) Nasal (/m/, /n/) 79.6 (SD = 26.1) 93.9 (SD = 11.4) Fricative (/x/) 80.4 (SD = 17.5) 86.9 (SD = 21.3) Context before het Group P1 Table 60 shows no significant differences within group P1 with respect to performance on the contexts before het. Table 60 Results of group P1 (N = 14) for the contexts before het Stop-het vs. Nasal-het Z = -.134; p = .893 Stop-het vs. Fricative-het Z = -.063; p = .950 Nasal-het vs. Fricative-het Z = -.252; p = .801 Group P2 Table 61 shows no significant differences within group P2 with respect to performance on the contexts before het. Table 61 Results of group P2 (N = 20) for the contexts before het Stop-het vs. Nasal-het Z = -.022; p = .982 Stop-het vs. Fricative-het Z = -.121; p = .904 Nasal-het vs. Fricative-het Z = -.223; p = .823 74 Comparison of the Groups This section compares the performance of groups P1, P2 and the Dutch control group on the contexts before het. The results of only 31 second language learner participants (34 - 3) are used in this analysis, as the 3 outliers are excluded. Two of the outliers are part of group P1 and one is part of group P2. Consequently, group P1 consists of only 12 second language learners (14 - 2), and group P2 of only 19 second language learners (20 - 1). Table 62 shows the results of the comparison of the groups analysis. Groups P1 (N = 12), P2 (N = 19) and the Dutch control group (N = 15) are compared. Performance on the stops before het is given both with and without the t-het-t contexts. Table 62 Comparison of the groups: the Dutch control group vs. group P1 vs. group P2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Context before het Dutch control group (N = 15) Group P1 (N = 12) Group P2 (N = 19) Stop (/k/, /t/) 92.9 (SD = 7.0) 81.1 (SD = 11.7) 93.3 (SD = 8.3) Comparison of the groups (df = 2) χ2 = 12.575; p = .002 Stop (/k/, /t/) without t-het-t 97.2 (SD = 6.0) 88.2 (SD = 11.8) 97.1 (SD = 8.1) χ2 = 10.569; p = .005 Nasal (/m/, /n/) 98.1 (SD = 5.0) 88.7 (SD = 10.8) 95.5 (SD = 9.3) χ2 = 11.190; p = .004 Fricative (/x/) 99.2 (SD = 3.2) 85.4 (SD = 12.9) 90.8 (SD = 12.4) χ2 = 14.951; p = .001 Table 62 shows significant differences between the groups for performance on all contexts before het. Table 63 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 63 Comparison of the groups (Dutch control group, group P1 and group P2) in pairs Significant differences are shaded (Bonferroni adjustment) P1 vs. P2 P1 vs. Dutch control group P2 vs. Dutch control group Stop (/k/, /t/) Z = -3.310; p = .001 Z = -2.628; p = .009 Z = -.324; p = .746 Stop (/k/, /t/) without t-het-t Z = -2.704; p = .007 Z = -2.443; p = .015 Z = -.235; p = .814 Nasal (/m/, /n/) Z = -2.067; p = .039 Z = -2.791; p = .005 Z = .-731; p = .465 Fricative (/x/) Z = -1.372; p = .170 Z = -3.589; p = .000 Z = -2.580; p = .010 Context before het Table 63 shows that there is a significant difference between groups P1 and P2 for performance on the stop-het contexts (with and without t-het-t). Group P2 perceives these contexts significantly better than group P1. The fact that the difference remains significant when the t-het-t contexts are excluded from the analysis indicates that there are also other stop-het contexts which cause difficulties for group P1. There are also significant differences between group P1 and the Dutch control group for performance on all of the contexts before het. In all cases, the Dutch control group perceives the contexts significantly better than group P1. The fact that the difference remains significant when the thet-t contexts are excluded from the analysis indicates that there are also other stop-het contexts which cause difficulties for group P1. There is also a significant difference between group P2 and the 75 Dutch control group for performance on the fricative-het contexts. The Dutch control group perceives these contexts significantly better than group P2. 4.7.5.1.4 Summary This section has investigated the influence of the contexts before het on the perception of het. The contexts before het are stops (/k/, /t/), nasals (/m/, /n/) and fricatives (/x/). In addition, two factors have been controlled for: age of first exposure and proficiency. The comparison of the groups analysis shows significant differences between the Dutch control group and the second language learner group for performance on the fricative-het contexts. The Dutch control group perceives these contexts significantly better than the second language learners. Within the Dutch control group, the difference between performance on the nasal-het and fricative-het contexts is significant. The Dutch control group perceives the fricative-het contexts significantly better than the nasal-het contexts. Within the second language learner group, there are no significant differences in performance with respect to the contexts before het. When the factor age of first exposure is controlled for, the comparison of the groups analysis shows that there is a significant difference between groups AE1 and AE2 for performance on the stophet contexts. Group AE1 perceives these contexts significantly better than group AE2. There is also a significant difference between the Dutch control group and group AE2 for performance on the fricativehet contexts. The Dutch control group perceives these contexts significantly better than group AE2. There are no significant differences between group AE1 and the Dutch control group. Thus, group AE1 performs at native level. Within groups AE1 and AE2, there are no significant differences in performance with respect to the contexts before het. When the factor proficiency is controlled for, the comparison of the groups analysis shows that there are significant differences between groups P1 and P2 for performance on the stop-het contexts (with and without t-het-t). Group P2 perceives these contexts significantly better than group P1. There are also significant differences between the Dutch control group and group P1 for performance on all of the contexts before het (stop-het with and without t-het-t, nasal-het, fricative-het). In all cases, the Dutch control group perceives the contexts significantly better than group P1. Finally, there is a significant difference between the Dutch control group and group P2 for performance on the fricativehet contexts. The Dutch control group perceives these contexts significantly better than group P2. Within groups P1 and P2, there are no significant differences in performance with respect to the contexts before het. 76 4.7.5.2 The Contexts After Het As pointed out in sections 4.4.1 and 4.4.2, the contexts after het are the initial phonemes of the nonce nouns. As noted in section 4.4.2, eighteen different phonemes are used for the onsets of the nonce nouns, namely /p/, /t/, /k/, /a/, /ɑ/, /e/, /ε/, /ø/, /I/, /i/, /εi/, /ɔ/, /u/, /o/, /ɔu/, /œy/, /ʏ/ and /y/. These phonemes can be subdivided into two phonetic categories, namely stops (/p/, /t/, /k/) and vowels (/a/, /ɑ/, /e/, /ε/, /ø/, /I/, /i/, /εi/, /ɔ/, /u/, /o/, /ɔu/, /œy/, /ʏ/, /y/). Both of these phonetic categories occur after het. Table 64 shows the phonetic categories that occur after het. Table 64 The phonetic categories of the contexts after het stops vowels /p/, /t/, /k/ /a/, /ɑ/, /e/, /ε/, /ø/, /I/, /i/, /εi/, /ɔ/, /u/, /o/, /ɔu/, /œy/, /ʏ/, /y/ 4.7.5.2.1 Group Results General Results Tables 65, 66 and 67 show the performance of the Dutch control group and the second language learners on the contexts after het. Table 65 Group results for performance on the contexts after het (Mean in percentages correct) Dutch control group (N = 15) Second language learners (N = 34) Stop 91.9 (SD = 8.1) 80.9 (SD = 18.4) Vowel 100 (SD = 0.0) 89.2 (SD = 17.4) Context after het Table 66 Range of performance on the contexts after het (Mean in percentages correct): the Dutch control group (N = 15) Minimum Maximum Range Median Stop 78 100 22 94.4 Vowel 100 100 - - Context after het Table 67 Range of performance on the contexts after het (Mean in percentages correct): the second language learners (N = 34) Minimum Maximum Range Median Stop 17 100 83 88.9 Vowel 44 100 56 100 Context after het 77 The Dutch Control Group Table 68 shows no significant difference within the Dutch control group with respect to performance on the contexts after het. Table 68 Results of the Dutch control group (N = 15) for the contexts after het Het-Stop vs. Het-Vowel Z = .000; p = 1.000 The Second Language Learners Table 69 shows no significant difference within the second language learner group with respect to performance on the contexts after het. Table 69 Results of the second language learner group (N = 34) for the contexts after het Het-Stop vs. Het-Vowel Z = -.224; p = .823 Comparison of the Groups This section compares the performance of the Dutch control group and the second language learner group on the contexts after het. The results of only 31 second language learner participants are used (34 - 3), as the 3 outliers are excluded from the analysis. Table 70 shows the results of the comparison of the groups analysis. The second language learner group (N = 31) is compared to the Dutch control group (N = 15). Performance on the stops after het is given both with and without the t-het-t contexts. Table 70 Comparison of the groups for performance on the contexts after het The Dutch control group vs. the second language learners Mean in percentages correct for each group and outcome of the Mann Whitney test Significant differences are shaded Dutch control group (N = 15) Second language learners (N = 31) Dutch control group vs. second language learners Stop 91.9 (SD = 8.1) 85.5 (SD = 10.5) Z = -1.932; p = .053 Stop without t-het-t 97.3 (SD = 3.4) 91.0 (SD = 9.2) Z = -2.427; p = .015 Vowel 100 (SD = 0.0) 92.5 (SD = 14.2) Z = -2.438; p = .015 Context after het Table 70 shows that the Dutch control group performs better than the second language learners on all of the contexts after het. There are significant differences between the groups for performance on the het-stop contexts without t-het-t and the het-vowel contexts. The Dutch control group perceives these contexts significantly better than the second language learners. Table 70 shows that there is a significant difference between the groups for performance on the het-stop contexts without t-het-t, while the difference in performance on the het-stop contexts is not significant. For this reason, the 78 significant difference in performance on the het-stop contexts without t-het-t appears to be solely due to the exclusion of the t-het-t contexts. This suggests that it is especially other het-stop contexts which cause the significant difference in performance. 4.7.5.2.2 Group Results per Factor: Age of First Exposure General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for age of first exposure: groups AE1 and AE2. Group AE1 are the early acquirers of Dutch (before 18 years of age; puberty learners) and group AE2 are the later acquirers (after 18 years of age; adult learners). Table 71 shows the performance of the two groups on the contexts after het. Table 71 Group results for performance on the contexts after het: group AE1 and AE2 (Mean in percentages correct) AE1 (N = 7) AE2 (N = 27) Stop 92.9 (SD = 4.2) 77.8 (SD = 19.4) Vowel 98.4 (SD = 4.2) 86.8 (SD = 18.8) Context after het Group AE1 Table 72 shows no significant difference within group AE1 with respect to performance on the contexts after het. Table 72 Results of group AE1 (N = 7) for the contexts after het Het-Stop vs. Het-Vowel Z = -.343; p = .732 Group AE2 Table 73 shows no significant difference within group AE2 with respect to performance on the contexts after het. Table 73 Results of group AE2 (N = 27) for the contexts after het Het-Stop vs. Het-Vowel Z = -.217; p = .828 79 Comparison of the Groups This section compares the performance of groups AE1, AE2 and the Dutch control group on the contexts after het. The results of only 31 second language learner participants (34 - 3) are used in this analysis, as the 3 outliers are excluded. The outliers are all part of group AE2. Consequently, group AE2 consists of only 24 second language learners (27 - 3). Table 74 shows the results of the comparison of the groups analysis. Groups AE1 (N = 7), AE2 (N = 24) and the Dutch control group (N = 15) are compared. Performance on the stops after het is given both with and without the t-het-t contexts. Table 74 Comparison of the groups: the Dutch control group vs. group AE1 vs. group AE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Context after het Dutch control group (N = 15) Group AE1 (N = 7) Group AE2 (N = 24) Stop 91.9 (SD = 8.1) 92.9 (SD = 4.2) 83.3 (SD = 10.9) Comparison of the groups (df = 2) χ2 = 8.394; p = .015 Stop without t-het-t 97.3 (SD = 3.4) 96.2 (SD = 3.6) 89.4 (SD = 9.8) χ2 = 8.390; p = .015 Vowel 100 (SD = 0.0) 98.4 (SD = 4.2) 90.7 (SD = 15.6) χ2 = 7.893; p = .019 Table 74 shows significant differences between the groups for performance on all of the contexts after het (with and without t-het-t). Table 75 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 75 Comparison of the groups (Dutch control group, group AE1 and group AE2) in pairs Significant differences are shaded (Bonferroni adjustment) AE1 vs. AE2 AE1 vs. Dutch control group AE2 vs. Dutch control group Stop Z = -2.245; p = .025 Z = -.109; p = .913 Z = -2.423; p = .015 Stop without t-het-t Z = -1.643; p = .100 Z = -.735; p = .462 Z = -2.693; p = .007 Vowel Z = -1.256; p = .209 Z = -1.464; p = .143 Z = -2.642; p = .008 Context after het Table 75 shows no significant differences between groups AE1 and AE2 for performance on the contexts after het. There are also no significant differences between group AE1 and the Dutch control group for performance on the contexts after het. Thus, group AE1 performs at native level. There are significant differences between group AE2 and the Dutch control group for performance on all of the contexts after het (with and without t-het-t). In all cases, the Dutch control group performs significantly better than group AE2. The fact that the difference in performance remains significant when the t-het-t contexts are excluded from the analysis indicates that there are also other het-stop contexts which cause difficulties for group AE2. 80 4.7.5.2.3 Group Results per Factor: Proficiency General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for proficiency: groups P1 and P2. Group P1 are the second language learners with an average proficiency level and group P2 are the second language learners with a high proficiency level. Table 76 shows the performance of the two groups on the contexts after het. Table 76 Group results for performance on the contexts after het: group P1 and P2 (Mean in percentages correct) P1 (N = 14) P2 (N = 20) Stop 71.4 (SD = 20.9) 87.5 (SD = 13.4) Vowel 83.3 (SD = 21.7) 93.3 (SD = 12.7) Context after het Group P1 Table 77 shows no significant difference within group P1 with respect to performance on the contexts after het. Table 77 Results of group P1 (N = 14) for the contexts after het Het-Stop vs. Het-Vowel Z = -.350; p = .726 Group P2 Table 78 shows no significant difference within group P2 with respect to performance on the contexts after het. Table 78 Results of group P2 (N = 20) for the contexts after het Het-Stop vs. Het-Vowel Z = -.022; p = .983 Comparison of the Groups This section compares the performance of groups P1, P2 and the Dutch control group on the contexts after het. The results of only 31 second language learner participants (34 - 3) are used for this analysis, as the 3 outliers are excluded. Two of the outliers are part of group P1 and one is part of group P2. Consequently, group P1 consists of only 12 second language learners (14 - 2) and group P2 of only 19 second language learners (20 - 1). Table 79 shows the results of the comparison of the groups analysis. Groups P1 (N = 12), P2 (N = 19) and the Dutch control group (N = 15) are compared. Performance on the stops after het is given both with and without the t-het-t contexts. 81 Table 79 Comparison of the groups: the Dutch control group vs. group P1 vs. group P2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Context after het Dutch control group (N = 15) Group P1 (N = 12) Group P2 (N = 19) Stop 91.9 (SD = 8.1) 78.2 (SD = 11.2) 90.1 (SD = 7.1) Comparison of the groups (df = 2) χ2 = 11.809; p = .003 Stop without t-het-t 97.3 (SD = 3.4) 86.1 (SD = 11.2) 94.0 (SD = 6.2) χ2 = 9.685; p = .008 Vowel 100 (SD = 0.0) 88.0 (SD = 19.2) 95.3 (SD = 9.3) χ2 = 7.250; p = .027 Table 79 shows significant differences between the groups for performance on all of the contexts after het (with and without t-het-t). Table 80 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 80 Comparison of the groups (Dutch control group, group P1 and group P2) in pairs Significant differences are shaded (Bonferroni adjustment) P1 vs. P2 P1 vs. Dutch control group Stop Z = -2.984; p = .003 Z = -3.027; p = .002 Z = -.712; p = .477 Stop without t-het-t Z = -2.062; p = .039 Z = -2.867; p = .004 Z = -1.618; p = .106 Vowel Z = -1.053; p = .292 Z = -2.701; p = .007 Z = -2.113; p = .035 Context after het P2 vs. Dutch control group Table 80 shows a significant difference between groups P1 and P2 for performance on the het-stop contexts. Group P2 perceives these contexts significantly better than group P1. The difference in performance on the het-stop contexts without t-het-t, however, is not significant. For this reason, the significant difference in performance on the het-stop contexts appears to be solely due to the inclusion of the t-het-t contexts. There are also significant differences between group P1 and the Dutch control group for performance on all of the contexts after het (with and without t-het-t). In all cases, the Dutch control group perceives the contexts significantly better than group P1. The fact that the difference in performance remains significant when the t-het-t contexts are excluded from the analysis indicates that there are also other het-stop contexts which cause difficulties for group P1. There are no significant differences between group P2 and the Dutch control group for performance on the contexts after het. Thus, group P2 performs at native level. 82 4.7.5.2.4 Summary This section has investigated the influence of the contexts after het on the perception of het. The contexts after het are stops (/p/, /t/, /k/) and vowels (/a/, /ɑ/, /e/, /ε/, /ø/, /I/, /i/, /εi/, /ɔ/, /u/, /o/, /ɔu/, /œy/, /ʏ/, /y/). In addition, two factors have been controlled for: age of first exposure and proficiency. The comparison of the groups analysis shows that there are significant differences between the Dutch control group and the second language learner group for performance on the het-stop (without t-het-t) contexts and the het-vowel contexts. The Dutch control group perceives these contexts significantly better than the second language learners. Within the Dutch control group and the second language learner group, there are no significant differences in performance with respect to the contexts after het. When controlling for the factor age of first exposure, the comparison of the groups analysis shows significant differences between group AE2 and the Dutch control group for performance on all of the contexts after het (het-stop with and without t-het-t, het-vowel). The Dutch control group perceives these contexts significantly better than group AE2. There are no significant differences between group AE1 and the Dutch control group. Thus, group AE1 performs at native level. There are also no significant differences between groups AE1 and AE2. Within groups AE1 and AE2, there are no significant differences in performance with respect to the contexts after het. When the factor proficiency is controlled for, the comparison of the groups analysis shows that there is a significant difference between groups P1 and P2 for performance on the het-stop contexts. Group P2 perceives these contexts significantly better than group P1. There are also significant differences between the Dutch control group and group P1 for performance on all of the contexts after het (het-stop with and without t-het-t, het-vowel). The Dutch control group perceives these contexts significantly better than group P1. There are no significant differences between group P2 and the Dutch control group. Thus, group P2 performs at native level. Within groups P1 and P2, there are no significant differences in performance with respect to the contexts after het. 83 4.7.5.3 The Het-Stop Contexts In Chapter 3, section 3.3, it was hypothesised that het should be more difficult to perceive when followed by a /t/ than when followed by other consonants. In order to investigate whether this is indeed the case, performance on the het-/t/ and het-/p/k/ contexts (the het-stop contexts) can be compared. 4.7.5.3.1 Group Results General Results Table 81, 82 and 83 shows the performance of the Dutch control group and the second language learners on the het-stop contexts. Table 81 Group results for performance on the het-stop contexts (Mean in percentages correct) Dutch control group (N = 15) Second language learners (N = 34) het-t 85.2 (SD = 14.3) 71.2 (SD = 22.6) het-p/k 98.5 (SD = 3.9) 90.5 (SD = 18.4) Het-Stop Context Table 82 Range of performance on the het-stop contexts (Mean in percentages correct): the Dutch control group (N = 15) Minimum Maximum Range Median het-t 67.0 100 33.0 88.9 het-p/k 89.0 100 11.0 100 Het-Stop Context Table 83 Range of performance on the het-stop contexts (Mean in percentages correct): the second language learners (N = 34) Het-Stop Context het-t het-p/k Minimum Maximum Range Median 0 100 100 77.8 22.0 100 78.0 100 The Dutch Control Group Table 84 shows no significant difference within the Dutch control group with respect to performance on the het-stop contexts. Table 84 Results of the Dutch control group (N = 15) for the het-stop contexts Het-/t/ vs. Het-/p/k/ Z = -.344; p = .731 84 The Second Language Learners Table 85 shows no significant difference within the second language learner group with respect to performance on the het-stop contexts. Table 85 Results of the second language learner group (N = 34) for the het-stop contexts Het-/t/ vs. Het-/p/k/ Z = -.232; p = .816 Comparison of the Groups This section compares the performance of the Dutch control group and the second language learners on the het-stop contexts. The results of only 31 second language learner participants are used (34 - 3), as the 3 outliers are excluded from the analysis. Table 86 shows the results of the comparison of the groups analysis. The second language learners (N = 31) are compared to the Dutch control group (N = 15). Performance on the het-/t/ contexts is given both with and without the t-het-t contexts. Table 86 Comparison of the groups for performance on the het-stop contexts The Dutch control group vs. the second language learners Mean in percentages correct for each group and outcome of the Mann Whitney test Significant differences are shaded Dutch control group (N = 15) Second language learners (N = 31) Dutch control group vs. second language learners het-/t/ 85.2 (SD = 14.3) 75.6 (SD = 16.8) Z = -1.669; p = .095 het-/t/ without t-het-t 95.6 (SD = 7.6) 84.4 (SD = 15.5) Z = -2.470; p = .014 het-/p/k/ 98.5 (SD = 3.9) 95.3 (SD = 8.0) Z = -1.273; p = .203 Het-Stop Context Table 86 shows that the Dutch control group performs better than the second language learners on all of the het-stop contexts. There is a significant difference between the groups for performance on the het-/t/ contexts without t-het-t. The Dutch control group perceives these contexts significantly better than the second language learners. Table 86 shows a significant difference between the groups for performance on the het-/t/ contexts without t-het-t, while the difference in performance on the het-/t/ contexts is not significant. For this reason, the significant difference in performance on the het-/t/ contexts without t-het-t appears to be solely due to the exclusion of the t-het-t contexts. This suggests that it is other het-/t/ contexts which cause the significant difference in performance. 85 4.7.5.3.2 Group Results per Factor: Age of First Exposure General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for age of first exposure: groups AE1 and AE2. Group AE1 are the early acquirers of Dutch (before 18 years of age; puberty learners) and group AE2 are the later acquirers (after 18 years of age; adult learners). Table 87 shows the performance of the two groups on the het-stop contexts. Table 87 Group results for performance on the het-stop contexts: group AE1 and AE2 (Mean in percentages correct) AE1 (N = 7) AE2 (N = 27) het-/t/ 85.7 (SD = 8.4) 67.5 (SD = 23.7) het-/p/k/ 100 (SD = 0.0) 88.1 (SD = 20.0) Het-Stop Context Group AE1 Table 88 shows no significant difference within group AE1 with respect to performance on the het-stop contexts. Table 88 Results of group AE1 (N = 7) for the het-stop contexts Het-/t/ vs. Het-/p/k/ Z = -.171; p = .864 Group AE2 Table 89 shows no significant difference within group AE2 with respect to performance on the het-stop contexts. Table 89 Results of group AE2 (N = 27) for the het-stop contexts Het-/t/ vs. Het-/p/k/ Z = -.024; p = .981 Comparison of the Groups This section compares the performance of groups AE1, AE2 and the Dutch control group on the het-stop contexts. The results of only 31 second language learner participants (34 - 3) are used in this analysis, as the 3 outliers are excluded. The outliers are all part of group AE2. Consequently, group AE2 consists of only 24 second language learners (27 - 3). Table 90 shows the results of the comparison of the groups analysis. Groups AE1 (N = 7), AE2 (N = 24) and the Dutch control group (N = 15) are compared. Performance on the het-/t/ contexts is given both with and without the t-het-t contexts. 86 Table 90 Comparison of the groups: the Dutch control group vs. group AE1 vs. group AE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Het-Stop Context Dutch control group (N = 15) Group AE1 (N = 7) Group AE2 (N = 24) het-/t/ 85.2 (SD = 14.3) 85.7 (SD = 8.4) 72.7 (SD = 17.6) Comparison of the groups (df = 2) χ2 = 6.184; p = .045 het-/t/ without t-het-t 95.6 (SD = 7.6) 90.5 (SD = 8.9) 82.6 (SD = 16.7) χ2 = 6.943; p = .031 het-/p/k/ 98.5 (SD = 3.9) 100 (SD = 0.0) 94.0 (SD = 8.7) χ2 = 5.806; p = .055 Table 90 shows significant differences between the groups for performance on the het-/t/ contexts (with and without the t-het-t contexts). Table 91 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 91 Comparison of the groups (Dutch control group, group AE1 and group AE2) in pairs Significant differences are shaded (Bonferroni adjustment) AE1 vs. AE2 AE1 vs. Dutch control group AE2 vs. Dutch control group het-/t/ Z = -1.963; p = .050 Z = .000; p = 1.000 Z = -2.056; p = .040 het-/t/ without t-het-t Z = -1.026; p = .305 Z = -1.352; p = .176 Z = -2.536; p = .011 het-/p/k/ Z = -1.865; p = .062 Z = -.990; p = .322 Z = -1.737; p = .082 Het-Stop Context Table 91 shows no significant differences between groups AE1 and AE2 for performance on the hetstop contexts. There are also no significant differences between group AE1 and the Dutch control group for performance on the contexts after het. Thus, group AE1 performs at native level. There is a significant difference between group AE2 and the Dutch control group for performance on the het-/t/ contexts (without t-het-t). The Dutch control group performs significantly better on these contexts than group AE2. Table 91 shows a significant difference between group AE2 and the Dutch control group for performance on the het-/t/ contexts without t-het-t, while the difference in performance on the het-/t/ contexts is not significant. For this reason, the significant difference in performance on the het-/t/ contexts without t-het-t appears to be solely due to the exclusion of the t-het-t contexts. This suggests that it is other het-/t/ contexts which cause the significant difference in performance. It should be observed that the Kruskal-Wallis test shows a significant difference between the groups for performance on the het-/t/ contexts (Table 90), and that this difference is not borne out by the MannWhitney test (Table 91). 87 4.7.5.3.3 Group Results per Factor: Proficiency General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for proficiency: groups P1 and P2. Group P1 are the second language learners with an average proficiency level and group P2 are the second language learners with a high proficiency level. Table 92 shows the performance of the two groups on the het-stop contexts. Table 92 Group results for performance on the het-stop contexts: group P1 and P2 (Mean in percentages correct) P1 (N = 14) P2 (N = 20) het-/t/ 55.6 (SD = 25.4) 82.2 (SD = 11.6) het-/p/k/ 87.3 (SD = 18.9) 92.8 (SD = 18.1) Het-Stop Context Group P1 Table 93 shows no significant difference within group P1 with respect to performance on the het-stop contexts. Table 93 Results of group P1 (N = 14) for the het-stop contexts Het-/t/ vs. Het-/p/k/ Z = -.039; p = .969 Group P2 Table 94 shows no significant difference within group P2 with respect to performance on the het-stop contexts. Table 94 Results of group P2 (N = 20) for the het-stop contexts Het-/t/ vs. Het-/p/k/ Z = -.040; p = .968 Comparison of the Groups This section compares the performance of groups P1, P2 and the Dutch control group on the het-stop contexts. The results of only 31 second language learner participants (34 - 3) are used for this analysis, as the 3 outliers are excluded. Two of the outliers are part of group P1 and one is part of group P2. Consequently, group P1 consists of only 12 second language learners (14 - 2) and group P2 of only 19 second language learners (20 - 1). Table 95 shows the results of the comparison of the groups analysis. Groups P1 (N = 12), P2 (N = 19) and the Dutch control group (N = 15) are compared. Performance on the het-/t/ contexts is given both with and without the t-het-t contexts. 88 Table 95 Comparison of the groups: the Dutch control group vs. group P1 vs. group P2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Het-Stop Context Dutch control group (N = 15) Group P1 (N = 12) Group P2 (N = 19) het-/t/ 85.2 (SD = 14.3) 63.0 (SD = 17.9) 83.6 (SD = 10.0) Comparison of the groups (df = 2) χ2 = 12.837; p = .002 het-/t/ without t-het-t 95.6 (SD = 7.6) 75.0 (SD = 18.1) 90.4 (SD = 10.1) χ2 = 11.455; p = .003 het-/p/k/ 98.5 (SD = 3.9) 93.5 (SD = 8.8) 96.5 (SD = 7.5) χ2 = 3.221; p = .200 Table 95 shows significant differences between the groups for performance the het-/t/ contexts (with and without t-het-t). Table 96 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 96 Comparison of the groups (Dutch control group, group P1 and group P2) in pairs Significant differences are shaded (Bonferroni adjustment) P1 vs. P2 P1 vs. Dutch control group P2 vs. Dutch control group het-/t/ Z = -3.349; p = .001 Z = -2.953; p = .003 Z = -.378; p = .706 het-/t/ without t-het-t Z = -2.449; p = .014 Z = -3.054; p = .002 Z = -1.571; p = .116 het-/p/k/ Z = -1.143; p = .253 Z = -1.751; p = .080 Z = -.680; p = .497 Het-Stop Context Table 96 shows a significant difference between groups P1 and P2 for performance on the het-/t/ contexts (with and without t-het-t). Group P2 perceives these contexts significantly better than group P1. There are also significant differences between group P1 and the Dutch control group for performance on the het-/t/ contexts (with and without t-het-t). The Dutch control group perceives these contexts significantly better than group P1. There are no significant differences between group P2 and the Dutch control group for performance on the het-stop contexts. Thus, group P2 performs at native level. 4.7.5.3.4 Summary This section has investigated the influence of the het-stop contexts on the perception of het. The influence of the het-/t/ contexts and the het-/p/k/ contexts has been compared. In addition, two factors have been controlled for: age of first exposure and proficiency. The comparison of the groups analysis shows a significant difference between the Dutch control group and the second language learner group for performance on the het-/t/ contexts (without the t-het-t contexts). The Dutch control group perceives these contexts significantly better than the second language learners. Within the Dutch control group and the second language learner group, no significant differences in performance with respect to the het-stop contexts have been found. When the factor age of first exposure is controlled for, the comparison of the groups analysis shows a significant difference between group AE2 and the Dutch control group for performance on the het-/t/ contexts (with and without the t-het-t contexts). The Dutch control group perceives these contexts significantly better than group AE2. There is no significant difference between group AE1 and 89 the Dutch control group for performance on the het-stop contexts. Thus, group AE1 performs at native level on these contexts. There is also no significant difference in performance between groups AE1 and AE2. Within groups AE1 and AE2, there are no significant differences in performance with respect to the het-stop contexts. When the factor proficiency is controlled for, the comparison of the groups analysis shows a significant difference between groups P1 and P2 for performance on the het-/t/ contexts (with and without t-het-t). There is also a significant difference between group P1 and the Dutch control group with respect to performance on the het-/t/ contexts (with and without t-het-t). There are no significant differences between group P2 and the Dutch control group for performance on het het-stop contexts. Thus, group P2 performs at native level. Within groups P1 and P2, there are no significant differences in performance with respect to the contexts after het. 4.7.5.4 The Full Contexts of Het Section 4.7.5.1 has discussed the contexts before het (the final phonemes of the manner adverbs) and section 4.7.5.2 has discussed the contexts after het (the initial phonemes of the nonce nouns). The full contexts of het are the final phonemes of the manner adverbs and the initial phonemes of the nonce nouns taken together. Thus, the phonetic categories before and after het as discussed in sections 4.7.5.1 and 4.7.5.2 can be combined. Table 97 shows the phonetic categories of the full contexts of het. Table 97 The phonetic categories of the full contexts of het Full contexts of het Before het After het Stops Stops Stops Vowels Nasals Stops Nasals Vowels Fricatives Stops Fricatives Vowels 4.7.5.4.1 Group Results General Results Tables 98, 99 and 100 show the performance of the Dutch control group and the second language learners on the full contexts of het. 90 Table 98 Group results for performance on the full contexts of het (Mean in percentages correct) Full context of het Stop-Stop Dutch control group (N = 15) Second language learners (N = 34) 84.2 (SD = 16.0) 77.2 (SD = 19.1) Stop-Vowel 100 (SD = 0.0) 87.1 (SD = 22.5) Nasal-Stop 97.8 (SD = 5.9) 88.2 (SD = 20.7) Nasal-Vowel 100 (SD = 0.0) 90.2 (SD = 19.3) Fricative-Stop 98.3 (SD = 6.5) 77.2 (SD = 23.3) Fricative-Vowel 100 (SD = 0.0) 97.1 (SD = 17.2) Table 99 Range of performance on the full contexts of het (Mean in percentages correct): the Dutch control group (N = 15) Minimum Maximum Range Median Stop-Stop 63 100 38 87.5 Stop-Vowel 100 100 - - Nasal-Stop 83 100 17 100 Nasal-Vowel 100 100 - - Fricative-Stop 75 100 25 100 Fricative-Vowel 100 100 - - Full context of het Table 100 Range of performance on the full contexts of het (Mean in percentages correct): the second language learners (N = 34) Minimum Maximum Range Median Stop-Stop 25 100 75 75 Stop-Vowel 20 100 80 100 Nasal-Stop 0 100 100 100 Nasal-Vowel 33 100 67 100 Fricative-Stop 25 100 75 75 Fricative-Vowel 0 100 100 100 Full context of het The Dutch Control Group Table 101 shows that the Dutch control group perceives the stop-het vowel contexts significantly better than the fricative-het-stop contexts. In addition, the Dutch control group perceives the nasal-het-stop contexts significantly better than the fricative-het-stop contexts. The Dutch control group also perceives the nasal-het-vowel contexts significantly better than the fricative-het-stop contexts. Finally, the Dutch control group perceives the fricative-het-vowel contexts significantly better than the fricative-het-stop contexts. Thus, the Dutch control group especially appears to have difficulties with the fricative-het-stop contexts. 91 Table 101 Results of the Dutch control group (N = 15) for the full contexts of het Significant differences are shaded Full context of het Stop-Stop Stop-Vowel Nasal-Stop Nasal-Vowel Stop-Stop - Stop-Vowel - Z = -.479; p = .632 - Nasal-Stop - - Z= -.144; p = .885 Z = -1.907; p = .057 - Nasal-Vowel - - - Z = -.479; p = .632 Z = .000; p = 1.000 Z = -1.907; p = .057 - Fricative-Stop - - - - Fricative-Vowel - - - - FricativeStop Z = .000; p = 1.000 Z = -2.828; p = .005 Z = -1.993; p = .046 Z = -2.828; p = .005 - FricativeVowel Z = -.479; p = .632 Z = .000; p = 1.000 Z = -1.907; p = .057 Z = .000; p = 1.000 Z = -2.828; p = .005 - The Second Language Learners Table 102 shows no significant differences within group AE1 with respect to performance on the full contexts of het. Table 102 Results of the second language learners (N = 34) for the full contexts of het Full context of het Stop-Stop Stop-Vowel Nasal-Stop Nasal-Vowel Stop-Stop - Stop-Vowel - Z = -.526; p = .599 - Nasal-Stop - - Z = -.328; p = .743 Z = -1.127; p = .260 - Nasal-Vowel - - - Z = -.566; p = .571 Z = -1.585; p = .113 Z = -1.451; p = .147 - Fricative-Stop - - - - Fricative-Vowel - - - - FricativeStop Z = -.188; p = .851 Z = -.549; p = .583 Z = -.163; p = .870 Z = -1.001; p = .317 - FricativeVowel Z = -.473; p = .636 Z = -.044; p = .965 Z = -1.169; p = .242 Z = -1.106; p = .269 Z = -.182; p = .855 - Comparison of the Groups This section compares the performance of the Dutch control group and the second language learner group on the full contexts of het. The results of only 31 second language learner participants are used (34 - 3) for the analysis, as the 3 outliers are excluded. Table 103 shows the results of the comparison of the groups analysis. The second language learner group (N = 31) is compared to the Dutch control group (N = 15). Performance on the full contexts of het is given both with and without the t-het-t contexts. 92 Table 103 The Dutch control group vs. the second language learners Mean in percentages correct for each group and outcome of the Mann Whitney test Significant differences are shaded Dutch control group (N = 15) Second language learners (N = 31) Dutch control group vs. second language learners Stop-Stop 84.2 (SD = 16.0) 81.1 (SD = 14.7) Z = -.676; p = .499 Stop-Stop without t-het-t 96.0 (SD = 8.3) 94.8 (SD = 10.3) Z = -.245; p = .806 Stop-Vowel 100 (SD = 0.0) 90.3 (SD = 19.9) Z = -2.130; p = .033 Nasal-Stop 98.8 (SD = 5.9) 93.6 (SD = 9.3) Z = -1.571; p = .116 Nasal-Vowel 100 (SD = 0.0) 93.6 (SD = 13.4) Z = -1.807; p = .071 Fricative-Stop 98.3 (SD = 6.5) 82.3 (SD = 17.3) Z = -3.260; p = .001 Fricative-Vowel 100 (SD = 0.0) 100 (SD = 0.0) Z = .000; p = 1.000 Full context of het Table 103 shows that the Dutch control group performs better than the second language learners on all of the full contexts of het except the fricative-het-vowel contexts. Both groups perform equally well on the fricative-het-vowel contexts with a score of 100% correct (SD = 0.0). There are significant differences between the groups for performance on the stop-het-vowel and fricative-het-stop contexts. In both cases, the Dutch control group performs significantly better than the second language learners. 4.7.5.4.2 Group Results per Factor: Age of First Exposure General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for age of first exposure: groups AE1 and AE2. Group AE1 are the early acquirers of Dutch (before 18 years of age; puberty learners) and group AE2 are the later acquirers (after 18 years of age; adult learners). Table 104 shows the performance of the groups on the full contexts of het. Table 104 Group results for performance on the full contexts of het: group AE1 and AE2 (Mean in percentages correct) AE1 (N = 7) AE2 (N = 27) 91.2 (SD = 6.1) 73.6 (SD = 19.7) Stop-Vowel 100 (SD = 0.0) 83.7 (SD = 24.2) Nasal-Stop 97.6 (SD = 6.3) 85.8 (SD = 22.5) Nasal-Vowel 95.2 (SD = 12.6) 88.9 (SD = 20.7) Fricative-Stop 89.3 (SD = 13.4) 74.1 (SD = 24.5) 100 (SD = 0.0) 96.3 (SD = 19.2) Full context of het Stop-Stop Fricative-Vowel 93 Group AE1 Table 105 shows no significant differences within group AE1 with respect to performance on the full contexts of het. Table 105 Results of group AE1(N = 7) for the full contexts of het Full context of het Stop-Stop Stop-Vowel Nasal-Stop Nasal-Vowel Stop-Stop - Stop-Vowel - Z = -.176; p = .860 - Nasal-Stop - - Z = -.694; p = .488 Z = -1.265; p = .206 - Nasal-Vowel - - - Z = -.694; p = .488 Z = -1.265; p = .206 Z = .000; p = 1.000 - Fricative-Stop - - - - Fricative-Vowel - - - - FricativeStop Z = -.171; p = .864 Z = -.694; p = .488 Z = -.171; p = .864 Z = -.171; p = .864 - FricativeVowel Z = -.176; p = .860 Z = .000; p = 1.000 Z = -1.265; p = .206 Z = -1.265; p = .206 Z = -.694; p = .488 - Group AE2 Table 106 shows no significant differences within group AE2 with respect to performance on the full contexts of het. Table 106 Results of group AE2 (N = 27) for the full contexts of het Full context of het Stop-Stop Stop-Vowel Nasal-Stop Nasal-Vowel Stop-Stop - Stop-Vowel - Z = -.169; p = .866 - Nasal-Stop - - Z = -.196; p = .845 Z = -.757; p = .449 - Nasal-Vowel - - - Z = -.277; p = .782 Z = -.949; p = .343 Z = -.597; p = .550 - Fricative-Stop - - - - Fricative-Vowel - - - - FricativeStop Z = -.025; p = .980 Z = -.396; p = .692 Z = .000; p = 1.000 Z = -.556; p = .578 - FricativeVowel Z = -.400; p = .689 Z = -.888; p = .375 Z = -1.399; p = .162 Z = -.703; p = .482 Z = -.852; p = .394 - Comparison of the Groups This section compares the performance of groups AE1, AE2 and the Dutch control group on the full contexts of het. The results of only 31 second language learner participants (34 - 3) are used in this analysis, as the 3 outliers are excluded. The outliers are all part of group AE2. Consequently, group AE2 consists of only 24 second language learners (27 - 3). Table 107 shows the results of the comparison of the groups analysis. Groups AE1 (N = 7), AE2 (N = 24) and the Dutch control group (N = 15) are compared. Performance on the full contexts of het is given both with and without the t-het-t contexts. 94 Table 107 Comparison of the groups: the Dutch control group vs. group AE1 vs. group AE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Full context of het Dutch control group (N = 15) Group AE1 (N = 7) Group AE2 (N = 24) Stop-Stop 84.2 (SD = 16.0) 91.2 (SD = 6.1) 78.1 (SD = 15.3) Comparison of the groups (df = 2) χ2 = 4.287; p = .117 Stop-Stop without t-het-t 96.0 (SD = 8.3) 100 (SD = 0.0) 93.3 (SD = 11.3) χ2 = 2.746; p = .253 Stop-Vowel 100 (SD = 0.0) 100 (SD = 0.0) 87.5 (SD = 21.9) χ2 = 8.594; p = .014 Nasal-Stop 97.8 (SD = 5.9) 97.6 (SD = 6.3) 92.4 (SD = 9.8) χ2 = 4.496; p = .106 Nasal-Vowel 100 (SD - 0.0) 95.2 (SD = 12.6) 93.1 (SD = 13.8) χ2 = 3.467; p = .177 Fricative-Stop 98.3 (SD = 6.5) 89.3 (SD = 13.4) 80.2 (SD = 18.0) χ2 =11.911; p = .003 Fricative-Vowel 100 (SD = 0.0) 100 (SD = 0.0) 100 (SD = 0.0) χ2 = .000; p = 1.000 Table 107 shows significant differences between the groups for performance on the stop-het-vowel and fricative-het-stop contexts. Table 108 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). Table 108 Comparison of the groups (Dutch control group, group AE1 and group AE2) in pairs Significant differences are shaded (Bonferroni adjustment) AE1 vs. AE2 AE1 vs. Dutch control group AE2 vs. Dutch control group Stop-Vowel Z = -1.723; p = .085 Z = .000; p = 1.000 Z = -2.457; p = .014 Fricative-Stop Z = -1.165; p = .244 Z = -2.003; p = .045 Z = -3.383; p = .001 Full context of het Table 108 shows no significant differences in performance between groups AE1 and AE2. There are also no significant differences in performance between group AE1 and the Dutch control group. Thus, group AE1 performs at native level. There are significant differences between group AE2 and the Dutch group for performance on the stop-het-vowel and the fricative-het-stop contexts. The Dutch control group perceives these contexts significantly better than group AE2. 4.7.5.4.3 Group Results per Factor: Proficiency General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for proficiency: groups P1 and P2. Group P1 are the second language learners with an average proficiency level and group P2 are the second language learners with a high proficiency level. Table 109 shows the performance of the two groups on the full contexts of het. 95 Table 109 Group results for performance on the full contexts of het: group P1 and P2 (Mean in percentages correct) P1 (N = 14) P2 (N = 20) 66.1 (SD = 18.6) 85.0 (SD = 15.5) Stop-Vowel 80.0 (SD - 27.2) 92.0 (SD = 17.7) Nasal-Stop 79.8 (SD = 27.1) 94.2 (SD = 12.4) Nasal-Vowel 83.3 (SD = 25.3) 95.0 (SD = 12.2) Fricative-Stop 69.6 (SD = 24.4) 82.5 (SD = 21.6) 100 (SD = 0.0) 95.0 (SD = 22.4) Full context of het Stop-Stop Fricative-Vowel Group P1 Table 110 shows no significant differences within group P1 with respect to performance on the full contexts of het. Table 110 Results of group P1(N = 14) for the full contexts of het Full context of het Stop-Stop Stop-Vowel Nasal-Stop Nasal-Vowel Stop-Stop - Stop-Vowel - Z = -.598; p = .550 - Nasal-Stop - - Z = -.198; p = .843 Z = -.095; p = .925 - Nasal-Vowel - - - Z = -.045; p = .964 Z = -.321; p = .748 Z = -.443; p = .658 - Fricative-Stop - - - - Fricative-Vowel - - - - FricativeStop Z = .000; p = 1.000 Z = -.189; p = .850 Z = -.315; p = .753 Z = -.237; p = .812 - FricativeVowel Z = -.190; p = .850 Z = -.223; p = .823 Z = -.319; p = .749 Z = -.486; p = .627 Z = -.542; p = .588 - Group P2 Table 111 shows that group P2 perceives the nasal-het-vowel contexts significantly better than the stop-het-vowel contexts. Group P2 also perceives the nasal-het-vowel contexts significantly better than the nasal-het-stop contexts. Finally, group P2 perceives the nasal-het-vowel contexts significantly better than the fricative-het-vowel contexts. Thus, group P2 especially seems to have difficulties with the stop-het-vowel contexts, the nasal-het-stop contexts and the fricative-het-vowel contexts. The nasal-het-vowel contexts appear to cause the least difficulties for group P2. 96 Table 111 Results of group P2(N = 20) for the full contexts of het Significant differences are shaded Full context of het Stop-Stop Stop-Vowel Nasal-Stop Nasal-Vowel Stop-Stop - Stop-Vowel - Z = -.204; p = .838 - Nasal-Stop - - Z = -.243; p = .808 Z = -1.905; p = .057 - Nasal-Vowel - - - Z = -.122; p = .903 Z = -2.044; p = .041 Z = -2.595; p = .009 - Fricative-Stop - - - - Fricative-Vowel - - - - FricativeStop Z = -.202; p = .840 Z = -.558; p = .577 Z = -.244; p = .807 Z = -.794; p = .427 - FricativeVowel Z = -.578; p = .563 Z = -1.777; p = .076 Z = -1.070; p = .285 Z = -2.029; p = .042 Z = -1.235; p = .217 - Comparison of the Groups This section compares the performance of groups P1, P2 and the Dutch control group on the full contexts of het. The results of only 31 second language learner participants (34 - 3) are used for this analysis, as the 3 outliers are excluded. Two of the outliers are part of group P1 and one is part of group P2. Consequently, group P1 consists of only 12 second language learners (14 - 2) and group P2 of only 19 second language learners (20 - 1). Table 112 shows the results of the comparison of the groups analysis. Groups P1 (N = 12), P2 (N = 19) and the Dutch control group (N = 15) are compared. Performance on the full contexts of het is given both with and without the t-het-t contexts. Table 112 Comparison of the groups: the Dutch control group vs. group P1 vs. group P2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Significant differences are shaded Full context of het Dutch control group (N = 15) Group P1 (N = 12) Group P2 (N = 19) Stop-Stop 84.2 (SD = 16.0) 70.8 (SD = 14.4) 87.5 (SD = 11.0) Comparison of the groups (df = 2) χ2 = 8.664; p = .013 Stop-Stop without t-het-t 96.0 (SD - 8.3) 90.0 (SD = 13.5) 97.9 (SD = 6.3) χ2 = 4.442; p = .109 Stop-Vowel 100 (SD = 0.0) 83.3 (SD = 26.7) 94.7 (SD = 13.1) χ2 = 7.856; p = .020 Nasal-Stop 97.8 (SD = 5.9) 88.9 (SD = 10.9) 96.5 (SD = 7.0) χ2 = 7.775; p = .020 Nasal-Vowel 100 (SD = 0.0) 91.7 (SD = 15.1) 94.7 (SD = 12.5) χ2 = 3.804; p = .149 Fricative-Stop 98.3 (SD = 6.5) 77.1 (SD = 16.7) 85.5 (SD = 17.3) χ2 = 12.808; p = .002 Fricative-Vowel 100 (SD = 0.0) 100 (SD = 0.0) 100 (SD = 0.0) χ2 = .000; p = 1.000 Table 112 shows significant differences between the groups for performance on the stop-het-stop contexts, the stop-het-vowel contexts, the nasal-het-stop contexts and the fricative-het-stop contexts. Table 113 is more specific by showing the results of the comparison of the groups in pairs (Mann Whitney test). 97 Table 113 Comparison of the groups (Dutch control group, group P1 and group P2) in pairs Significant differences are shaded (Bonferroni adjustment) P1 vs. P2 P1 vs. Dutch control group P2 vs. Dutch control group Stop-Stop Z = -3.036; p = .002 Z = -1.982; p = .047 Z = -.417; p = .677 Stop-Vowel Z = -1.558; p = .119 Z = -2.703; p = .007 Z = -1.586; p = .113 Nasal-Stop Z = -2.159; p = .031 Z = -2.458; p = .014 Z = -.578; p = .564 Fricative-Stop Z = -1.400; p = .162 Z = -3.582; p = .000 Z = -2.567; p = .010 Full context of het Table 113 shows a significant difference between groups P1 and P2 for performance on the stop-hetstop contexts. Group P2 perceives these contexts significantly better than group P1. Because the Kruskal-Wallis test shows a significant difference between the groups for performance on the stop-hetstop contexts, but no significant difference for these contexts when the t-het-t contexts are excluded from the analysis, the significant difference between groups P1 and P2 for performance on the stophet-stop contexts appears to be solely due to the inclusion of the t-het-t contexts. There are also significant differences between group P1 and the Dutch control group for performance on the stop-hetvowel, nasal-het-stop and fricative-het-stop contexts. The Dutch control group perceives these contexts significantly better than group P1. There is also a significant difference between group P2 and the Dutch control group for performance on the fricative-het-stop contexts. The Dutch control group perceives these contexts significantly better than group P2. 4.7.5.4.4 Summary This section has investigated the influence of the full contexts of het on the perception of het. The full contexts of het are a combination of the contexts before and after het. In addition, two factors have been controlled for: age of first exposure and proficiency. The comparison of the groups analysis shows that there are significant differences between the Dutch control group and the second language learner group for performance on the stop-het-vowel and fricative-het-stop contexts. The Dutch control group perceives these contexts significantly better than the second language learners. Within the Dutch control group, there are significant differences between performance on the stop-het-vowel and the fricative-het-stop contexts, performance on the nasal-het-stop and the fricative-het-stop contexts, performance on the nasal-het-vowel and the fricative-het-stop contexts and performance on the fricative-het-stop and fricative-het-vowel contexts. In all cases, the Dutch control group perceives the fricative-het-stop contexts significantly worse than the other contexts of het. Within the second language learner group, there are no significant differences in performance with respect to the full contexts of het. When controlling for the factor age of first exposure, the comparison of the groups analysis shows significant differences between the Dutch control group and group AE2 for performance on the stop-het-vowel and fricative-het-stop contexts. The Dutch control group perceives these contexts significantly better than group AE2. There are no significant differences in performance between group AE1 and the Dutch control group. This means that group AE1 performs at native level. There are also 98 no significant differences between groups AE1 and AE2. Within groups AE1 and AE2, there are no significant differences in performance with respect to the full contexts of het. When the factor proficiency is controlled for, the comparison of the groups analysis shows that there is a significant difference between groups P1 and P2 for performance on the stop-het-stop contexts. Group P2 perceives these contexts significantly better than group P1. There are also significant differences between the Dutch control group and group P1 for performance on the stop-hetvowel contexts, the nasal-het-stop contexts and the fricative-het-stop contexts. The Dutch control group perceives these contexts significantly better than group P1. Finally, there is also a significant difference between the Dutch control group and group P2 for performance on the fricative-het-stop contexts. The Dutch control group perceives these contexts significantly better than group P2. Within group P1, there are no significant differences in performance with respect to the full contexts of het. Within group P2, there are significant differences between performance on the stop-het-vowel and nasal-het-vowel contexts, the nasal-het-stop and nasal-het-vowel contexts and the nasal-het-vowel and fricative-het-vowel contexts. In all cases, group P2 perceives the nasal-het-vowel contexts significantly better than the other contexts of het. 4.7.5.5 The Stop-Het-Stop Contexts The previous sections illustrate that, in some cases, the in-and exclusion of the t-het-t contexts from the analysis has an effect on the statistical results. Therefore, it is useful to have a closer look at performance on the stop-het-stop contexts in general and the t-het-t contexts in particular. This section compares the t-het-t contexts (t-het-tef, t-het-taaf, t-het-techt) to the t-het-p/k contexts (t-het-pank, t-het-parp, t-het-kag) and the k-het-p/t contexts (k-het-puig, k-het-tirs). Table 114 shows the stop-het-stop contexts subdivided into the three different categories. Table 114 The stop-het-stop contexts t-het-t contexts t-het-p/k contexts k-het-p/t contexts t-het-tef t-het-pank k-het-puig t-het-taaf t-het-parp k-het-tirs t-het-techt t-het-kag 4.7.5.5.1 Group Results General Results Tables 115, 116 and 117 show the performance of the Dutch control group and the second language learner group on the t-het-t contexts and the other stop-het-stop contexts. 99 Table 115 Group results for performance on the stop-het-stop contexts (Mean in percentages correct) Dutch control group (N = 15) Second language learners (N = 34) 64.4 (SD = 40.8) 54.9 (SD = 34.7) -t-het-p/k 97.8 (SD = 8.6) 92.2 (SD = 16.5) -k-het-p/t 93.3 (SD = 17.6) 88.2 (SD = 27.7) Stop-het-Stop Context -t-het-t Table 116 Range of performance on the stop-het-stop contexts (Mean in percentages correct): the Dutch control group (N = 15) Minimum Maximum Range Median -t-het-t 0 100 100 66.7 -t-het-p/k 67 100 33 100 -k-het-p/t 50 100 50 100 Stop-het-Stop Context Table 117 Range of performance on the stop-het-stop contexts (Mean in percentages correct): the second language learners (N = 34) Minimum Maximum Range Median -t-het-t 0 100 100 66.7 -t-het-p/k 33 100 67 100 -k-het-p/t 0 100 100 100 Stop-het-Stop Context The Dutch Control Group Table 118 shows that the Dutch control group perceives the t-het-p/k contexts significantly better than the k-het-p/t contexts. Table 118 Results of the Dutch control group (N = 15) for the stop-het-stop contexts Significant differences are shaded t-het-t t-het-p/k k-het-p/t t-het-t - Z = .000; p = 1.000 Z = -.057; p = .954 t-het-p/k - - Z = -1.993; p = .046 k-het-p/t - - - Stop-het-Stop context 100 The Second Language Learner Group Table 119 shows that the second language learner group perceives the t-het-p/k contexts significantly better than the k-het-p/t contexts. Table 119 Results of the second language learner group (N = 34) for the stop-het-stop contexts Significant differences are shaded t-het-t t-het-p/k k-het-p/t t-het-t - Z = -1.057; p = .291 Z = -.190; p = .849 t-het-p/k - - Z = -2.898; p = .004 k-het-p/t - - - Stop-het-Stop context Comparison of the Groups This section compares the performance of the Dutch control group and the second language learners on the stop-het-stop contexts. The results of only 31 second language learner participants are used (34 - 3), as the 3 outliers are excluded from the analysis. Table 120 shows the results of the comparison of the groups analysis. The second language learners (N = 31) are compared to the Dutch controls (N = 15). Table 120 The Dutch control group vs. the second language learners Mean in percentages correct for each group and outcome of the Mann Whitney test Dutch control group (N = 15) Second language learners (N = 31) Dutch control group vs. second language learners t-het-t 64.4 (SD = 40.8) 58.1 (SD = 33.3) Z = -.756; p = .450 t-het-p/k 97.8 (SD = 8.6) 95.7 (SD = 11.4) Z = -.630; p = .529 k-het-p/t 93.3 (SD = 17.6) 93.6 (SD = 17.0) Z = -.040; p = .968 Stop-het-Stop Context Table 120 shows no significant differences between the Dutch control group and the second language learners for performance on the stop-het-stop contexts. This means that the two groups have a similar perception of these contexts. 4.7.5.5.2 Group Results per Factor: Age of First Exposure General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for age of first exposure: groups AE1 and AE2. Group AE1 are the early acquirers of Dutch (before 18 years of age; puberty learners) and group AE2 are the later acquirers (after 18 years of age; adult learners). Table 121 shows the performance of the two groups on the stop-het-stop contexts. 101 Table 121 Group results for performance on the stop-het-stop contexts: group AE1 and AE2 (Mean in percentages correct) AE1 (N = 7) AE2 (N = 27) 76.2 (SD = 16.3) 49.4 (SD = 36.2) t-het-p/k 100 (SD = 0.0) 90.1 (SD = 18.1) k-het-p/t 100 (SD = 0.0) 85.2 (SD = 30.4) Stop-het-Stop Context t-het-t Group AE1 Table 122 shows no significant differences within group AE1 with respect to performance on the stop-het-stop contexts. Table 122 Results of group AE1 (N = 7) for the stop-het-stop contexts t-het-t t-het-p/k k-het-p/t t-het-t - Z = -.176; p = .860 Z = -.176; p = .860 t-het-p/k - - Z = .000; p = 1.000 k-het-p/t - - - Stop-het-Stop context Group AE2 Table 123 shows that group AE2 perceives the t-het-p/k contexts significantly better than the k-het-p/t contexts. Table 123 Results of group AE2 (N = 27) for the stop-het-stop contexts Significant differences are shaded t-het-t t-het-p/k k-het-p/t t-het-t - Z = -.304; p = .761 Z = -.229; p = .819 t-het-p/k - - Z = -2.117; p = .034 k-het-p/t - - - Stop-het-Stop context Comparison of the Groups This section compares the performance of groups AE1, AE2 and the Dutch control group on the stop-het-stop contexts. The results of only 31 second language learner participants (34 - 3) are used in this analysis, as the 3 outliers are excluded. The outliers are all part of group AE2. Consequently, group AE2 consists of only 24 second language learners (27 - 3). Table 124 shows the results of the comparison of the groups analysis. Groups AE1 (N = 7), AE2 (N = 24) and the Dutch control group (N = 15) are compared. 102 Table 124 Comparison of the groups: the Dutch control group vs. group AE1 vs. group AE2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Stop-het-Stop context Dutch control group (N = 15) Group AE1 (N = 7) Group AE2 (N = 24) t-het-t 64.4 (SD = 40.8) 76.2 (SD = 16.3) 52.8 (SD = 35.3) Comparison of the groups (df = 2) χ2 = 2.621; p = .270 t-het-p/k 97.8 (SD = 8.6) 100 (SD = 0.0) 94.4 (SD = 12.7) χ2 = 1.917; p = .382 k-het-p/t 93.3 (SD = 17.6) 100 (SD = 0.0) 91.7 (SD = 19.0) χ2 = 1.300; p = .522 Table 124 shows no significant differences between the groups for performance on the stop-het-stop contexts. This means that the groups have a similar perception of these contexts. 4.7.5.5.3 Group Results per Factor: Proficiency General Results As noted in section 4.7.2, the second language learners are subdivided into two groups for proficiency: groups P1 and P2. Group P1 are the second language learners with an average proficiency level and group P2 are the second language learners with a high proficiency level. Table 125 shows the performance of the two groups on the stop-het-stop contexts. Table 125 Group results for performance on the stop-het-stop contexts: group P1 and P2 (Mean in percentages correct) P1 (N = 14) P2 (N = 20) t-het-t 33.3 (SD = 37.0) 70.0 (SD = 23.9) t-het-p/k 88.1 (SD = 16.6) 95.0 (SD = 16.3) k-het-p/t 82.1 (SD = 31.7) 92.5 (SD = 24.5) Stop-het-Stop Context Group P1 Table 126 shows no significant differences within group P1 with respect to performance on the stop-het-stop contexts. Table 126 Results of group P1 (N = 14) for the stop-het-stop contexts t-het-t t-het-p/k k-het-p/t t-het-t - Z = -.284; p = .776 Z = -.158; p = .875 t-het-p/k - - Z = -1.104; p = .270 k-het-p/t - - - Stop-het-Stop context 103 Group P2 Table 127 shows no significant differences within group P2 with respect to performance on the stop-het-stop contexts. Table 127 Results of group P2 (N = 20) for the stop-het-stop contexts t-het-t t-het-p/k t-het-t - Z = -.227; p = .820 Z = -.189; p = .850 t-het-p/k - - Z = .000; p = 1.000 k-het-p/t - - - Stop-het-Stop context k-het-p/t Comparison of the Groups This section compares the performance of groups P1, P2 and the Dutch control group on the stop-het-stop contexts. The results of only 31 second language learner participants (34 - 3) are used for this analysis, as the 3 outliers are excluded. Two of the outliers are part of group P1 and one is part of group P2. Consequently, group P1 consists of only 12 second language learners (14 - 2) and group P2 of only 19 second language learners (20 - 1). Table 128 shows the results of the comparison of the groups analysis. Groups P1 (N = 12), P2 (N = 19) and the Dutch control group (N = 15) are compared. Table 128 Comparison of the groups: the Dutch control group vs. group P1 vs. group P2 Mean in percentages correct for each group and outcome of the Kruskal-Wallis test Stop-het-Stop context Dutch control group (N = 15) Group P1 (N =12) Group P2 (N = 19) t-het-t 64.4 (SD = 40.8) 38.9 (SD = 37.2) 70.2 (SD = 24.6) Comparison of the groups (df = 2) χ2 = 5.415; p = .067 t-het-p/k 97.8 (SD = 8.6) 91.7 (SD = 15.1) 98.3 (SD = 7.6) χ2 = 3.290; p = .193 k-het-p/t 93.3 (SD = 17.6) 87.5 (SD = 22.6) 97.4 (SD = 11.5) χ2 = 2.473; p = .290 Table 128 shows no significant differences between the groups for performance on the stop-het-stop contexts. This means that the groups have a similar perception of these contexts. 4.7.5.5.4 Summary This section has investigated the influence of the stop-het-stop contexts on the perception of het. The stop-het-stop contexts are the t-het-t contexts, the t-het-p/k contexts and the k-het-p/t contexts. In addition, two factors have been controlled for: age of first exposure and proficiency. The comparison of the groups analysis shows that there are no significant differences between the Dutch control group and the second language learners for performance on the stop-hetstop contexts. Thus, the two groups appear to have a similar perception of these contexts. The mean performance on the t-het-t contexts is very low for both of the groups, however. Within the Dutch control group and the second language learner group, there is a significant difference between 104 performance on the -t-het-p/k and the -k-het-p/t contexts. Both groups perceive the t-het-p/k contexts significantly better than the k-het-p/t contexts. When the factor age of first exposure is controlled for, the comparison of the groups analysis shows no significant differences in performance between the groups. The groups appear to have a similar perception of the stop-het-stop contexts. Within group AE1, there are no significant differences in performance with respect to performance on the stop-het-stop contexts. Within group AE2, performance on the t-het-p/k and the k-het-p/t contexts is significantly different. Group AE2 perceives the t-het-p/k contexts significantly better than the k-het-p/t contexts. When the factor proficiency is controlled for, the comparison of the groups analysis shows no significant differences in performance between the groups. Again, the groups appear to have a similar perception of the stop-het-stop contexts. Within groups P1 and P2, there are no significant differences in performance with respect to the stop-het-stop contexts. 4.7.6 Individual Results This section analyses the individual results of the second language learners for performance on het on the perception task. Table 20 shows that the Dutch control group has a mean score of 94.6% correct (SD = 5.3) for performance on het, and that the second language learner group has a mean score of 83.7% correct (SD = 16.9) for performance on het. In order to subdivide the second language learners into groups with respect to targetlike vs. non-targetlike, the mean score of the Dutch control group has been taken as the cut-off point. Table 129 shows the number and percentage of second language learners who are targetlike for performance on het and the number and percentage who are non-targetlike. Table 129 Targetlike vs. non-targetlike: the number and percentage of second language learners (total N = 34) for performance on het. Second language learners Number (N) Percentage (%) Targetlike ( ≥ 94.6% correct) 8 23.5 Non-targetlike (< 94.6% correct) 26 76.5 Table 129 shows that the majority of the second language learners is non-targetlike for performance on het: 76.5% of the second language learners is non-targetlike and only 23.5% is targetlike. It is also possible to look at individual performance by controlling for the factors age of first exposure and proficiency. Table 130 shows the number and percentage of second language learners in each group (AE1, AE2, P1, P2) for performance on het. 105 Table 130 The number and percentage of second language learners for performance on het: groups (AE1: N = 7; AE2: N = 27; P1: N = 14; P2: N = 20) Group AE1 Group AE2 Group P1 Group P2 Second language learners N % N % N % N % Targetlike (≥ 94.6 % correct) 4 57.1 4 14.8 1 7.1 7 35.0 Non-targetlike (< 94.6% correct) 3 42.9 23 85.2 13 92.9 13 65.0 Table 130 shows that the factors age of first exposure and proficiency both have an effect on the second language learners' performance on het. The individuals in group AE1 (the early acquirers) perform better than the individuals in group AE2 (the late acquirers), and the individuals in group P2 (the second language learners with a high proficiency level) perform better than the individuals in group P1 (the second language learners with an average proficiency level). The targetlike individuals mainly appear to be in groups AE1 and P2. However, it should be held in mind that the difference in number of subjects in groups AE1 and AE2 is considerable, and that this may affect the comparison. 4.7.7 General Summary This section gives a general summary of the perception results. First the general group results are summarised, then the group results per factor and finally the individual results. Group Results The results of the present study indicate that the second language learners have difficulties with the perception of het. The Dutch control group perceives het significantly better than the second language learners. The Dutch control group also perceives the determiners and the test as a whole significantly better than the second language learners. In addition, the second language learners have difficulties with the perception of dat; the Dutch control group perceives dat significantly better than the second language learners. However, when the various factors are controlled for, it is revealed that the second language learners' difficulties with dat are due to the performance of groups LE1 (the group which has had medium exposure) and P1 (the group with an average proficiency level) only, whereas most of the groups have difficulties with het. This indicates that the second language learners mainly have difficulties with the perception of het. They perform at native level on de, deze, die, dit, the nonneuter demonstratives, the neuter demonstratives and the demonstratives taken together. A comparison of performance within the groups shows no significant differences within the Dutch control group and the second language learner group with respect to performance on de vs. het, performance on the determiners vs. the demonstratives and performance on the non-neuter demonstratives vs. the neuter demonstratives. The misperception of het appears to be due to the following phonological contexts: fricativehet, het-stop (without t-het-t), het-vowel, het-/t/ (without t-het-t), stop-het-vowel and fricative-het-stop. The Dutch controls perceive these contexts significantly better than the second language learners. The fact that the difference between the groups is significant for performance on the het-stop contexts 106 and the het-/t/ contexts without the t-het-t contexts, and not significant for the het-stop contexts and the het-/t/ contexts with the t-het-t contexts, indicates that it is not the t-het-t contexts which cause the significant difference in perception but rather other het-stop and other het-/t/ contexts. An additional analysis of performance on the stop-het-stop contexts shows that, for these contexts, there are no significant differences in perception between the Dutch controls and the second language learners. However, for both groups, performance on the t-het-t contexts is very low, which indicates that both groups have serious difficulties in perceiving these contexts. A comparison of performance within the groups shows that the Dutch control group perceives the fricative-het contexts significantly better than the nasal-het contexts. Thus, the Dutch control group appears to have difficulties with the nasal-het contexts. The Dutch control group also appears to have difficulties with the fricative-het-stop contexts, as these contexts are perceived significantly worse than the stop-het-vowel contexts, the nasal-het-stop contexts, the nasal-het-vowel contexts and the fricative-het-vowel contexts. The analysis of performance on the stop-het-stop contexts shows that both the Dutch control group and the second language learner group perceive the t-het-p/k contexts significantly better than the k-het-p/t contexts. Thus, both the Dutch control group and the second language learners appear to have difficulties with the k-het-p/t contexts. Group Results per Factor The results of the present study suggest that the factors age of first exposure and proficiency have an effect on the second language learners' perception of het. Earlier acquisition and a higher proficiency level appear to lead to a better perception of het. The factors length of exposure and intensity of exposure do not seem to have an effect on the perception of het. A comparison of performance within the groups shows no significant differences within groups AE1, AE2, LE1, LE2, IE1, IE2 and P2 with respect to performance on de vs. het, performance on the determiners vs. the demonstratives and performance on the non-neuter demonstratives vs. the neuter demonstratives. Group P1 perceives the non-neuter demonstratives significantly better than the neuter demonstratives, however. When the factor age of first exposure is controlled for, the misperception of het appears to be due to the following phonological contexts: fricative-het, het-stop (with and without t-het-t), het-vowel, het-/t/ (without t-het-t), stop-het-vowel and fricative-het-stop. The Dutch control group perceives these contexts significantly better than group AE2. The fact that the difference between group AE2 and the Dutch control group for performance on the het-stop contexts remains significant when the t-het-t contexts are excluded from the analysis indicates that there are also other het-stop contexts which cause difficulties for group AE2. The fact that the difference between group AE2 and the Dutch control group is significant for performance on the het-/t/ contexts without the t-het-t contexts and not for performance on the het-/t/ contexts with the t-het-t contexts, indicates that it is not the t-het-t contexts which cause the significant difference in perception but rather other het-/t/ contexts. The analysis of performance on the stop-het-stop contexts shows no significant differences in perception between the groups for these contexts. However, all of the groups have a very low performance on the t-het-t contexts, which indicates that all groups have serious difficulties with these contexts. 107 The comparison of performance within the groups shows that there is a significant difference in performance within group AE2. The analysis of performance on the stop-het-stop contexts shows that group AE2 perceives the t-het-p/k contexts significantly better than the k-het-p/t contexts. Consequently, group AE2 appears to have difficulties with the k-het-p/t contexts. When the factor proficiency is controlled for, the misperception of het appears to be due to the following phonological contexts: stop-het (with and without t-het-t), nasal-het, fricative-het, het-stop (with and without t-het-t), het-vowel, het-/t/ (with and without t-het-t), stop-het-vowel, nasal-het-stop and fricative-het-stop. The Dutch control group perceives these contexts significantly better than group P1. The Dutch control group also perceives the fricative-het and fricative-het-stop contexts significantly better than group P2. The fact that the differences between group P1 and the Dutch control group for performance on the stop-het, het-stop and het-/t/ contexts remain significant when the t-het-t contexts are excluded from the analysis, indicates that there are also other stop-het, het-stop and het-/t/ contexts which cause difficulties for group P1. The analysis of performance on the stop-het-stop contexts shows that, for these contexts, there are no significant differences in perception between the groups. However, for all groups, performance on the t-het-t contexts is very low, which suggests that all groups have serious difficulties in perceiving these contexts. The comparison of performance within the groups shows that there are significant differences in performance within group P2. Group P2 perceives the nasalـhetـvowel contexts significantly better than the stopـhetـvowel contexts, the nasalـhetـvowel contexts significantly better than the nasalـhetـstop contexts and the nasalـhetـvowel contexts significantly better than the fricativeـhetـvowel contexts. Thus, group P2 especially appears to have difficulties with the stop-het-vowel contexts, the nasal-het-stop contexts and the fricative-het-vowel contexts. The nasal-het-vowel contexts appear to cause the least difficulties for group P2. Individual Results The individual analysis shows that the majority of the second language learners (76.5%) is non-targetlike for performance on het. Only 23.5% of the second language learners is targetlike for performance on het. Like the previous analyses, the individual analysis suggests that the factors age of first exposure and proficiency have an effect on the second language learners' performance on het. Overall, the individuals in group AE1 (the early acquirers) perform better than those in group AE2 (the late acquirers), and the individuals in group P2 (the learners with a high proficiency level) perform better than those in group P1 (the learners with an average proficiency level). The analysis shows that the targetlike individuals mainly appear to be in groups AE1 and P2. However, the difference in number of subjects in groups AE1 and AE2 is considerable, which may affect the comparison. 108 5 The Production Task 5.1 Introduction This chapter outlines the production task. The production task is designed to obtain additional production data from the second language learners who have participated in the perception experiment. The production task tests whether the second language learners can correctly produce gender. In Chapter 3, it is hypothesised that the second language learners will show non-targetlike behaviour in the production task. It is predicted that the second language learners will mainly overgeneralise de, but that they may also use het for common nouns. In addition, it is predicted that second language learners who have had lengthy and intensive exposure will overgeneralise to a lesser extent than the learners who have had little exposure. Moreover, it is predicted that the second language learners will produce the targetlike determiner more often for high frequency nouns than for low frequency nouns. This chapter is organised as follows. Section 5.2 outlines the participants, section 5.3 outlines the experimental design, section 5.4 describes the method, and section 5.5 presents the results. 5.2 Participants The subject group taking part in the production task was the same group of second language learners (N = 34) which participated in the perception experiment. A difference, however, was that the Dutch control group did not participate. This group could be expected to perform at native level and the participation of the Dutch native speakers was therefore deemed unnecessary. 5.3 Experimental Design: Stimuli The chosen task is a gap-filling exercise. The participants had to indicate whether 60 high-and medium frequency nouns needed the determiner de or het. The stimuli for the production task have been obtained from Dr. Laura Sabourin, who uses them in her PhD study on grammatical gender and second language processing (Sabourin, 2003). The stimuli consist of Dutch de- and het-nouns of medium and high frequency. Our research team has chosen 60 suitable nouns from Dr. Sabourin’s list of nouns. Following Dr. Sabourin's advice, nouns which indicate humans (e.g. dokter, doctor; lerares, teacher; mevrouw, lady) have been avoided as much as possible, since the second language learners might have knowledge that these nouns usually have de. Half of the nouns are high frequency nouns, and the other half of the nouns are medium frequency nouns. Likewise, half of the nouns are denouns, and the other half het-nouns. Table 131 shows that this results in 15 high-frequency de-nouns, 15 high-frequency het-nouns, 15 medium-frequency de-nouns and 15 medium-frequency het-nouns. 109 Table 131 Distribution of the nouns in the production task De Het Total High frequency 15 15 30 Medium frequency 15 15 30 Total 30 30 60 5.4 Procedure: Method Trial order was randomised via the online Random Sequence Generator. In addition, two lists were made, A being the reverse of B. This was done to prevent order effects. Brief written instructions were included immediately before the production task. In these instructions the participants were asked to write down de or het before every noun. The instructions emphasised that the participants should give a spontaneous answer instead of thinking their answer through. The participants were instructed to write down a question mark if they really did not know whether a noun needs de or het. This was done to prevent guessing. If subjects did not know a noun altogether, they could indicate this by writing down an ‘X’. As mentioned in Chapter 4, section 4.2, the participants were given the production task after the perception experiment had taken place. Right after they had finished the perception experiment, the participants filled in the production task. Half of the participants received version A and the other half received version B. The production task took the participants approximately 5 minutes. The time spent depended on individual speed. The production task is provided in Appendix 5 in the questionnaire under Lidwoordentest (Determiner Test), section D. 5.5 Results 5.5.1 Introduction Like the results of the perception experiment, the results of the production task are analysed with SPSS, version 14.0. Again, the standard significance level of 0.05 (5%) is used. As in the analysis of the perception results, the factors age of first exposure, proficiency, length of exposure and intensity of exposure are controlled for when analysing performance on the production task, and the same subdivision into groups is used. Again, the early acquirers are expected to perform better than the late acquirers, the second language learners with a high proficiency level are expected to perform better than the learners with an average proficiency level, and the learners who have had lengthy and intensive exposure are expected to perform better than the learners who have had medium and average exposure. It was found that, for performance on de, de high frequency and de medium frequency, the second language learners show an uneven distribution (Kolmogorov-Smirnov and Shapiro-Wilk: p < 0.05). They also show an uneven distribution for performance on het high frequency and the total 110 performance on het (Shapiro-Wilk: p < 0.05). For these reasons, and because all of the groups consist of less than 30 subjects when the various factors are controlled for, non-parametric tests are used to compare (performance within) the group(s). The Wilcoxon signed-ranks test is used for comparisons within groups and the Mann-Whitney test is used for comparisons between groups. This section is organised as follows. Section 5.5.2 gives the general results, section 5.5.3 gives the results per factor, section 5.5.4 gives the individual results and section 5.5.5 gives a general summary. 5.5.2 General Results Tables 132 and 133 show the second language learners' performance on the production task. Table 132 General production results (Mean in percentages correct) De Het Total High frequency 85.7 (SD = 15.8) 77.4 (SD = 18.4) 81.5 (SD = 14.3) Medium frequency 81.8 (SD = 15.3) 51.8 (SD = 18.7) 66.8 (SD = 14.2) Total 83.7 (SD = 14.2) 64.7 (SD = 16.0) 74.2 (SD = 12.9) Table 133 Range of performance (Mean in percentages correct) Minimum Maximum Range Median De 40 100 60 87 Het 30 100 70 65 High frequency 43 100 57 81.5 Medium frequency 43 93 50 63 De High frequency 47 100 53 93 De Medium frequency 33 100 67 87 Het High frequency 20 100 80 76.5 Het Medium frequency 20 100 80 53 Whole Test 50 97 47 75 Table 134 shows no significant differences within the second language learner group with respect to performance on the production task. Table 134 Production results of the second language learners (N = 34) De vs. Het Z = -.037; p = .970 High frequency vs. Medium frequency Z = -.068; p = .945 De High frequency vs. De Medium frequency Z = -.054; p = .957 Het High frequency vs. Het Medium frequency Z = -.286; p = .775 High frequency: De vs. Het Z = -.342; p = .732 Medium frequency: De vs. Het Z = .000; p = 1.000 111 5.5.3 Results per Factor This section analyses the group results per factor. It is organised as follows. Section 5.5.3.1 gives the results for the factor age of first exposure, section 5.5.3.2 gives the results for the factor proficiency, section 5.5.3.3 gives the results for the factor length of exposure and section 5.5.3.4 gives the results for the factor intensity of exposure. Section 5.5.3.5 gives the results of the correlation between the factors and performance, and section 5.5.3.6 gives the results of the regression analysis. 5.5.3.1 Age of First Exposure General Results As noted in section 4.7.2, the second language learners were subdivided into two groups for age of first exposure: groups AE1 and AE2. Group AE1 are the early acquirers of Dutch (before 18 years of age; puberty learners) and group AE2 are the later acquirers (after 18 years of age; adult learners). Tables 135 and 136 show the performance of the two groups. Table 135 Results of group AE1 (Mean in percentages correct) De Het Total High frequency 91.4 (SD = 9.2) 84.9 (SD = 15.7) 88.0 (SD = 12.3) Medium frequency 85.7 (SD = 16.0) 70.6 (SD = 21.5) 78.1 (SD = 16.4) Total 88.4 (SD = 11.9) 77.7 (SD = 17.7) 83.1 (SD = 13.9) Group AE1 (N = 7) Table 136 Results of group AE2 (Mean in percentages correct) De Het Total High frequency 84.2 (SD = 16.9) 75.4 (SD = 18.8) 79.9 (SD = 14.5) Medium frequency 80.8 (SD = 15.3) 46.9 (SD = 14.7) 63.9 (SD = 12.2) Total 82.4 (SD = 14.6) 61.3 (SD = 13.9) 71.9 (SD = 11.9) Group AE2 (N = 27) Group AE1 Table 137 shows no significant differences within group AE1 with respect to performance on the production task. Table 137 Production results of group AE1 (N = 7) De vs. Het Z = -.210; p = .833 High frequency vs. Medium frequency Z = .000; p = 1.000 De High frequency vs. De Medium frequency Z = -.086; p = .931 Het High frequency vs. Het Medium frequency Z = -.106; p = .915 High frequency: De vs. Het Z = .000; p = 1.000 Medium frequency: De vs. Het Z = .000; p = 1.000 112 Group AE2 Table 138 shows no significant differences within group AE2 with respect to performance on the production task. Table 138 Production results of group AE2 (N = 27) De vs. Het Z = -.241; p = .809 High frequency vs. Medium frequency Z = -.012; p = .990 De High frequency vs. De Medium frequency Z = -.229; p = .819 Het High frequency vs. Het Medium frequency Z = -.129; p = .898 High frequency: De vs. Het Z = -.289; p = .773 Medium frequency: De vs. Het Z = .000; p = 1.000 Comparison of the Groups Table 139 shows the comparison of the groups analysis. Groups AE1 (N = 7) and AE2 (N = 27) are compared. Table 139 Comparison of the groups: group AE1 vs. group AE2 Mean in percentages correct for each group and outcome of the Mann-Whitney test Significant differences are shaded Group AE1 (N = 7) Group AE2 (N = 27) Group AE1 vs. group AE2 De 88.4 (SD = 11.9) 82.4 (SD = 14.6) Z = -1.330; p = .183 Het 77.7 (SD = 17.7) 61.3 (SD = 13.9) Z = -2.246; p = .025 High frequency 88.0 (SD = 12.3) 79.9 (SD = 14.5) Z = -1.136; p = .256 Medium frequency 78.1 (SD = 16.4) 63.9 (SD = 12.2) Z = -2.467; p = .014 De High frequency 91.4 (SD = 9.2) 84.2 (SD = 16.9) Z = -.851; p = .395 De Medium frequency 85.7 (SD = 16.0) 80.8 (SD = 15.3) Z = -.975; p = .329 Het High frequency 84.9 (SD = 15.7) 75.4 (SD = 18.8) Z = -1.182; p = .237 Het Medium frequency 70.6 (SD = 21.5) 46.9 (SD = 14.7) Z = -2.581; p = .010 Whole Test 83.1 (SD = 13.9) 71.9 (SD = 11.9) Z = -1.962; p = .050 Table 139 shows significant differences between groups AE1 and AE2 for performance on het, the medium frequency nouns, the medium frequency het-nouns and the test as a whole. In all cases, group AE1 performs significantly better than group AE2. Thus, age of first exposure seems to have an effect on the second language learners' production of het, the medium frequency nouns, the medium frequency het-nouns and the test as a whole. Earlier acquisition seems to lead to a better production of het, the medium frequency nouns, the medium frequency het-nouns and the test as a whole. Because group AE1 performs significantly better than group AE2 on the medium frequency nouns and the medium frequency het-nouns, and because there are no differences between the groups for performance on the high frequency nouns and the high frequency het-nouns, there seems to be an effect of frequency here. 113 5.5.3.2 Proficiency General Results As noted in section 4.7.2, the second language learners were subdivided into two groups for proficiency: groups P1 and P2. Group P1 are the second language learners with an average proficiency level and group P2 are the second language learners with a high proficiency level. Tables 140 and 141 show the performance of the two groups. Table 140 Results of group P1 (Mean in percentages correct) De Het Total High frequency 79.1 (SD = 18.7) 68.4 (SD = 18.9) 73.7 (SD = 13.7) Medium frequency 73.8 (SD = 15.5) 41.4 (SD = 11.2) 57.6 (SD = 9.9) Total 76.4 (SD = 15.8) 55.1 (SD = 12.1) 65.8 (SD = 10.0) Group P1 (N = 14) Table 141 Results of group P2 (Mean in percentages correct) De Het Total High frequency 90.3 (SD = 11.8) 83.7 (SD = 15.5) 87.0 (SD = 12.3) Medium frequency 87.4 (SD = 12.7) 59.1 (SD = 19.7) 73.3 (SD = 13.3) Total 88.8 (SD = 10.6) 71.4 (SD = 15.1) 80.1 (SD = 11.6) Group P2 (N = 20) Group P1 Table 142 shows no significant differences within group P1 with respect to performance on the production task. Table 142 Production results of group P1 (N = 14) De vs. Het Z = -.472; p = .637 High frequency vs. Medium frequency Z = .000; p = 1.000 De High frequency vs. De Medium frequency Z = -.105; p = .916 Het High frequency vs. Het Medium frequency Z = .000; p = 1.000 High frequency: De vs. Het Z = -.251; p = .801 Medium frequency: De vs. Het Z = -.189; p = .850 114 Group P2 Table 143 shows no significant differences within group P2 with respect to performance on the production task. Table 143 Production results of group P2 (N = 20) De vs. Het Z = -.371; p = .711 High frequency vs. Medium frequency Z = -.393; p = .694 De High frequency vs. De Medium frequency Z = -.300; p = .765 Het High frequency vs. Het Medium frequency Z = -.019; p = .985 High frequency: De vs. Het Z = .000; p = 1.000 Medium frequency: De vs. Het Z = -.222; p = .825 Comparison of the Groups Table 144 shows the comparison of the groups analysis. Groups P1 (N = 14) and P2 (N = 20) are compared. Table 144 Comparison of the groups: group P1 vs. group P2 Mean in percentages correct for each group and outcome of the Mann-Whitney test Significant differences are shaded Group P1 (N = 14) Group P2 (N = 20) Group P1 vs. group P2 De 76.4 (SD = 15.8) 88.8 (SD = 10.6) Z = -2.979; p = .003 Het 55.1 (SD = 12.1) 71.4 (SD = 15.1) Z = -3.005; p = .003 High frequency 73.7 (SD = 13.7) 87.0 (SD = 12.3) Z = -2.518; p = .012 Medium frequency 57.6 (SD = 9.9) 73.3 (SD = 13.3) Z = -3.366; p = .001 De High frequency 79.1 (SD = 18.7) 90.3 (SD = 11.8) Z = -1.917; p = .055 De Medium frequency 73.8 (SD = 15.5) 87.4 (SD = 12.7) Z = -2.671; p = .008 Het High frequency 68.4 (SD = 18.9) 83.7 (SD = 15.5) Z = -2.260; p = .024 Het Medium frequency 41.4 (SD = 11.2) 59.1 (SD = 19.7) Z = -2.810; p = .005 Whole Test 65.8 (SD = 10.0) 80.1 (SD = 11.6) Z = -3.277; p = .001 Table 144 shows significant differences between groups P1 and P2 for performance on all of the comparisons except the high frequency de-nouns, although this is almost significant. In all cases, group P2 performs significantly better than group P1. 115 5.5.3.3 Length of Exposure General Results As noted in section 4.7.2, the second language learners were subdivided into two groups for length of exposure: groups LE1 and LE2. Group LE1 are the learners who have had exposure to Dutch for less than 12 years (medium exposure) and group LE2 are the learners who have had exposure to Dutch for 12 years or more (lengthy exposure). Tables 145 and 146 below show the performance of the two groups. Table 145 Results of group LE1 (Mean in percentages correct) De Het Total High frequency 82.7 (SD = 16.6) 72.8 (SD = 21.3) 77.7 (SD = 16.1) Medium frequency 80.0 (SD = 14.8) 46.2 (SD = 20.8) 63.1 (SD = 16.0) Total 81.3 (SD = 14.1) 59.6 (SD = 18.4) 70.4 (SD = 14.6) Group LE1 (N = 15) Table 146 Results of group LE2 (Mean in percentages correct) De Het Total High frequency 88.0 (SD = 15.1) 81.0 (SD = 15.3) 84.5 (SD = 12.4) Medium frequency 83.2 (SD = 16.0) 56.2 (SD = 16.0) 69.7 (SD = 12.2) Total 85.6 (SD = 14.3) 68.6 (SD = 12.9) 77.1 (SD = 11.0) Group LE2 (N = 19) Group LE1 Table 147 shows no significant differences within group LE1 with respect to performance on the production task. Table 147 Production results of group LE1 (N = 15) De vs. Het Z = -.630; p = .529 High frequency vs. Medium frequency Z = .000; p = 1.000 De High frequency vs. De Medium frequency Z = -.283; p = .777 Het High frequency vs. Het Medium frequency Z = -.031; p = .975 High frequency: De vs. Het Z = -.456; p = .648 Medium frequency: De vs. Het Z = -.089; p = .929 116 Group LE2 Table 148 shows no significant differences within group LE2 with respect to performance on the production task. Table 148 Production results of group LE2 (N = 19) De vs. Het Z = -.342; p = .732 High frequency vs. Medium frequency Z = -.020; p = .984 De High frequency vs. De Medium frequency Z = -.240; p = .811 Het High frequency vs. Het Medium frequency Z = -.047; p = .962 High frequency: De vs. Het Z = -.060; p = .952 Medium frequency: De vs. Het Z = -.261; p = .794 Comparison of the Groups Table 149 shows the comparison of the groups analysis. Groups LE1 (N = 15) and LE2 (N = 19) are compared. Table 149 Comparison of the groups: group LE1 vs. group LE2 Mean in percentages correct for each group and outcome of the Mann-Whitney test Group LE1 (N = 15) Group LE2 (N = 19) Group LE1 vs. group LE2 De 81.3 (SD = 14.1) 85.6 (SD = 14.3) Z = -1.101; p = .271 Het 59.6 (SD = 18.4) 68.6 (SD = 12.9) Z = -1.881; p = .060 High frequency 77.7 (SD = 16.1) 84.5 (SD = 12.4) Z = -1.291; p = .197 Medium frequency 63.1 (SD = 16.0) 69.7 (SD = 12.2) Z = -1.275; p = .202 De High frequency 82.7 (SD = 16.6) 88.0 (SD = 15.1) Z = -1.066; p = .287 De Medium frequency 80.0 (SD = 14.8) 83.2 (SD = 16.0) Z = -.847; p = .397 Het High frequency 72.8 (SD = 21.3) 81.0 (SD = 15.3) Z = -1.085; p = .278 Het Medium frequency 46.2 (SD = 20.8) 56.2 (SD = 16.0) Z = -1.839; p = .066 Whole Test 70.4 (SD = 14.6) 77.1 (SD = 11.0) Z = -1.528; p = .126 Table 149 shows no significant differences between groups LE1 and LE2 for performance on the various test categories. However, the difference between the groups for performance on het and the medium frequency het-nouns is approaching significance. Thus, there seems to be an effect of frequency here. Overall, length of exposure does not seem to have an effect on the second language learners' production of the various test categories. 117 5.5.3.4 Intensity of Exposure General Results As noted in section 4.7.2, the second language learners were subdivided into two groups for intensity of exposure: groups IE1 and IE2. Group IE1 are the learners who have had average exposure to Dutch and group IE2 are the learners who have had more than average or intensive exposure to Dutch. Tables 150 and 151 show the performance of the two groups. Table 150 Results of group IE1 (Mean in percentages correct) De Het Total High frequency 83.7 (SD = 18.5) 77.3 (SD = 15.5) 80.6 (SD = 13.1) Medium frequency 77.4 (SD = 16.5) 45.9 (SD = 15.0) 61.7 (SD = 11.7) Total 80.5 (SD = 16.0) 61.7 (SD = 12.5) 71.1 (SD = 11.3) Group IE1 (N = 18) Table 151 Results of group IE2 (Mean in percentages correct) De Het Total High frequency 87.9 (SD = 12.2) 77.5 (SD = 21.6) 82.6 (SD = 16.0) Medium frequency 86.7 (SD = 12.6) 58.4 (SD = 20.6) 72.5 (SD = 14.9) Total 87.3 (SD = 11.2) 68.0 (SD = 19.1) 77.6 (SD = 14.1) Group IE2 (N = 16) Group IE1 Table 152 shows no significant differences within group IE1 with respect to performance on the production task. Table 152 Production results of group IE1 (N = 18) De vs. Het Z = -.371; p = .711 High frequency vs. Medium frequency Z = -.218; p = .827 De High frequency vs. De Medium frequency Z = -.237; p = .813 Het High frequency vs. Het Medium frequency Z = -.109; p = .913 High frequency: De vs. Het Z = -.284; p = .777 Medium frequency: De vs. Het Z = -.196; p = .844 118 Group IE2 Table 153 shows no significant differences within group IE2 with respect to performance on the production task. Table 153 Production results of group IE2 (N = 16) De vs. Het Z = -.028; p = .977 High frequency vs. Medium frequency Z = -.513; p = .608 De High frequency vs. De Medium frequency Z = .000; p = 1.000 Het High frequency vs. Het Medium frequency Z = -.233; p = .816 High frequency: De vs. Het Z = .000; p = 1.000 Medium frequency: De vs. Het Z = .000; p =1.000 Comparison of the Groups Table 154 shows the comparison of the groups analysis. Groups IE1 (N = 18) and IE2 (N = 16) are compared. Table 154 Comparison of the groups: group IE1 vs. group IE2 Mean in percentages correct for each group and outcome of the Mann-Whitney test Significant differences are shaded Group IE1 (N = 18) Group IE2 (N = 16) Group IE1 vs. group IE2 De 80.5 (SD = 16.0) 87.3 (SD = 11.2) Z = -1.425; p = .154 Het 61.7 (SD = 12.5) 68.0 (SD = 19.1) Z = -1.265; p = .206 High frequency 80.6 (SD = 13.1) 82.6 (SD = 16.0) Z = -.590; p = .555 Medium frequency 61.7 (SD = 11.7) 72.5 (SD = 14.9) Z = -2.050; p = .040 De High frequency 83.7 (SD = 18.5) 87.9 (SD = 12.2) Z = -.194; p = .846 De Medium frequency 77.4 (SD = 16.5) 86.7 (SD = 12.6) Z = -1.720; p = .085 Het High frequency 77.3 (SD = 15.5) 77.5 (SD = 21.6) Z = -.366; p = .715 Het Medium frequency 45.9 (SD = 15.0) 58.4 (SD = 20.6) Z = -1.742; p = .081 Whole test 71.1 (SD = 11.3) 77.6 (SD = 14.1) Z = -1.244; p = .214 Table 154 shows that there is a significant difference between groups IE1 and IE2 for performance on the medium frequency nouns only. Group IE2 performs significantly better on the medium frequency nouns than group IE1. There are no significant differences between the groups for performance on the other test categories. However, there does seem to be an effect of frequency here, since the difference between the groups for performance on both the medium frequency de-nouns and the medium frequency het-nouns shows a tendency towards significance. 119 5.5.3.5 Correlation Between the Factors and Performance Table 155 shows the correlation between the factors and the second language learners' production of the test categories. Table 156 shows the strength of the correlations. Table 155 Correlation coëfficients for the factors and the second language learners' performance on the production task Significant correlations are shaded Age of First Exposure Proficiency Length of Exposure Intensity of Exposure De r = -.274; p = .116 r = .661; p = .000 r = .182; p = .304 r = .171; p = .332 Het r = -.598; p = .000 r = .399; p = .019 r = .431; p = .011 r = .174; p = .326 High frequency r = -.406; p = .017 r = .502; p = .003 r = .403; p = .018 r = .071; p = .690 Medium frequency r = -.536; p = .001 r = .614; p = .000 r = .264; p = .131 r = .282; p = .106 De High frequency r = -.262; p = .134 r = .618; p = .000 r = .215; p = .223 r = .153; p = .389 De Medium frequency r = -.240; p = .171 r = .592; p = .000 r = .107; p = .546 r = .157; p = .377 Het High frequency r = -.417; p = .014 r = .254; p = .147 r = .439; p = .009 r = -.014; p = .936 Het Medium frequency r = -.623; p = .000 r = .440; p = .009 r = .315; p = .070 r = .310; p = .075 Whole Test r = -.519; p = .002 r = .608; p = .000 r = .366; p = .033 r = .201; p = .254 Proficiency Length of Exposure Intensity of Exposure Table 156 Strength of the correlations Age of First Exposure De r² = -.075 r² = .436 r² = .033 r² = .029 Het r² = -.357 r² = .159 r² = .185 r² = .030 High frequency r² = -.164 r² = .252 r² = .162 r² = .005 Medium frequency r² = -.287 r² = .376 r² = .069 r² = .079 De High frequency r² = -.068 r² = .381 r² = .046 r² = .023 De Medium frequency r² = -.057 r² = .350 r² = .011 r² = .024 Het High frequency r² = -.173 r² = .064 r² = .192 r² = -.000 Het Medium frequency r² = -.388 r² = .193 r² = .099 r² = .096 Whole Test r² = -.269 r² = .369 r² = .133 r² = .040 Table 155 shows significant negative correlations between performance on het, the high frequency nouns, the medium frequency nouns, the high frequency het-nouns, the medium frequency het-nouns and the test as a whole and the factor age of first exposure. When age of first exposure increases, performance on these test categories deteriorates, i.e. the later Dutch is learnt, the worse the test categories are produced. Table 155 reveals significant positive correlations between proficiency and the second language learners' performance on most test categories. When proficiency increases, performance on these test categories improves, i.e. the higher the proficiency level, the better the test categories are produced. Table 155 also reveals significant positive correlations between the second language learners' performance on het, the high frequency nouns, the high frequency het-nouns and the test as a whole and the factor length of exposure. When length of exposure increases, performance on these test 120 categories improves, i.e. the more exposure the second language learners have had, the better the test categories are produced. As Table 155 shows, there are no significant correlations between performance and the factor intensity of exposure. This factor does not appear to have an effect on performance on the task. 5.5.3.6 Regression Analysis A multiple regression analysis can reveal which factor best predicts the second language learners' production of de, het, etc. Table 157 shows the results of the regression analysis (method: Enter). Table 157 Regression coëfficients for the factors and the second language learners' performance on the production task (standardised coëfficient Beta); Significant coëfficients are shaded Age of First Exposure Proficiency Length of Exposure Intensity of Exposure De β = -.116; p = .459 β = .691; p = .000 β = -.091; p = .557 β = -.108; p = .502 Het β = -.506; p = .002 β = .164; p = .306 β = .262; p = .093 β = -.044; p = .779 High frequency β = -.298; p = .081 β = .392; p = .028 β = .190; p = .254 β = -.151; p = .378 Medium frequency β = -.386; p = .014 β = .498; p = .003 β = .008; p = .959 β = -.025; p = .871 De High frequency β = -.108; p = .514 β = .628; p = .001 β = -.034; p = .837 β = -.097; p = .565 De Medium frequency β = -.110; p = .512 β = .645; p = .001 β = -.148; p = .374 β = -.111; p = .520 Het High frequency β = -.382; p = .035 β = .074; p = .680 β = .317; p = .075 β = -.151; p = .400 Het Medium frequency β = -.502; p = .003 β = .212; p = .188 β = .140; p = .362 β = .066; p = .677 Whole Test β = -.376; p = .016 β = .479; p = .004 β = .112; p = .448 β = -.086; p = .570 Table 157 shows that the factors age of first exposure and proficiency appear to be the best predictors of performance on the test categories. Age of first exposure best predicts performance on het, the high frequency het-nouns and the medium frequency het-nouns. Proficiency best predicts performance on de, the high frequency nouns, the high frequency de-nouns and the medium frequency de-nouns. Both age of first exposure and proficiency predict performance on the medium frequency nouns and the test as a whole. When compared to the factors age of first exposure and proficiency, the factors length of exposure and intensity of exposure do not have a substantial effect. 121 5.5.4 Individual Results As Table 132 shows, the second language learners have a mean score of 83.7% correct for performance on de, and a mean score of only 64.7% correct for performance on het. No significant differences have been found between the second language learners' performance on de and het (Table 134). However, the mean performance does show that the second language learners produce de better than het. The results also suggest that the second language learners overgeneralise de, and that they use het with common nouns. An analysis of the individual results can shed more light on these issues. The 34 second language learners can be categorised according to their response patterns for performance on the production task. The results are given in Table 158. A tick ( ۷ ) indicates that 20% or more of the nouns of a given category are produced with the respective determiner (cf. Unsworth, in press: 18-21). The cut-off point of 20% is arbitrary. Table 158 Individual response patterns for performance on the production task Target is shaded L2 learners Common Pattern N % De 1 6 17.6 ۷ 2 18 52.9 ۷ 3 10 29.4 ۷ Neuter Het De Het ۷ ۷ ۷ ۷ ۷ ۷ Table 158 shows that three different response patterns are found in the data. In the targetlike pattern 1, de is consistently used with common nouns and het is consistently used with neuter nouns. In this pattern, de is not produced with neuter nouns (i.e. de is not produced with neuter nouns at a rate higher than 20%) and het is not produced with common nouns (i.e. het is not produced with common nouns at a rate higher than 20%). In pattern 2, de is consistently used with common nouns and both de and het are used with neuter nouns. In pattern 3, de and het are used with both types of nouns. Response pattern 1 is the targetlike pattern, but only 17.6% of the second language learners shows this pattern. Pattern 2, which is the most frequent pattern, shows that more than half of the second language learners (52.9%; 18/34) also uses de with neuter nouns. The second most frequent pattern is pattern 3, which shows that almost one third of the second language learners (29.4%; 10/34) not only uses de with neuter nouns, but also het with common nouns. Pattern 3 could indicate random behaviour (using de and het with all or most nouns). Patterns 2 and 3 together show that as many as 82.4% of the second language learners (18+10 = 28; 28/34) use de with more than 20% of the neuter nouns. This points to a default strategy. 122 5.5.5 General Summary This section gives a general summary of the production results. First the general results are summarised, then the results per factor and finally the individual results. General Results No significant differences have been found within the second language learner group for performance on de vs. het, the high frequency nouns vs. the medium frequency nouns, the high frequency de-nouns vs. the medium frequency de-nouns, the high frequency het-nouns vs. the medium frequency het-nouns, the high frequency de-nouns vs. the high frequency het-nouns and the medium frequency de-nouns vs. the medium frequency het-nouns. Results per Factor Various factors interacting with the second language learners' performance on the production task have been controlled for. These factors are age of first exposure, proficiency, length of exposure and intensity of exposure. The factor age of first exposure appears to have an effect on performance on the test. There are significant differences between groups AE1 and AE2, and earlier acquisition seems to lead to a better production of the medium frequency nouns, the medium frequency het-nouns and the test as a whole. Moreover, significant negative correlations have been found between age of first exposure and the second language learners' performance on het, the high frequency nouns, the medium frequency nouns, the high frequency het-nouns, the medium frequency het-nouns and the test as a whole. When age of first exposure increases, performance on these test categories deteriorates, i.e. the later Dutch is learnt, the worse these test categories are produced. There also seems to be an effect of frequency here, because group AE1 performs significantly better on the medium frequency nouns and the medium frequency het-nouns than group AE2, and because there are no differences between the groups for performance on the high frequency nouns and the high frequency het-nouns. Proficiency also appears to have an effect on performance. There are significant differences between groups P1 and P2, and a higher proficiency level seems to lead to a better production of de, het, the high frequency nouns, the medium frequency nouns, the medium frequency de-nouns, the high frequency het-nouns, the medium frequency het-nouns and the test as a whole. Moreover, significant positive correlations have been found between proficiency and the second language learners' performance on de, het, the high frequency nouns, the medium frequency nouns, the high frequency de-nouns, the medium frequency de-nouns, the medium frequency het-nouns and the test as a whole. When proficiency increases, performance on these test categories improves, i.e. the higher the proficiency level, the better these test categories are produced. There are no significant differences between groups LE1 and LE2, but significant positive correlations have been found between length of exposure and the second language learners' performance on het, the high frequency nouns, the high frequency het-nouns and the test as a whole. When length of exposure increases, performance on these test categories improves, i.e. the more 123 exposure the second language learners have had, the better these test categories are produced. Thus, length of exposure appears to have an effect on performance. There also seems to be an effect of frequency here, because the difference between groups LE1 and LE2 for performance on het and the medium frequency het-nouns is approaching significance, with group LE2 performing better than group LE1. There is a significant difference between groups IEI and IE2 for performance on the medium frequency nouns. Group IE2 produces the medium frequency nouns significantly better than group IE1. No significant correlations have been found between intensity of exposure and performance, however. Consequently, intensity of exposure does not appear to have a substantial effect on performance. There does appear to be an effect of frequency here, because group IE2 produces the medium frequency nouns significantly better than group IE1 and because the difference between the groups for performance on both the medium frequency de-nouns and the medium frequency hetnouns shows a tendency towards significance, group IE2 performing better than group IEI. A multiple regression analysis suggests that the factors age of first exposure and proficiency are the best predictors of performance on the various test categories. Age of first exposure best predicts performance on het, the high frequency het-nouns and the medium frequency het-nouns. Proficiency best predicts performance on de, the high frequency nouns, the high frequency de-nouns and the medium frequency de-nouns. Both age of first exposure and proficiency best predict performance on the medium frequency nouns and the test as a whole. When compared to the factors age of first exposure and proficiency, the factors length of exposure and intensity of exposure do not have a substantial effect. Individual Results An analysis of the individual results reveals three different response patterns in the data. The targetlike pattern 1 (de is consistently used with common nouns and het is consistently used with neuter nouns) is only witnessed in 17.6% of the second language learners. The most frequent pattern is pattern 2. In pattern 2 de is used with common nouns and both de and het are used with neuter nouns. Pattern 2 is witnessed in 52.9% of the second language learners. The second most frequent pattern is pattern 3. In pattern 3 de and het are used with both types of nouns. Pattern 3 is witnessed in 29.4% of the second language learners. The results show that only 17.6% of the second language learners is targetlike, that as many as 82.4% of the second language learners (patterns 2 and 3 together) use de with neuter nouns, and that 29.4% of the second language learners also use het with common nouns. Due to the frequent use of de with neuter nouns, these results point to a default strategy. 124 6 Perception and Production Introduction The previous chapters have analysed the second language learners' performance on the perception experiment (Chapter 4) and the production task (Chapter 5). Chapter 4 has shown that the second language learners have difficulties with the perception of het, but not with the perception of de. Chapter 5 has shown that the learners have difficulties with the production of both de and het. As Llisterri (1995) and Rochet (1995) point out, both perception and production are important in the process of second language acquisition, but the relationship between the perception and production of second language speech sounds is still rather unclear. The question in this thesis, is whether a better perception of het leads to a better production of het, or whether a better production of het leads to a better perception of het. In order to investigate this, the results of the perception experiment and the production task can be combined. Because the focus in this thesis is on the perception of het, and because the second language learners in this study mainly show difficulties with the perception of het, only the results of performance on het will be used (perception and production). The results of all 34 second language learner participants are used for the current analyses, because the comparisons are made within the group of second language learners. Therefore, the 3 outliers need not be excluded from the analyses. The results of the Dutch control group cannot be compared to those of the second language learners here, because the Dutch control group has not participated in the production task. Does Good Perception Lead To Good Production? According to the saliency hypothesis as posed in this thesis, a misperception of het should lead to production difficulties. Conversely, a good perception of het should lead to a good production of het. This section investigates whether those participants who have a better perception of het also show a better production of het. Table 20 shows that the second language learners have a mean score of 83.7% correct for performance on het (perception). The second language learners are subdivided into two groups with respect to their performance on het on the perception task: a group which shows a low to average performance and a group which shows a higher than average performance. The total mean performance (83.7% correct) serves as the cut-off point. This has resulted in the subdivision as shown in Table 159. Table 159 Subdivision into groups for performance on het (perception; Mean in percentages correct) Group N Mean Std.Deviation Minimum Maximum Range PercHet1 11 64.8 17.9 26.0 81.0 55.0 PercHet2 23 92.7 4.7 85.0 100 15.0 125 The groups shown in Table 159 can be used to compare performance on the production task. Tables 160 and 161 show the performance of groups PercHet1 and PercHet2 on the production task. Table 160 The performance of group PercHet1 on the production task (Mean in percentages correct) Group PercHet1 (N = 11) De Het Total High frequency 81.3 (SD = 19.2) 66.6 (SD = 20.6) 73.9 (SD = 17.0) Medium frequency 75.8 (SD = 17.8) 45.5 (SD = 12.5) 60.6 (SD = 9.9) Total 78.6 (SD = 17.6) 56.1 (SD = 12.5) 67.3 (SD = 11.8) Table 161 The performance of group PercHet2 on the production task (Mean in percentages correct) Group PercHet2 (N = 23) De Het Total High frequency 87.8 (SD = 13.8) 82.6 (SD = 15.0) 85.2 (SD = 11.6) Medium frequency 84.7 (SD = 13.4) 54.8 (SD = 20.6) 69.8 (SD = 15.1) Total 86.1 (SD = 11.8) 68.7 (SD = 16.1) 77.4 (SD = 12.4) Table 162 shows the results of the comparison of the groups analysis. Groups PercHet1 and PercHet2 are compared by means of a Mann-Whitney test. Table 162 Comparison of the groups for performance on the production task Group PercHet1 vs. group PercHet2 Mean in percentages correct for each group and outcome of the Mann Whitney test Significant differences are shaded Group PercHet2 (N = 33) 86.1 (SD = 11.8) Group PercHet1 vs. Group PercHet2 De Group PercHet1 (N = 11) 78.6 (SD = 17.6) Het 56.1 (SD = 12.5) 68.7 (SD = 16.1) Z = -2.089; p = .037 High frequency 73.9 (SD = 17.0) 85.2 (SD = 11.6) Z = -1.797; p = .072 Medium frequency 60.6 (SD = 9.9) 69.8 (SD = 15.1) Z = -1.706; p = .088 De High frequency 81.3 (SD = 19.2) 87.8 (SD = 13.8) Z = -.886; p = .376 Z = -1.465; p = .143 De Medium frequency 75.8 (SD = 17.8) 84.7 (SD = 13.4) Z = -1.480; p = .139 Het High frequency 66.6 (SD = 20.6) 82.6 (SD = 15.0) Z = -2.247; p = .025 Het Medium frequency 45.5 (SD = 12.5) 54.8 (SD = 20.6) Z = -1.264; p = .206 Whole Test 67.3 (SD = 11.8) 77.4 (SD = 12.4) Z = -2.065; p = .039 Table 162 shows significant differences between the groups for performance on het, the high frequency het-nouns and the test as a whole. Group PercHet2 produces these test categories significantly better than group PercHet1. Group PercHet2 is the group with a higher perception score, and thus, it seems that a better perception of het may indeed contribute to a better production of het. There is no significant difference between the two groups for performance on the medium frequency het-nouns. The reason for this is probably the fact that both groups have a very low performance on these nouns. 126 An additional correlation analysis shows that the second language learners' performance on het (perception) significantly correlates with performance on het (production). Table 163 shows the result of the correlation analysis. Table 163 Correlation: performance on het (perception) vs. performance on het (production) The second language learners (N = 34) Significant correlations are shaded Het perception vs. Het production r = .362; p = .035 Table 163 shows a significant positive correlation between performance on het (perception) and performance on het (production). When performance on het (perception) improves, performance on het (production) improves, i.e. a better perception appears to contribute to a better production of het. However, the reverse is also true: When performance on het (production) improves, performance on het (perception) improves, i.e. a better production appears to contribute to a better perception of het. The correlation is rather weak, however, as it explains only 13.1% of the variance (r² = .131). Does Good Production Lead To Good Perception? In the previous section it was found that a better perception of het may further a better production of het. Possibly, a better production of het (i.e. a better knowledge of het) may also lead to a better perception of het. It is possible that second language learners of Dutch first need to have knowledge about gender before they can perceive het in everyday speech. This section investigates whether those participants who show a better production of het also have a better perception of het. Table 132 shows that the second language learners have a mean score of 64.7% correct for performance on het (production). The second language learners are subdivided into two groups with respect to their performance on het on the production task: a group which shows a low to average performance and a group which shows a higher than average performance. The total mean performance (64.7% correct) serves as the cut-off point. This has resulted in the subdivision shown in Table 164. Table 164 Subdivision into groups for performance on het (production; Mean in percentages correct) Group N Mean Std.Deviation Minimum Maximum Range ProdHet1 17 52.4 9.7 30 63 33 ProdHet2 17 76.9 10.6 67 100 33 The groups shown in Table 164 can be used to compare performance on the perception task. Table 165 shows the performance of groups ProdHet1 and ProdHet2 on the perception task. Table 165 also shows the results of the comparison of the groups analysis. Groups ProdHet1 and ProdHet2 are compared by means of a Mann-Whitney test. 127 Table 165 Comparison of the groups for performance on the perception task Group ProdHet1 vs. group ProdHet2 Mean in percentages correct for each group and outcome of the Mann Whitney test Significant differences are shaded Group ProdHet1 (N = 17) Group ProdHet2 (N = 17) Group ProdHet1 vs. Group ProdHet2 De 95.1 (SD = 8.6) 99.6 (SD = 1.7) Z = -3.318; p = .001 Het 76.9 (SD = 19.0) 90.4 (SD = 11.5) Z = -2.904; p = .004 All Determiners 86.1 (SD = 12.0) 94.9 (SD = 6.5) Z = -3.406; p = .001 Non-neuter demonstratives 96.5 (SD = 13.3) 100 (SD = 0.0) Z = -1.435; p = .151 Neuter demonstratives 93.7 (SD = 11.8) 99.2 (SD = 2.1) Z = -2.491; p = .013 All Demonstratives 95.0 (SD = 12.0) 99.5 (SD = 1.1) Z = -2.463; p = .014 Whole Test (all test items) 90.6 (SD = 11.6) 97.3 (SD = 3.5) Z = -3.475; p = .001 Table 165 shows significant differences between the groups for performance on all test categories except the non-neuter demonstratives. Group ProdHet2 perceives these categories significantly better than group ProdHet1. Group ProdHet2 is the group with a higher production score, and thus, it seems that a better production of het may indeed contribute to a better perception of het. Surprisingly, a better production of het also appears to contribute to a better perception of de, the determiners taken together, the neuter demonstratives, the demonstratives taken together and the test as a whole. As mentioned earlier, the results of the correlation analysis as given in Table 163 also suggest that a better production of het may lead to a better perception of het, i.e. that when performance on het (production) improves, performance on het (perception) improves. Summary When the perception and production results are combined, the comparison of the groups analysis suggests that a better perception of het may contribute to a better production of het. The results also suggest that a better production of het may contribute to a better perception of het. These findings are confirmed by an additional correlation analysis. Thus, the results suggest that a good perception of het may indeed further a good production of het and vice versa. Surprisingly, a better production of het also appears to contribute to a better perception of de and the demonstratives. 128 7 General Discussion 7.1 Perception This section discusses the results of the perception experiment in terms of the two main research questions as posed in Chapter 3. It discusses whether the predictions based on research questions 1 and 3 are borne out. Research Question 1 The first research question asked whether second language learners of Dutch can perceive the difference between het and de in the experiment or whether they fail to perceive het. It was asked whether the performance of the second language learners significantly differs from that of the Dutch native speaker control group, and whether phonological context has an influence on perception. As in natural speech, het systematically features as a non-salient form in the perception experiment as a result of its phonological context, while de mostly features as a salient form. For these reasons (and because de is the default and de is used for all plural nouns in Dutch), it was hypothesised that the second language learners should systematically fail to perceive het, perceiving de instead and thereby in a sense 'overgeneralising' de in perception. It was hypothesised that the Dutch control group should perceive het significantly better than the second language learners. It was also hypothesised that both the second language learners and the Dutch native speakers should perceive de significantly better than het. With respect to the influence of phonological context, the following hierarchy of difficulty was hypothesised: het should be more difficult to perceive when followed by consonants than when followed by vowels, and more difficult to perceive when followed by /t/ than when followed by other consonants. Both the second language learners and the Dutch control group should especially have difficulties with the t-het-t contexts. The results of the present study indicate that the second language learners have difficulties with the perception of het. The Dutch control group perceives het significantly better than the group of second language learners. This means that the second language learners regularly fail to perceive het, perceiving de instead and thereby 'overgeneralising' de in perception. Thus, in this respect, the prediction is borne out. There are no significant differences between the perception of de and het within the second language learner group and the Dutch control group. When the various factors are controlled for, there are also no significant differences between the perception of de and het within groups AE1, AE2, LE1, LE2, IE1, IE2, P1 and P2. Thus, the prediction that the second language learners and the Dutch control group should perceive de significantly better than het is not borne out. The influence of phonological context has been investigated extensively. The results of the present study indicate that the second language learners' misperception of het is due to the following phonological contexts: fricative-het, het-stop (without t-het-t), het-vowel, het-/t/ (without t-het-t), stophet-vowel and fricative-het-stop. The Dutch control group perceives these contexts significantly better 129 than the second language learners. The analysis of the influence of phonological context indicates that the predictions made in this respect are not entirely borne out. A clear pattern which shows that the second language learners find het more difficult to perceive when het is followed by consonants than when followed by vowels, and more difficult to perceive when followed by /t/ than when followed by other consonants (the hypothesised hierarchy of difficulty) has not been found. Indeed, the second language learners perceive the het-/t/ contexts (without t-het-t) significantly worse than the Dutch control group, but they perform at native level on the het-/p/k/ contexts. This indicates that the second language learners find het more difficult to perceive in contexts in which het is followed by /t/ (the het-/t/ contexts) than in contexts in which het is followed by another consonant (the het-/p/k/ contexts). Thus, in this respect, the prediction is borne out. The analysis of performance on the stop-het-stop contexts shows no significant differences in performance between the Dutch control group and the second language learners. However, the mean performance on the t-het-t contexts is very low for both of the groups (mean 64.4% correct, SD = 40.8 and mean 54.9% correct, SD = 34.7 respectively, see Table 139). The reason why no significant difference is found between the groups is probably that both groups have a very low performance on these contexts, and not just one of the groups. Thus, both groups appear to have serious difficulties with the t-het-t contexts. This is in line with the prediction that especially the t-het-t contexts should be difficult for both the Dutch control group and the second language learners. Thus, although the results are not statistically corroborated, this prediction, too, appears to be borne out. The second language learners also perceive the het-stop contexts (without t-het-t) and the fricative-het-stop contexts significantly worse than the Dutch control group. This indicates that the second language learners find het followed by consonants more difficult than the Dutch control group. However, the second language learners also perceive the het-vowel contexts and the stop-het-vowel contexts significantly worse than the Dutch control group. The finding that the second language learners have difficulties with het in contexts in which a vowel occurs after het was not predicted. Rather, it was hypothesised that this would not be the case. Because /t/-deletion is improbable before vowels, the phonological contrast between [də] and [ət] should be maximally clear in this case. The reason why this does not appear to be so in the present study is unclear. It is possible that the second language learners perceive the /t/ of [ət] as the onset of the nonce noun that follows. This would weaken the phonological contrast between [də] and [ət]. However, section 4.4.5 illustrates how a glottal stop should be inserted immediately before nonce nouns with an initial vowel in the present study. Section 4.4.5 describes the phonology of the contexts of the demonstratives, but in the contexts of the determiners the glottal stop should be inserted in the same way. A glottal stop should be inserted between the final /t/ of [ət] and the onset of the nonce noun. As noted in section 2.5.3, the glottal stop is not a separate phoneme in Dutch, but it is inserted before vowel-initial syllables within words after /a/ and /ə/ and often also at the beginning of a word (Jongenburger & van Heuven, 1991). A glottal stop between [ət] and the following nonce noun would maintain the phonological contrast between [ət] and [də]. It is possible, however, that the speaker in the experiment (Dr. Quené), does not use a glottal stop before nonce nouns with initial vowel. Therefore, future research on the perception of het will have to pay specific attention to contexts in which het is followed by nonce nouns with initial 130 vowel. If future studies propose a similar hierarchy of difficulty as the present study, they will have to make sure that a glottal stop occurs before nonce nouns with initial vowel. It is also striking that the second language learners only have difficulties with the stop-hetvowel contexts, and not with the nasal-het-vowel contexts and the fricative-het-vowel contexts as well. Apparently, het is easier to perceive in the nasal-het-vowel and the fricative-het-vowel contexts than in the stop-het-vowel contexts. Thus, especially a stop before het seems to cause difficulties with respect to the het-vowel contexts. Although /t/-deletion is improbable before vowels, it is possible that stops before het do cause /t/-deletion in het in the stop-het-vowel contexts, and that this is not (or to a lesser extent) the case for nasals and fricatives before het in the het-vowel contexts. T-deletion would weaken the phonological contrast between [də] and [ət] in the stop-het-vowel contexts. Future research will have to shed more light on the perception of het in stop-het-vowel contexts. If future studies propose a similar hierarchy of difficulty as the present study, they will have to make sure that het does not undergo /t/-deletion in stop-het-vowel contexts. Although the finding that the second language learners have difficulties with het in contexts in which a vowel occurs after het has not been predicted, this is not strange. Like all other contexts of het in the present study, the het-vowel and stop-het-vowel contexts also render het non-salient. Het does not occur in its full form but is reduced to [ət] in these contexts. It was not predicted that these contexts should be easy to perceive, but rather that they should be easier to perceive when compared to the other contexts of het. Moreover, as mentioned previously, it is possible that the second language learners perceive the /t/ of [ət] as the onset of the following nonce noun, or that /t/-deletion does occur in these contexts. Overall, however, the prediction that the second language learners should find het more difficult to perceive when het is followed by consonants than when followed by vowels does not appear to be borne out. Another unpredicted result is the finding that the second language learners have difficulties with the fricative-het contexts, i.e. the finding that the context before het also plays a role in the misperception of het. Because the second language learners have difficulties with the fricative-hetstop contexts, and not with the fricative-het-vowel contexts, especially a stop after het seems to cause difficulties with respect to the fricative-het contexts. This is also supported by the fact that the second language learners perceive the het-stop contexts (without t-het-t) significantly worse than the Dutch control group. A brief analysis of performance on the individual fricative-het-stop contexts reveals that the second language learners especially have difficulties with the context g-het-tiemp. The second language learners have a mean score of only 53% correct (SD = 50.7) for this context, while their mean scores for the other fricative-het-stop contexts are all 85% correct. Thus, the difficulties with the fricative-het and fricative-het-stop contexts especially appear to be due to the context g-het-tiemp. Apparently, the second language learners find it impossible to make out whether they hear de or het in this context. The difficulties with the context g-het-tiemp may also explain why the second language learners have difficulties with the fricative-het-stop contexts but not with the nasal-het-stop contexts and the stop-het-stop contexts, and why they have difficulties with the fricative-het contexts, but not with the stop-het and nasal-het contexts. The analysis of performance within the groups shows that the Dutch control group also has difficulties with the fricative-het-stop contexts. These difficulties also 131 seem to be due to the context g-het-tiemp. The Dutch controls have a mean score of 93% correct (SD = 25.8) for g-het-tiemp, while they score 100% correct on the other fricative-het-stop contexts. Finally, the analysis of performance on the stop-het-stop contexts shows that both the second language learners and the Dutch control group perceive the t-het-p/k contexts significantly better than the k-het-p/t contexts, i.e. the k-het-p/t contexts are perceived significantly worse than the t-het-p/k contexts. This indicates that both the Dutch control group and the second language learners have difficulties with the k-het-p/t contexts. Apparently, the combination k-het-p/t renders het more difficult to perceive than the combination t-het-p/k. The results of the present study indicate that the second language learners do not only have difficulties with het, but with dat as well. The Dutch control group perceives dat significantly better than the second language learners. This was not predicted. When the various factors are controlled for, however, it becomes clear that the second language learners' difficulties with dat are due to the performance of groups LE1 (the group which has had medium exposure) and P1 (the group with an average proficiency level) only, whereas most of the groups have difficulties with het. This means that the second language learners mainly have difficulties with het. Because groups LE2 and P2 do not have difficulties with dat, it appears that lengthier exposure and a higher proficiency level may lead to a better perception of dat, i.e. the misperception of dat may correspond to a lack of exposure and a lower proficiency level. Because the second language learners do not have difficulties with de, deze, die and dit, and because only groups LE1 and P1 have problems with dat, the second language learners do not show signs of having a general problem with Dutch gender. The second language learners rather show signs of having a problem with het specifically. Research Question 3 The third research question asked whether there is evidence for an effect of length and intensity of exposure in the second language learners' perception of het, i.e. whether second language learners who have had lengthy and intensive exposure to Dutch perceive het better than second language learners who have had little exposure. It was hypothesised and predicted that this should be the case. The results of the present study indicate that the factors length of exposure and intensity of exposure do not have an effect on the second language learners' perception of het. There are only differences between the short vs. long exposure groups for performance on dat and the test as a whole (without t-het-t). There are no differences between the medium vs. intensive exposure groups. In addition, no significant correlations have been found between length and intensity of exposure and performance. Thus, the predictions based on research question 3 are not borne out. It is surprising that length and intensity of exposure do not have an effect on the second language learners' perception of het. Second language learners who have had lengthy and intensive exposure should perceive het better than second language learners who have had little exposure, because they have gained more phonological knowledge about the reduction-and assimilation 132 processes which affect het in fast, informal speech. Some quick calculations reveal that 73.7% of the second language learners who have had lengthy exposure are late acquirers while 26.3% of them are early acquirers, that 31.6% of them have an average proficiency level while 68.4% of them have a high proficiency level, and that 52.6% of them do not use Dutch intensively while 47.4% of them do use Dutch intensively. These results suggest that age of first exposure may have an effect on the (mis)perception of het. It is possible that the second language learners who have had lengthy exposure do not perform better than those who have had little exposure because so many of them are late acquirers. Moreover, more than 50% of them do not use Dutch intensively. Similar calculations reveal that 62.5% of the second language learners who have had intensive exposure are late acquirers while 37.5% of them are early acquirers, that 25% of them have an average proficiency level while 75% of them have a high proficiency level, and that 43.8% of them have not had lengthy exposure while 56.3% of them have had lengthy exposure. In this case, too, age of first exposure may have an effect. Possibly, the second language learners who have had intensive exposure do not perform better than those who have had little exposure because so many of them are late acquirers. However, it should be observed that only the factor proficiency has been found to correlate with performance on het, and that the influence of the factors age of first exposure and intensity of exposure is not statistically corroborated. As a consequence, it is not possible to draw conclusions for the second language learner group as a whole as to why length of exposure and intensity of exposure do not have an effect on performance. 7.2 Production This section discusses the results of the production task in terms of research question 2 as posed in Chapter 3. It discusses whether the predictions based on research question 2 are borne out. The second research question asked whether the second language learners of Dutch can produce de and het correctly. It was asked whether they overgeneralise, and if they do, in what way (i.e. using de as the default or not). Observing previous research, it was hypothesised that the second language learners should show non-targetlike behaviour in the production task. It was predicted that they should mainly overgeneralise de, but that they may also use het with common nouns. It was also predicted that second language learners who have had lengthy and intensive exposure to Dutch may show targetlike behaviour in the production task (on both de and het). Learners who have had little exposure to Dutch should not show targetlike behaviour, however. Therefore, a significant difference between the performance of learners with lengthy and intensive exposure and that of learners with little exposure was predicted. A frequency effect was also predicted, i.e. learners should perform significantly better on high-frequency nouns than on medium-frequency nouns. The analysis of the individual results shows that three different response patterns are attested in the production data. The second language learners who show the targetlike pattern 1 consistently use de with common nouns and het with neuter nouns. Only a minority of the second language learners shows this pattern, however. The most frequent pattern is pattern 2. Learners who show this 133 pattern use de with common nouns and both de and het with neuter nouns. The second most frequent pattern is pattern 3. Learners who show this pattern use de and het with both common and neuter nouns. Patterns 2 and 3 together show that the vast majority of the second language learners uses de with neuter nouns. Pattern 3 shows that the learners also use het with common nouns. Because only a minority of the second language learners is targetlike, the prediction that they should show non-targetlike behaviour in the production task is borne out. The prediction that the learners should mainly overgeneralise de is also borne out, since the majority of them uses de with neuter nouns. The frequent use of de with neuter nouns points to a default strategy. The prediction that the learners should also use het with common nouns is borne out as well. The comparison of the groups analysis suggests that the factor length of exposure does not have an effect on production. No significant differences have been found between groups LE1 and LE2. Thus, the prediction that this should be the case is not borne out. However, the correlation analysis shows that the factor length of exposure correlates significantly with performance on het, the high frequency nouns, the high frequency het-nouns and the test as a whole. When length of exposure increases, performance on these test categories improves, i.e. the more exposure the second language learners have had, the better the test categories are produced. Because length of exposure correlates significantly with performance on het, the high frequency nouns, the high frequency hetnouns and the test as a whole, the prediction that length of exposure should have an effect on performance on the production task is partly borne out. The prediction is not fully borne out, since length of exposure only has an effect on the performance on het and not on performance on de. It was also predicted that the factor intensity of exposure should have an effect on performance. However, since there is only a significant difference between groups IEI and IE2 for performance on the medium frequency nouns, and since no significant correlations have been found between intensity of exposure and performance on the production task, this prediction does not seem to be borne out. It is surprising that length of exposure does not have an effect on the production of de, and that intensity of exposure does not have an effect on the production of de and het altogether. Second language learners who have had lengthy and intensive exposure should produce de and het better than learners who have had little exposure, because they have had more time and opportunity to store the gender of individual nouns in the mental lexicon. As mentioned in section 2.2, there are hardly any morphological and semantic regularities which could point to the gender of nouns in Dutch, and the regularities that exist have many exceptions (Donaldson, 1987: 27-33; Geerts et al., 1984: 41-49). Therefore, the acquisition of Dutch grammatical gender largely has to be done through word learning and learners need sufficient input in order to be able to acquire a targetlike grammatical gender system (Unsworth, in press). As mentioned in the previous section, the majority of the second language learners who have had lengthy and intensive exposure are late acquirers. This may explain why learners who have had lengthy exposure do not perform better on de, and why learners who have had intensive exposure do not perform better on both de and het than learners who have had little exposure. Table 155 shows that age of first exposure correlates with the production of het, but not with the production of de, however. Still, it should be observed that this correlation analysis includes all second language learners, and not just the late acquirers. As a consequence, it is not possible to draw 134 conclusions for the second language learner group as a whole as to why length of exposure does not have an effect on the production of de, and why intensity of exposure does not have an effect on production altogether. No significant differences have been found between the second language learners' production of the high-and medium frequency nouns. Thus, at first glance, frequency effects do not seem to occur. However, when the various factors influencing performance are controlled for, frequency effects are observed. When age of first exposure is controlled for, group AE1 performs significantly better on the medium frequency nouns and the medium frequency het-nouns than group AE2, but there are no differences between the groups for performance on the high frequency nouns and the high frequency het-nouns. Thus, the medium frequency nouns and the medium frequency het-nouns seem to be more difficult for late acquirers than the high frequency nouns and the high frequency het-nouns, and earlier acquisition seems to lead to a better production of the medium frequency nouns and the medium frequency het-nouns. When length of exposure is controlled for, the difference between groups LE1 and LE2 for performance on het and the medium frequency het-nouns is approaching significance, with group LE2 performing better than group LE1. This is not the case for performance on the high frequency het-nouns. Thus, the medium frequency het-nouns seem to be more difficult for learners who have had little exposure, and lengthy exposure seems to lead to a better production of the medium frequency het-nouns. Finally, when intensity of exposure is controlled for, group IE2 produces the medium frequency nouns significantly better than group IE1 and the difference between the groups for performance on both the medium frequency de-nouns and the medium frequency hetnouns shows a tendency towards significance, with group IE2 performing better than group IEI. Because this is not the case for the high frequency nouns, the medium frequency nouns (both the deand het-nouns) seem to be more difficult for learners who have had little exposure, and intensive exposure seems to lead to a better production of the medium frequency nouns. Consequently, the prediction that an effect of frequency should be visible in performance also appears to be borne out. 7.3 Perception and Production When the results of the perception experiment and the production task are combined, the results suggest that a better perception of het may lead to a better production of het and vice versa. Surprisingly, a better production of het also seems to lead to a better perception of de and the demonstratives. However, the way in which the relation between perception and production has been investigated here is of a rather basic nature, and a more thorough approach may be of more use in order to reveal the relationship between the second language learners' perception and production of het. For instance, this study has not investigated whether the phonological contexts the second language learners have difficulties with in perception are the same contexts they have difficulties with in production. Therefore, it is not certain that the second language learners' production difficulties are solely due to their misperception of het. Future research will have to determine the nature of the relationship between the second language learners' perception and production of het and the effect of phonological context on this relationship. 135 7.4 Theoretical Implications As stated in Chapter 2, previous studies into the acquisition of Dutch grammatical gender have shown that second language learners of Dutch have serious difficulties with the definite determiners, especially with neuter het. Second language learners of Dutch systematically show a delay in the acquisition of the definite determiners as well as signs of fossilisation in a non-targetlike stage of overgeneralisation (Hulk & Cornips 2006a,b; Cornips et al., 2006; Blom et al., in press; Unsworth, in press; Cornips & Hulk, in press; Brouwer et al., in press). Learners mainly overgeneralise de, but make errors in the other direction as well by also using het for common nouns (Cornips et al., 2006; Unsworth, in press). Unsworth observes that the use of het for common nouns may be a marker of bilingualism in Dutch. Unsworth's study suggests that lengthy and intensive exposure may lead to more targetlike responses in bilinguals, i.e. that the factors length and intensity of exposure have a significant effect on acquisition. Sabourin et al. (2006) have found that second language acquirers of Dutch gender produce the targetlike determiner more often for high frequency nouns than for low frequency nouns, i.e. a frequency effect is found. Brouwer et al. (in press), Hulk & Cornips (2006b) and Cornips & Hulk (in press) propose that the problems second language learners of Dutch experience with Dutch neuter gender may be caused by an underspecification of the grammatical gender feature, i.e. second language learners are aware of gender but do not have a complete knowledge and do not know the right gender specification (yet). Most of the previous studies are on production. The study by Brouwer et al. (in press) is the only study to date which focuses on comprehension. The present study has attempted to gain more insight into the role of comprehension in the second language acquisition of Dutch grammatical gender. It was hypothesised that the reason for second language learners' difficulties with Dutch neuter gender (het) may be a misperception of het caused by the non-saliency of het as induced by its phonological context. When het is systematically missed in perception, second language learners are likely to fail to store het in the mental lexicon and only de may be represented in the mental lexicon. This may lead to an underspecification of the grammatical gender feature in second language learners. This could explain second language learners' overgeneralisation of de. This proposal was put forward as the saliency hypothesis. The results of the present study indicate that second language learners of Dutch indeed have difficulties with the perception of het. Because the hypothesised hierarchy of difficulty is not entirely borne out in the results, the influence of phonological context on the misperception of het is not entirely clear. However, the hypothesised hierarchy of difficulty is largely borne out in the results: the second language learners find het more difficult to perceive when followed by /t/ than when followed by other consonants, and both the second language learners and the Dutch control group have serious difficulties with the t-het-t contexts. Although het is also difficult to perceive when followed by vowels, and the prediction that het should be more difficult to perceive when followed by consonants than when followed by vowels does not appear to be borne out, the second language learners also have difficulties with het when het is followed by consonants (i.e. with the het-stop and the fricativehet-stop contexts). Moreover, like all other contexts of het in the present study, the het-vowel and the stop-het-vowel contexts also render het non-salient, and the second language learners' misperception 136 of het in these contexts is therefore not surprising. Furthermore, the results of the perception experiment and the production task combined suggest that a better perception of het may lead to a better production of het. As to the question of whether the results of the present study support the saliency hypothesis, I therefore propose a tentative 'yes'. Because the results of the present study appear to support the saliency hypothesis, they also (indirectly) support the theory of an underspecification of the grammatical gender feature in second language learners as proposed by Brouwer et al. (in press), Hulk & Cornips (2006b) and Cornips & Hulk (in press). Furthermore, since the results suggest that a better perception of het may lead to a better production of het and vice versa, both perception and production appear to be important for the acquisition process. This is in line with what Llisterri (1995) and Rochet (1995) propose. The results of the present study also seem to support the proposal made by Holmes & Dejean de la Bâtie (1999), who state that lexical and sublexical cues are crucial to the process of gender acquisition. Holmes & Dejean de la Bâtie state that lexical information is likely to be most important, while sublexical information such as word endings indicating gender may reinforce the strength of gender knowledge in a postaccess procedure of gender confirmation. The production results of the present study show that the second language learners have problems with the acquisition of het. As the saliency hypothesis proposes, acquisition difficulties may be due to the fact that lexical gender cues (het, de) are to a large extent non-salient and may therefore be missed in perception. Holmes & Dejean de la Bâtie's proposal is supported by the finding that the saliency hypothesis appears to be borne out, and that a better perception of het may lead to a better production of het. It should also be observed that the present study has investigated the second language acquisition of Dutch grammatical gender by native speakers of English, while most of the previous research investigates the acquisition of Dutch grammatical gender by native speakers of languages other than English. Therefore, future research will have to determine whether the results of the present study are also borne out when the perception and production of second language learners who are native speakers of languages other than English are investigated. The production results show that the present study supports the previous research. The vast majority of the second language learners uses de with neuter nouns. This points to overgeneralisation by means of a default strategy. Almost one third of the learners also uses het with common nouns. The results also show an effect of frequency. When the factors influencing performance are controlled for, the medium frequency nouns appear to be more difficult for the second language learners than the high frequency nouns. A possible fossilisation in the non-targetlike stage of overgeneralisation has not been investigated in any depth here, but it is very well possible that some second language learners have indeed fossilised. Tables 146 and 151 suggest that there are learners who have received lengthy and intensive exposure and who are nonetheless non-targetlike in their production of de and het. With respect to the influence of the factors length of exposure, intensity of exposure, age of first exposure and proficiency, the following can be observed. Age of first exposure and proficiency appear to have an effect on perception. Earlier acquisition and a higher proficiency level seem to lead to a better perception of het. Furthermore, age of first exposure, proficiency and length of exposure appear to have an effect on production. Earlier acquisition and lengthier exposure seem to lead to a 137 better production of het, and a higher proficiency level seems to lead to a better production of both de and het. The finding that length of exposure has an effect on production is in line with Unsworth's finding (in press). Intensity of exposure does not appear to have an effect on performance, however. 8 Conclusion The present study has investigated the perception of the definite determiners het and de by second language learners of Dutch in order to determine whether they can perceive het despite its non-saliency as induced by its phonological context. A failure to perceive het is likely to lead to a failure to store het in the mental lexicon. Since de is mostly salient, only de may be represented in the mental lexicon. This may lead to an underspecification of the grammatical gender feature in second language learners, which could explain second language learners' overgeneralisation of de (saliency hypothesis). The present study has also investigated the second language learners' production capacities in order to determine whether they can correctly produce de and het. The results of the perception experiment indicate that the second language learners have difficulties with the perception of het. The second language learners perceive het significantly worse than the Dutch control group. Thus, the second language learners fail to perceive het on a regular basis, perceiving de instead and thereby 'overgeneralising' de in perception. The influence of phonological context on the misperception of het has extensively been investigated. The hypothesised hierarchy of difficulty is largely borne out in the results and the unpredicted results (difficulties with the het-vowel and the stop-het-vowel contexts) are unsurprising. Moreover, the results of the perception experiment and the production task combined suggest that a better perception of het may lead to a better production of het. For these reasons, the saliency hypothesis appears to be borne out. Because the hypothesised hierarchy of difficulty is not entirely borne out in the results, the influence of phonological context on the misperception of het is not entirely clear, however. Moreover, the present study has not investigated whether the phonological contexts the second language learners have difficulties with in perception are the same contexts they have difficulties with in production. Therefore, it cannot be stated with certainty that the acquisition difficulties the second language learners show in production are solely due to their misperception of het. The results of the perception experiment and the production task combined also suggest that a better production of het may lead to a better perception of het, i.e. possibly, learners first need to have knowledge about gender before they can perceive het in everyday speech. Further research will have to determine the influence of phonological context on the misperception of het, the nature of the relationship between the second language learners' perception and production of het and the effect of phonological context on this relationship. The results of the production task support the previous research. The vast majority of the second language learners uses de with neuter nouns, which suggests overgeneralisation by means of a default strategy. Almost one third of the second language learners also uses het with common nouns. When the various factors are controlled for, frequency effects are found; the medium frequency nouns seem to be more difficult for the second language learners than the high frequency nouns. 138 The perception results show that the factors age of first exposure and proficiency appear to have an effect on perception. Earlier acquisition and a higher proficiency level seem to lead to a better perception of het. The production results show that age of first exposure, proficiency and length of exposure appear to have an effect on production. Earlier acquisition and lengthier exposure seem to lead to a better production of het and a higher proficiency level seems to lead to a better production of both de and het. Finally, the present study has investigated the acquisition of Dutch grammatical gender by native speakers of English, but most of the previous research investigates the acquisition of Dutch grammatical gender by native speakers of languages other than English. Future research will therefore have to determine whether the results of the present study are also borne out when the perception and production of second language learners who are native speakers of languages other than English are investigated. 139 References Anderson, R.T. (1999). "Loss of gender agreement in L1 attrition: Preliminary results". In: Bilingual Research Journal 23 (4). pp. 319-38. Andersson, A.-B. (1992). Second Language Learner’s Acquisition of Grammatical Gender in Swedish. PhD thesis. University of Göteborg, Sweden. Baayen, H.R., Piepenbrock, R. & Van Rijn, H. (1993). The CELEX lexical data base. Dutch, English, German. Philadelphia, PA: Linguistics Data Consortium. Beck, M.L. (1998). "L2 acquisition and obligatory head movement: English-speaking learners of German and the Local impairment hypothesis". In: Studies in Second Language Acquisition 20 (3). pp. 311-48. Birdsong, D. (1999). "Whys and Why Nots of the Critical Period Hypothesis for Second Language Acquisition". In: D. Birdsong (Ed.), Second Language Acquisition and the Critical Period Hypothesis. Mahwah, NJ: Lawrence Erlbaum Associates. pp. 1-23. Birdsong, D. & Paik, J. (2008). "Second language acquisition and ultimate attainment". In: B. Spolsky & F. Hult (Eds.), Handbook of Educational Linguistics. London: Blackwell. pp. 424-437. Bley-Vroman, R. (1990). "The logical problem of foreign language learning". In: Linguistic Analysis 20. pp. 3-49. Blom, E., Polišenská, D. & Weerman, F. (2005). “Variation in Inflection: A comparison of agreement inflection in child L1, child L2 and adult L2 Dutch”. Paper presented at Variflex workshop, University of Amsterdam, 20th December. Blom, E., Polišenská, D. & Weerman, F. (in press). “Articles, Adjectives and Age of Onset: The Acquisition of Dutch Grammatical Gender”. In: Second Language Research. Boersma, P. & Weenink, D. (2007). PRAAT: doing phonetics by computer. Institute of Phonetic Sciences, University of Amsterdam, The Netherlands. <http://www.fon.hum.uva.nl/praat/>. Böhme, K. & Levelt, W.J.M. (1979). “Children’s use and awareness of natural and syntactic gender in possessive pronouns”. Paper presented at the International Reading Seminar on Linguistic Awareness and Learning to Read, Victoria, Canada. Bol, G.W. & Kuiken, F. (1988). Grammaticale analyse van taalontwikkelingsstoornissen. PhD dissertation, University of Amsterdam. Booij, G. (1995). The Phonology of Dutch. Oxford:Clarendon Press. Brouwer, S., Cornips, L. & Hulk, A. (in press). “Misrepresentation of Dutch neuter gender in older bilingual children?” In: B. Haznedar, and E. Gavruseva (Eds.), Current Trends in Child Second Language Acquisition: A Generative Perspective. Amsterdam: John Benjamins. 140 Bruhn de Garavito, J. & White, L. (2002). “The second language acquisition of Spanish DPs: The status of grammatical features”. In: A.T. Pérez-Leroux and J. Munoz Liveras (Eds.), The Acquisition of Spanish Morphosyntax. Dordrecht: Kluwer. pp.153-78. Cain, J., Weber-Olsen, M. & Smith, R. (1987). “Acquisition strategies in a first and second language: Are they the same?” In: Journal of Child Language 14 (2). pp. 333-52. Caramazza, A. (1997). "How many levels of processing are there in lexical access?" In: Cognitive Neuropsychology 14 (1). pp. 177-208. Carroll, S.E. (1989). “Second-language acquisition and the computational paradigm”. In: Language Learning 39 (4). pp. 535-94. Carroll, S.E. (1999). "Putting 'input' in its proper place". In: Second Language Research 15 (4). pp. 337-88. Carstens, V. (2000). "Concord in minimalist theory". In: Linguistic Inquiry 31 (2). pp. 319-55. Chierchia, G., Guasti, M.T. & Guilmini, A. (2001). Nouns and articles in child grammar and the syntax/semantics map. Ms. University of Milan, University of Siena, University of Maryland, College Park. Chomsky, N. (1980). Rules and Representations. New York: Columbia University Press and Oxford: Basil Blackwell Publishers. (Excerpted in The Behavioral and Brain Sciences 3 (1980). pp. 1-61). Clark, E. (1985). “The acquisition of Romance, with special reference to French”. In: D. Slobin (Ed.), The Crosslinguistic Study of Language Acquisition Volume 1: The Data. London: Lawrence Erlbaum Associates. pp. 255-372. Comrie, B. (1999). “Grammatical Gender Systems: A Linguist’s Assessment”. In: Journal of Psycholinguistic Research 28 (5). pp. 457-66. Corbett, G. (1991). Gender. Cambridge: Cambridge University Press. Cornips, L. & Hulk, A. (in press). "Factors of success and failure in the acquisition of grammatical gender in Dutch". In: Second Language Research. Cornips, L., van der Hoek, M. & Verwer, R. (2006). “The acquisition of grammatical gender in bilingual child acquisition of Dutch (by older Moroccan and Turkish children). The definite determiner, attributive adjective and relative pronoun”. In: B. Los & J van de Weijer (Eds.). Linguistics in The Netherlands 2006. Amsterdam: John Benjamins. pp 40-51. Corpus Gesproken Nederlands (CGN). Radboud University Nijmegen, Linguistics Department. Copyright 1 April 2004 De Nederlandse Taalunie. DeKeyser, R.M. (1998). "Beyond focus on form: Cognitive perspectives on learning and practicing second language grammar". In: C.J. Doughty & J. Williams (Eds.), Focus on Form in Classroom Second Language Acquisition. Cambridge: Cambridge University Press. pp. 42-63. 141 DeKeyser, R.M. (2001). "Automaticity and automatization". In: P. Robinson (Ed.), Cognition and Second Language Instruction. Cambridge: Cambridge University Press. pp. 125-51. Dell, G.S. (1986). "A spreading-activation model of retrieval in sentence production". In: Psychological Review 93. pp. 283-321. Dell, G.S. (1990). "Effects of frequency and vocabulary type on phonological speech errors". In: Language and Cognitive Processes 5. pp. 313-49. Desrochers, A. & Paivio, A. (1990). “Le phonème initial des noms inanimés et son effet sur l’identification du genre grammatical”. In: Revue Canadienne de Psychologie 49. pp. 240-62. Desrochers, A., Paivio, A. & Desrochers, S. (1989). “L’éffet de la fréquence d’usage des noms inanimés et de la valeur prédictive de leur terminasion sur l’identification du genre grammatical”. In: Revue Canadienne de Psychologie 43. pp. 62-73. Dewaele, J.-M. & Veronique, D. (2001). “Gender assignment and gender agreement in advanced French interlanguage: A cross-sectional study. In: Bilingualism: Language and Cognition 4 (3). pp. 275-97. Donaldson, B.C. (1987). Dutch Reference Grammar. Leiden: Martinus Nijhoff. Dörnyei, Z. (2003). Questionnaires in Second Language Research: Construction, Administration and Processing. Mahwah, NJ: Lawrence Erlbaum Associates. Ellis, N.C. (2002). "Frequency effects in language acquisition: A review with implications for theories of implicit and explicit language acquisition". In: Studies in Second Language Acquisition 24 (2). pp. 14388. Ellis, N.C. (2003). "Constructions, chunking, and connectionism: The emergence of second language structure". In: C.J. Doughty & M.H. Long (Eds.), The Handbook of Second Language Acquisition. Oxford: Blackwell. pp. 63-103. Ellis, N.C. (2005). "At the interface: Dynamic interactions of explicit and implicit language knowledge". In: Studies in Second Language Acquisition 27 (2). pp. 305-52. Eubank, L. & Gregg, K. (1999). "Critical periods and (second) language acquisition: divide et impera". In: D. Birdsong (Ed.), Second Language Acquisition and the Critical Period Hypothesis. Mahwah, NJ: Lawrence Erlbaum Associates. pp. 65-100. Favreau, M. & Segalowitz, N. (1983). "Automatic & controlled processes in the first-and secondlanguage reading of fluent bilinguals". In: Memory and Cognition 11 (1). pp. 56-74. Franceschina, F. (2005) Fossilised Second Language Grammars. The acquisition of grammatical gender. Amsterdam: John Benjamins. Garrett, M.F. (1988). “Processes in language production”. In: F.J. Newmeyer (Ed.), The Cambridge Survey of Linguistics, Vol. III: Biological and Psychological Aspects of Language. Cambridge, MA: Harvard University Press. pp. 69-96. 142 Garrett, M.F. (1992). “Disorders of lexical selection”. In: Cognition 42. pp. 143-80. Geerts, G. (1998). “Genusfouten: Hollanditis in Vlaanderen?” In: Verslagen en Mededelingen van de Koninklijke Academie voor Nederlandse Taal-en Letterkunde 1. pp. 68-78. Geerts, G., Haeseryn, W., Rooij, J.D. & Toorn, M.C.v.d. (1984). Algemene Nederlandse Spraakkunst. Groningen: Wolters-Noordhoff. Gillis, S. & de Houwer, A. (1998). “Dutch child language: An overview”. In: S. Gillis & A. de Houwer (Eds.), The Acquisition of Dutch. Amsterdam: John Benjamins. pp. 1-100. Goeman, T. (1999). T-deletie in de Nederlandse dialecten. Kwantitatieve analyse van structurele, ruimtelijke en temporele variatie. PhD thesis, Vrije Universiteit Amsterdam. Den Haag: Holland Academic Graphics. Gussenhoven, C. (1985). "Over de fonologie van de Nederlandse clitica". In: Spektator 15. pp. 180200. Hawkins, R. & Chan, C.Y.-H. (1997). "The partial availability of Universal Grammar in second language acquisition: The 'Failed functional features hypothesis'". In: Second Language Research 13. pp. 187-226. Hawkins, R. & Franceschina, F. (2004). "Explaining the acquisition of and non-acquisition of determiner-noun gender concord in French and Spanish". In: P. Prévost & J. Paradis (Eds.), The Acquisition of French in Different Contexts. Focus on Functional Categories. Amsterdam: John Benjamins. pp. 175-205. Holmes, V.M. & Segui, J. (2004). “Sublexical and Lexical Influences on Gender Assignment in French”. In: Journal of Psycholinguistic Research 33 (6). pp. 425-57. Holmes, V.M. & Segui, J. (2006). “Assigning Grammatical Gender During Word Production”. In: Journal of Psycholinguistic Research 35 (1). pp. 5-30. Holmes, V.M. & Dejean de la Bâtie, B. (1999). “Assignment of grammatical gender by native speakers and foreign learners of French”. In: Applied Psycholinguistics 20 (4). pp. 479-506. Houwer, A. de (1987). “Nouns and their companions: Or how a three-year-old handles the Dutch gender system”. In: Belgian Journal of Linguistics 2. pp. 55-73. Hulk, A. & Cornips, L. (2005). Bilingual acquisition and ‘fossilisation’ versus ‘acceleration’ of grammatical gender agreement of the definite article. Ms. Meertens Insitute/University of Amsterdam. Hulk, A. & Cornips, L. (2006a). “The boundaries between child L2 and 2L1: Do-support in child Dutch”. In: L Dekydtspotter, R.A. Sprouse & A. Liljestrand (Eds.). GASLA-7 Proceedings. Somerville, MA: Cascadilla. Hulk, A. & Cornips, L. (2006b). “Neuter gender and interface vulnerability in child L2/2L1 Dutch”. In: S. Unsworth, T. Parodi, A. Sorace, & M. Young-Scholten (Eds.), Paths of development in L1 and L2 Acquisition: In honor of Bonnie D. Schwartz. Amsterdam: John Benjamins. pp. 107-34. 143 Hulk, A. & Müller, N. (2000). “Bilingual first language acquisition at the interface between syntax and pragmatics”. In: Bilingualism: Language and Cognition 3. pp. 227-44. Hyltenstam, K. (1992). "Non-native features of near-native speakers". In: R.J. Harris (Ed.), Cognitive processes in bilinguals. Amsterdam: North Holland. pp. 351-68. Hyltenstam, K. & Abrahamsson, N. (2003). "Maturational constraints in SLA". In: C.J. Doughty & M.H. Long (Eds.), Handbook of Second Language Acquisition. Malden, MA: Blackwell. pp. 539-99. Jescheniak, J.-D. (1994). Word frequency effects in speech production. Unpublished doctoral dissertation, Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands. Johnson, E.K. (2004). “Grammatical gender and early word recognition in Dutch”. In: A. Brugos, M.R. Clark-Cotton & S. Ha (Eds.), Proceedings of the 29th Boston University Conference on Language Development, 2005. Somerville, MA: Cascadilla. pp. 320-30. Jongenburger, W. & van Heuven, V.J. (1991) "The distribution of (word-initial) glottal stop in Dutch". In: F. Drijkoningen & A. van Kemenade (Eds.), Linguistics in the Netherlands. Amsterdam: John Benjamins Publishing Company. pp. 101-110. Karmiloff-Smith, A. (1979). A Functional Approach to Child Language. A study of determiners and reference. Cambridge: Cambridge University Press. Kirkman, T.W. (1996). Statistics to Use. <http://www.physics.csbsju.edu/stats/>. (14 May 2008). Kooij, J.G. (1987). “Dutch”. In: B. Comrie (Ed.). The world’s major languages. London: Routledge. pp. 139-56. Lardiere, D. (2000). "Mapping features to forms in second language acquisition". In: J. Archibald (Ed.), Second Language Acquisition and Linguistic Theory. Oxford: Blackwell. pp. 102-29. Lenneberg, E. (1964). "The Capacity of Language Acquisition". In: J.A. Fodor & J.J. Katz (Eds.)., The structure of language. Englewood Cliffs, N.J.: Prentice-Hall. Lenneberg, E. (1967). Biological Foundations of Language. New York/London/Sydney: John Wiley & Sons, Inc. Levelt, W.J.M. (1989). Speaking: From Intention to Articulation. Cambridge, MA: MIT Press. Levelt, W.J.M. (1999). “Models of Word Production”. In: Trends in Cognitive Sciences 3 (6). pp. 22332. Levelt, W.J.M. (2001). “Spoken word production: A theory of lexical access”. In: Proceedings of the National Academy of the Sciences of the United States of America 98 (23). pp. 13464-471. Lightbown, P.M. & Spada, N. (2006). How Languages are Learned. Oxford: Oxford University Press. 3rd Ed. 144 Llisterri, J. (1995). “Relationships between speech production and speech perception in a second language”. In: K. Elenius and P. Branderud (Eds.), Proceedings of the XIIIth International Congress of Phonetic Sciences 4. Stockholm, Sweden, 13-19 August 1995. Stockholm: KTH / Stockholm University. pp. 92-99. MacWhinney, B. (1978). “The acquisition of morphophonology”. In: Mongraphs of the Society for Research in Child Development 43 (1-2, Serial no. 174). Meisel, J. (2007). "Exploring the limits of the LAD". In: Working papers in multilingualism 80. Hamburg: Hamburg University. pp. 3-31. Mills, A.E. (1986). The Acquisition of Gender: Evidence from German and English. Berlin: Springer. Mitterer, H. & Ernestus, M. (2006). "Listeners recover /t/s that speakers reduce: Evidence from /t/lenition in Dutch". In: Journal of Phonetics 34. pp. 73-103. Mϋller, N. (1990). “Developing two gender assignment systems simultaneously”. In: J.M. Meisel (Ed.), Two First Languages: Early Grammatical Development in Bilingual Children. Dordrecht: Foris. pp. 194232. Müller, N. & Hulk, A. (2001). “Cross-linguistic influence in bilingual first language acquisition: Italian and French as recipient languages”. In: Bilingualism: Language and Cognition 4. pp. 1-21. Paradis , J. & Genesee, F. (1995). “Language differentiation in early bilingual development”. In: Journal of Child Language 22 (3). pp. 611-31. Prévost, P. & White, L. (2000). "Missing surface in inflection or impairment in second language acquisition? Evidence from tense and agreement". In: Second Language Research 16 (2). pp. 103-33. RANDOM.ORG Random Sequence Generator. Ed. Dr. M. Haahr. 1998-2008. Department of Computer Science Trinity College Dublin, Ireland. March 2008. <http://www.random.org/sequences/>. Rochet, B.L. (1995). “Perception and production of L2 speech sounds by adults”. In: W. Strange (Ed.), Speech Perception and Linguistic Experience: Theoretical and Methodological Issues in CrossLanguage Speech Research. Timonium, MD: York Press Inc. pp. 379-410. Rogers, M. (1987). "Learner's difficulties with grammatical gender in German as a foreign language". In: Applied Linguistics 8. pp. 48-74. Sabourin, L. (2003). Grammatical Gender and Second Language Processing: an ERP Study. Doctoral Dissertation, Groningen University, The Netherlands. Groningen: Groningen University Press. Sabourin, L. & Haverkort, M. (2003). “Neural substrates of representation and processing of a second language”. In: R. van Hout, A. Hulk, F. Kuiken & R.J. Towell (Eds.), The lexicon-syntax interface in second language acquisition. Amsterdam: John Benjamins. pp. 175-95. 145 Sabourin, L., Stowe, L.A. & de Haan, G. (2006). “Transfer effects in learning a second language grammatical gender system”. In: Second Language Research 22 (1). pp. 1-29. Schachter, J. (1996). "Maturation and the issue of Universal Grammar in second language acquisition". In: V. Ritchie & T.K. Bhatia (Eds.), Handbook of Second Language Acquisition. San Diego: Academic Press. pp. 159-93. Schriefers, H. & Jescheniak, J.D. (1999). “Representation and Processing of Grammatical Gender in Language Production: A Review”. In: Journal of Psycholinguistic Research 28 (6). pp. 575-600. Segalowitz, N. (2003). "Automaticity". In: C.J. Doughty & M.H. Long (Eds.). The Handbook of Second Language Acquisition. Oxford: Blackwell. pp. 382-408. Singleton, D. & Ryan, L. (2004). Language Acquisition: The Age Factor. 2nd Ed. Clevedon: Multilingual Matters Ltd. Strozer, J.R. (1994). Language Acquisition After Puberty. Washington D.C: Georgetown University Press. Taft, M. & Meunier, F. (1998). “Lexical representation of gender: A quasi-regular domain”. In: Journal of Psycholinguistic Research 27 (1). pp. 23-45. Ullmann, M.T. (2001a). "A neurocognitive perspective on language: the declarative/procedural model". In: Nature Reviews (Neuroscience) 2. pp. 717-26. Ullmann, M.T. (2001b). "The neural basis of lexicon and grammar in first and second language: the declarative/procedural model". In: Bilingualism: Language and Cognition 4 (1). pp. 105-22. Ullmann, M.T. (2004). "Contributions of memory circuits to language: the declarative/procedural model". In: Cognition 92 (1-2). pp. 231-70. Umbel, V. & Oller, D.K. (1995). "Developmental changes in receptive vocabulary in Hispanic bilingual school children". In: B. Harley (Ed.), Lexical Issues in Language Learning. Amsterdam: John Benjamins. pp. 29-80. Unsworth, S. (in press). "Age and input in the acquisition of grammatical gender in Dutch". In: Second Language Research. Van Berkum, J.J.A. (1996). The psycholinguistics of grammatical gender: Studies in language comprehension and production. PhD Dissertation, Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands. Nijmegen: Nijmegen University Press. Van Berkum, J.J.A (1997). “Syntactic processes in speech production: the retrieval of grammatical gender”. In Cognition 64 (2). pp. 115-52. Vandeputte, O., Vincent, P. & Hermans, T. (1991). Dutch: The language of twenty million Dutch and Flemish people. Rekkem, Belgium: Stichting Ons Erfdeel. 146 Van der Velde, M. (2003). Déterminants et pronoms en Néerlandais et en Français: Syntaxe en acquisition. Phd thesis, Université Paris 8. Van der Velde, M. (2004). “L’Acquisition des déterminants en L1: Une étude comparative entre le Français et le Néerlandais”. In: Acquisition et Interaction en Langue Etrangère 21. pp. 9-46. Verrips, M. & Wijnen, F. (1998). "The acquisition of Dutch syntax". In: S. Gillis & A. De Houwer (Eds.), The Acquisition of Dutch. Amsterdam: John Benjamins. pp. 223-300. Vygotsky, L.S. (1978). Mind in Society. Cambridge, MA: Harvard University Press. Weerman, F., Bisschop, J. & Punt, L. (2002). "L1 and L2 Acquisition of Dutch Adjectival Flexion". Ms., Universiteit van Amsterdam. White, L., Valenzuela, E., Kozlowska-MacGregor, M. & Leung, I.Y.-K. (2004). “Gender and number agreement in nonnative Spanish”. In: Applied Psycholinguistics 25. pp. 105-33. Wijnen, F. & Deutsch, W. (1987). “The recognition of grammatical gender in Dutch”. Paper presented at the Tagung experimentell arbeitender Psychologen, Aachen, April 13. Zonneveld, R. van (1992). “Het jonge hoofd – De righthand head rule bij kinderen van 4 tot 7 jaar”. In: De nieuwe taalgids 85 (1). pp. 37-49. 147 Appendices Appendix 1 The nonce nouns used in the perception experiment T- (35) K-/P- (18/18) VOC- (37) taaf taag taam taar taat tacht tarp tauk taus techt terk teuk tieg tiemp tenk tift tilk tilp tils tir tirf tirs toem toft tolg tolm tolp tomp tuip tulg tung tups turp turs tuuf tuup kacht kag kecht keup kieg kijn kijp kilg koem koft kolg kolm kuft kuig kump kurp kuum kuun paaf pank parp peef pelk penk perg picht pijt pir pirk polt porf puig pung purg purt puuk aag ak aln eep eft eps ers euk eus icht ief ieg iens ies ift ijn ilg ilk irk ocht oem oerk oes olk olp omp oos org oup uim ulg ulp ump unk upt uuk irf (originally: tirf) 148 Appendix 2 The manner adverbs used in the perception experiment -X (18) -K (18) -M (18) -N (18) -T (36) angstig voorzichtig kinderachtig koppig behendig bereidwillig eenvoudig geduldig nieuwschierig duizelig vakkundig chagrijnig zenuwachtig eigenaardig schaapachtig wanhopig droevig hardhandig aanstekelijk afstandelijk energiek fatsoenlijk mogelijk ambachtelijk hartstochtelijk schriftelijk misselijk feestelijk gemeenschappelijk begrijpelijk geleidelijk bedenkelijk gemakkelijk duidelijk fanatiek gevaarlijk aangenaam langzaam spaarzaam stiekem ad rem anoniem eenzaam moeizaam unaniem behulpzaam kalm opmerkzaam vreedzaam gehoorzaam waakzaam bedachtzaam verdraagzaam behoedzaam aangedaan alleen voldaan modern clandestien beeldschoon ontdaan synchroon voortaan spontaan achteraan gewoon schuin monotoon gemeen meteen begaan haarfijn afgemat gekrenkt gericht verhit stipt nonchalant gerust gedempt efficiënt ingezakt onopgemaakt onopgemerkt resoluut nauwgezet naakt elegant decadent charmant zacht verontrust onbewust hypocriet beslist welbespraakt arrogant impliciet belust beheerst opgejut verdacht intelligent bewust verrukt ontzet attent enthousiast 149 English translation of the manner adverbs -X (18) -K (18) -M (18) -N (18) -T (36) frightened carefully childishly stubbornly dexterously willingly easily patiently curiously dizzy skilfully sulkily nervously strangely sheepishly desperately sorrowfully roughly infectiously standoffishly energetically decently possibly traditionally passionately in writing sickly festively jointly understandably gradually gravely easily lucidly fanatically dangerously pleasantly slowly sparingly stealthily quick-wittedly anonymously lonely laboriously unanimously obligingly calmly carefully peacefully obediently watchfully thoughtfully peacefully cautiously moved alone contently modernly illicitly beautifully thorougly upset synchronically from now on spontaneously last in line just disapprovingly monotonously wickedly immediately sympathetically minutely exhausted offendedly purposefully heatedly precisely carelessly safely subdued efficiently bent without make-up on unnoticedly resolutely scrupulously naked elegantly decadently charmingly softly uneasily unconsciously hypocritically certainly eloquently arrogantly implicitly keen calmly pent up suspiciously intelligently deliberately enrapturedly aghast considerately enthusiastically 150 Appendix 3 The test sentences used in the perception experiment (English translation included) DE (the) 1. Bang voor infecties, riep zij angstig de tieg terug. Afraid of infections, called she frightened the tieg back. (She was afraid of infections, and called back the tieg.) 2. Antiquairs geven altijd voorzichtig de koem terug aan hun klanten. Antiquarians give always carefully the koem back to their customers. (Antiquarians always carefully give the koem back to their customers.) 3. Zij negeerde kinderachtig de iens om zijn gedrag. She ignored childishly the iens because of his/its behaviour. (She childishly ignored the iens because of his/its behaviour.) 4. Om opstandig te zijn, verwijderde hij koppig de ilg uit zijn moeders computer. In order to rebellious be, removed he stubbornly the ilg from his mother’s computer. (As an act of rebellion, he stubbornly removed the ilg from his mother’s computer.) 5. Zij lachte zeer aanstekelijk de tilk toe. She laughed very infectiously the tilk at. (She laughed very infectiously at the tilk.) 6. Hij groet afstandelijk de irf, omdat hij haar eigenlijk niet aardig vindt. He greets standoffishly the irf, because he her in reality not like does. (He standoffishly greets the irf, because in reality he does not like her.) 7. Door haar ADHD rende ze energiek de kuft rond. Because of her ADHD ran she energetically the kuft around. (Because of her ADHD she ran energetically around the kuft.) 8. Vincent ruimde fatsoenlijk de pelk op voordat zijn date langs zou komen. Vincent cleared decently the pelk away before his date by would come. (Vincent decently cleared the pelk away before his date would come by.) 9. Hij heeft mogelijk de oem verdraaid en moet nu rust houden. 151 He has possibly the oem twisted and has to now rest. (He has possibly twisted the oem and has to rest now.) 10. Marga van Praag kondigde aangenaam de turp af. Marga van Praag proclaimed pleasantly the turp. (Marga van Praag pleasantly proclaimed the turp.) 11. Terroristen vernietigen langzaam de porf van alle presidenten. Terrorists destroy slowly the porf of all presidents. (Terrorists slowly destroy the porf of all presidents.) 12. Zij heeft altijd spaarzaam de oup bewaard, om later aan haar zoon te geven. She has always sparingly the oup kept, in order to later on to her son give. (She has always sparingly kept the oup, in order to give to her son later on.) 13. Suzanne kuste stiekem de ump tijdens wiskundeles. Suzanne kissed stealthily the ump during math class. (Suzanne stealthily kissed the ump during math class.) 14. Een klein wit hondje stapte aangedaan de tolg uit. A little white dog stepped moved the tolg out of. (A little white dog stepped moved out of the tolg.) 15. Een oude man liep alleen de tuip in om een boek uit te zoeken. An old man walked alone the tuip into in order to a book select. (An old man walked alone into the tuip in order to select a book.) 16. Zij levert voldaan de kuum in en verwacht een goed cijfer. She hands contently the kuum in and expects a good mark. (She contently hands in the kuum and expects a good mark.) 17. Zij had modern de pijt aangekleed in een prachtige jurk van Prada. She had modernly the pijt dressed in a magnificent dress by Prada. (She had modernly dressed the pijt in a magnificent dress by Prada.) 18. Drankhandelaren zijn clandestien de olp in gaan kopen. Beverage dealers have illicitly the olp started to purchase. (Beverage dealers have illicitly started to purchase the olp.) 152 19. Pinkeltje wandelde afgemat de tacht binnen met een zak vol stenen. Pinkeltje walked exhausted the tacht into with a bag full of stones. (Pinkeltje walked exhausted into the tacht with a bag full of stones.) 20. Opeens liet Mina gekrenkt de taat vallen en huilde. Suddenly Mina offendedly the taat dropped and cried. (Taking offence, Mina suddenly dropped the taat and cried.) 21. Zij probeert altijd gericht de taus verlegen te maken, wat erg onaangenaam is. She tries always purposefully the taus to embarrass, which very unpleasant is. (She always tries to embarrass the taus purposefully, which is very unpleasant.) 22. Hij begon verhit de kacht te slaan. He began heatedly the kacht to hit. (He heatedly began to hit the kacht.) 23. Gewoonlijk kijken postbezorgers stipt de kieg na. Usually check postal workers precisely the kieg. (Postal workers usually check the kieg in a precise way.) 24. Eva handelde nonchalant de paaf af en reed weg in haar auto. Eva settled carelessly the paaf and drove off in her car. (Eva carelessly settled the paaf and drove off in her car.) 25. Je kunt gerust de aag natmaken, want hij kan er tegen. You can safely the aag moisten, because he/it can it take. (You can safely moisten the aag, because he/it can take it.) 26. Wij zullen gedempt de eft inspreken, omdat er geluidsoverlast is. We will in a subdued voice the eft record, because there noise nuisance is. (We will record the eft in a subdued voice, because there is noise nuisance.) 27. Dagelijks maken studenten efficiënt de eus om wat bij te verdienen. Every day make students efficiently the eus in order to some make extra money. (Every day students efficiently make the eus in order to make some extra money.) 153 HET (the) 1. Hij ving behendig het tiemp op, zodat er geen schade was. He caught dexterously the tiemp, so there no damage was. (He dexterously caught the tiemp, thereby preventing damage.) 2. Een aantal mensen hielp bereidwillig het tilp om extra omzet te krijgen. A number of people helped willingly the tilp extra sales to gain. (A number of people willingly helped the tilp to gain extra sales.) 3. Omdat hij ervoor geleerd heeft, bouwt hij eenvoudig het kijn in elkaar. Because he for it studied has, builds he easily the kijn together. (Because he has studied for it, he easily builds the kijn together.) 4. Een grote groep krakers kondigde geduldig het pir aan. A large group of squatters announced patiently the pir. (A large group of squatters patiently announced the pir.) 5. Hij test nieuwsgierig het ies uit op zijn werk. He tests curiously the ies out at work. (He curiously tests out the ies at work.) 6. Daar wordt ambachtelijk het tirs gemaakt waar men dol op is. Over there is traditionally the tirs made which everybody loves. (Over there the tirs which everybody loves is traditionally made.) 7. Zij kuste hartstochtelijk het puig voordat hij weer vertrokken was. She kissed passionately the puig before he again left. (She passionately kissed the puig before he left again.) 8. Als twee mensen gaan trouwen, moet schriftelijk het ilk vastgelegd worden. When two people get married, has to be in writing the ilk laid down. (When two people get married, the ilk has to be laid down.) 9. Stefan moest misselijk het oerk uitspugen, waardoor hij niet meer kon leren. Stefan had to sickly the oerk spit out, because of which he not anymore could study. (Stefan had to spit out the oerk as he became sick, and could not study anymore as a result.) 154 10. Zij riep ad rem het tulg terug om nog één en ander uit te leggen. She called quick-wittedly the tulg back in order to a few things explain. (She quick-wittedly called back the tulg in order to explain a few things.) 11. Veel zwervers wonen anoniem het turs bij. Many tramps attend anonymously the turs. (Many tramps anonymously attend the turs.) 12. Zij kijkt elke nacht eenzaam het kuig over. She looks every night lonely the kuig out on. (Every night she lonely looks out on the kuig.) 13. Veel leerlingen bestuderen moeizaam het purt voor vrijdag. Many students study laboriously the purt for Friday. (Many students laboriously study the purt for Friday.) 14. Bij nieuwe verkiezingen zullen zij unaniem het unk wegstemmen. During new elections will they unanimously the unk vote against. (They will unanimously vote against the unk during new elections.) 15. Sylvia zong beeldschoon het tolm, maar won helaas geen beker. Sylvia sang beautifully the tolm, but did win unfortunately not a trophy. (Sylvia beautifully sang the tolm, but unfortunately she did not win a trophy.) 16. Moeder omhelsde ontdaan het koft op een middag in mei. Mother hugged thoroughly upset the koft on an afternoon in May. (Thoroughly upset, mother hugged the koft on an afternoon in May.) 17. Eva en Esther deden synchroon het omp na en dansten daarna moeiteloos verder. Eva and Esther imitated synchronically the omp and danced after that effortlessly on. (Eva and Esther synchronically imitated the omp and danced on effortlessly after that.) 18. Tim zal voortaan het uim doen, omdat hij erin geïnteresseerd is. Tim will from now on the uim do, because he in it interested is. (From now on Tim will do the uim, because he is interested in it.) 19. Zij verlaat ingezakt het tef en kijkt niet meer om. She leaves bent the tef and does look not anymore back. 155 (She leaves the tef with her back bent and does not look back anymore.) 20. Zaterdagavond kwam Aïcha onopgemaakt het taaf in en bestelde een cola. Saturday night came Aïcha without make-up on the taaf into and ordered a coke. (Saturday night Aïcha came into the taaf without make-up on and ordered a coke.) 21. Vader liet onopgemerkt het techt los in een vlaag van medelijden. Father let unnoticedly the techt go of in a fit of compassion. (Unnoticedly, father let go of the techt in a fit of compassion.) 22. Zonder nadenken stuurt oma resoluut het kag terug. Without thinking sends grandmother resolutely the kag back. (Without thinking grandmother resolutely sends the kag back.) 23. Elke dag doet opa nauwgezet het pank om scherp te blijven. Every day does grandfather scrupulously the pank in order to sharp stay. (Every day grandfather scrupulously does the pank in order to stay sharp.) 24. Zij weet naakt het parp veruit te overtreffen. She can naked the parp by a long shot surpass. (Naked she can surpass the parp by a long shot.) 25. Zij sprak elegant het ak toe en had veel succes. She addressed elegantly the ak and had much success. (She addressed the ak elegantly and had much success.) 26. Elaine roept decadent het eps bij zich wanneer er betaald moet worden. Elaine calls decadently the eps for when the bill paid has to be. (Elaine decadently calls for the eps when the bill has to be paid.) 27. Mijn vriendin palmde charmant het icht in. My friend won over charmingly the icht. (Because she acted charmingly, my friend won over the icht. DAT (that) 1. Veel mensen kwamen duizelig dat toft uit. Many people came dizzy that toft out of. (Many people came dizzy out of that toft.) 156 2. Schilders verven vakkundig dat ijn in een mooie kleur. Painters paint skilfully that ijn in a beautiful colour. (Painters skilfully paint that ijn in a beautiful colour.) 3. Mijn vriendin heeft feestelijk dat kurp versierd voor hun bruiloft. My friend has festively that kump decorated for their wedding. (My friend has festively decorated that kump for their wedding.) 4. Een aantal studenten test gemeenschappelijk dat olk uit, om te kijken of er nog fouten in zitten. A number of students tests jointly that olk, in order to see whether it still mistakes shows. (A number of students jointly tests that olk, in order to see whether it still shows mistakes.) 5. Marinus reikte behulpzaam dat tuup aan, zodat men weer verder kon werken. Marinus passed obligingly that tuup, so everybody could continue with the work. (Marinus obligingly passed that tuup, and everybody could continue with the work.) 6. Alle gevangenisbewaarders bewaken kalm dat kuun tegen opstand. All prison wardens guarded calmly that kuun against an uprising. (All prison wardens calmly guarded that kuun against an uprising.) 7. Diverse agenten luisteren dagelijks opmerkzaam dat purg af om boeven op te sporen. Various agents listen daily carefully that purg to in order to criminals find. (Every day, various agents carefully listen to that purg in order to find criminals.) 8. Op Kerstavond zingen zij vreedzaam dat uuk toe. On Christmas Eve sing they peacefully that uuk for. (On Christmas Eve they peacefully sing for that uuk.) 9. Regelmatig belt zij spontaan dat tups op om een afspraak te maken. Regularly calls she spontaneously that tups in order to an appointment make. (She has made a habit of spontaneously calling that tups to make an appointment.) 10. Ze reden achteraan dat ulp af, waardoor ze te laat thuis waren. They drove last in line that ulp down, because of which they late home were. (They drove down that ulp last in line, because of which they were home late.) 157 11. Midas Dekkers fluisterde zacht dat tauk toe op een mooie zomerdag. Midas Dekkers whispered softly that tauk to on a beautiful summer day. (Midas Dekkers whispered softly to that tauk on a beautiful summer day.) 12. Na een tijdje wachten ging zij verontrust dat tift controleren. After a while waiting went she uneasily that tift to check. (After she had waited a while, she uneasily went to check that tift.) 13. Tijdens zwemles zette zij onbewust dat kilg op stand zeven. During the swimming lesson turned she unconsciously that kilg to pitch seven. (During the swimming lesson she unconsciously turned that kilg to pitch seven.) 14. Hij gelastte hypocriet dat picht af, wat veel kritiek oogstte. He cancelled hypocritically that picht, which much criticism caused. (He hypocritically cancelled that picht, which caused much criticism.) 15. Politici zullen beslist dat euk in orde maken. Politicians will certainly that euk in order put. (Politicians will certainly put that euk in order.) DEZE (this) 1. Omdat hij straf had, schreef hij chagrijnig deze teuk over. Because he punished was, copied he sulkily this teuk. (Because he was punished, he sulkily copied this teuk.) 2. Hij nam zenuwachtig deze ocht over van zijn broer, maar wist niet wat hij ermee moest doen. He took nervously this ocht from his brother, but know did not what with it to do. (He nervously took this ocht from his brother, but did not know what to do with it.) 3. Marjan veranderde begrijpelijk deze tir om een beter uitzicht te hebben. Marjan changed understandably this tir in order to a better view to have. (Marjan understandably changed this tir to have a better view.) 4. Een visagist bracht geleidelijk deze polt aan op haar gezicht. A make-up artist applied gradually this polt to her face. (A make-up artist gradually applied this polt to her face.) 158 5. Kinderen imiteren elke dag gehoorzaam deze kolm op televisie. Children imitate every day obediently this kolm on television. (Every day, children obediently imitate this kolm on television.) 6. Fiona sluipt waakzaam deze upt door en kijkt voorzichtig om zich heen. Fiona sneaks watchfully this upt through and looks carefully her around. (Fiona watchfully sneaks through this upt and carefully looks around her.) 7. Ik zal gewoon deze tomp even bijvullen. I will just this tomp refill. (I will just refill this tomp.) 8. Tante Els maakte er een gewoonte van om schuin deze puuk na te kijken. Aunt Els made a habit of disapprovingly this puuk after looking. (Aunt Els made a habit of disapprovingly looking after this puuk.) 9. Mijn docent vertelt altijd heel monotoon deze org aan ons, zodat we er niks van snappen. My teacher tells always very monotonously this org to us, so we cannot a thing understand. (My teacher always very monotonously tells this org to us, so we cannot understand a thing.) 10. Zij zit welbespraakt deze taar bevelen te geven en eet alleen maar mandarijntjes. She sits eloquently this taar orders to give and only eats mandarins. (She is eloquently giving orders to this taar and only eats mandarins.) 11. Gijsbrecht oppert arrogant deze keup te laten activeren. Gijsbrecht suggests arrogantly this keup to have deactivated. (Gijsbrecht arrogantly suggests to have this keup deactivated.) 12. Zij beledigde impliciet deze peef wat haar niet in dank afgenomen werd. She offended implicitly this peef, which people clearly blamed her for. (She implicitly offended this peef, which people clearly blamed her for.) 13. Kim is erop belust deze eep te laten mislukken. Kim is keen on this eep the failure of. (Kim is keen on the failure of this eep.) 14. En paar oude heren spraken beheerst deze ieg uit en gingen naar huis. A few old gentleman pronounced calmly this ieg and went home. 159 (A few old gentlemen calmly pronounced this ieg and went home.) DIE (that) 1. Lisa bedankte eigenaardig die kecht voor zijn hulp. Lisa thanked strangely that kecht for his/its help. (Lisa strangely thanked that kecht for his/its help.) 2. Hij keek schaapachtig die pirk aan, maar kreeg geen reactie. He looked sheepishly that pirk at, but did get not a reaction. (He sheepishly looked at that pirk, but did not get a reaction.) 3. Tijdens een operatie sneed Peter wanhopig die ief door om een teken van leven te krijgen. During an operation cut Peter desperately that ief in order to a sign of life get. (During an operation Peter desperately cut that ief in order to get a sign of life.) 4. Om op een idee te komen draait hij bedenkelijk die tils om. In order to an idea hit upon turns he gravely that tils round. (In order to hit upon an idea he gravely turns that tils round.) 5. Hij hield gemakkelijk die irk tegen, waardoor er geen slachtoffers vielen. He stopped easily that irk, because of which there no victims were. (He easily stopped that irk, thereby preventing casualties.) 6. Hij beëindigde bedachtzaam die tuuf, omdat hij er genoeg van had. He thoughtfully ended that tuuf, because he it was fed up with. (He thoughtfully ended that tuuf, because he was fed up with it.) 7. Zij leven verdraagzaam die oos na, zoals van oudsher voorgeschreven staat. They live peacefully that oos according to, as from time immemorial dictated is. (They peacefully live according to that oos, as is dictated from time immemorial.) 8. Wij knijpen gemeen die tolp in zijn bil. We pinch wickedly that tolp in his buttock. (We wickedly pinch that tolp in his buttock.) 9. Ze schakelde meteen die kolg uit toen iemand aanbelde. 160 She turned immediately that kolg off when someone rang the bell. (She immediately turned off that kolg when someone rang the bell.) 10. Mijn zus liep opgejut die taam tegemoet en werd boos op oma. My sister walked pent up that taam towards and became angry with grandmother. (My sister walked towards that taam in a pent up state and became angry with grandmother.) 11. Een sjofele man rent verdacht die terk uit en agenten gaan er achteraan. A shabby man runs suspiciously that terk out of and constables go him after. (A shabby man suspiciously runs out of that terk and constables go after him.) 12. Kleine kinderen speelden intelligent die penk na en kregen veel applaus. Little children played intelligently that penk after and received much applause. (Little children intelligently played after that penk and received much applause.) 13. Wij proberen bewust die arf te negeren. We try deliberately that arf to ignore. (We deliberately try to ignore that arf.) DIT (this) 1. Drie meisjes verzorgen droevig dit tuul op zijn sterfbed. Three girls take care of sorrowfully this tuul on his deathbed. (Full of sorrow, three girls take care of this tuul on his deathbed.) 2. Een oudere jongen mishandelde hardhandig dit perg, maar hij werd niet opgepakt. An older boy ill-treated roughly this perg, but he was not arrested. (An older boy roughly ill-treated this perg, but he was not arrested.) 3. Hij heeft duidelijk dit toem uitgetekend, zodat we weten wat we moeten doen. He has lucidly this toem drawn, so we know what we have to do. (He has lucidly drawn this toem, and therefore we know what we have to do.) 4. Omdat hij wilde winnen, speelde hij erg fanatiek dit kump mee. Because he wanted to win, joined he very fanatically this kump in. (Because he wanted to win, he very fanatically joined in this kump.) 5. Om stoer te doen, sprong hij gevaarlijk dit oes af en hield er een zeer been aan over. 161 In order to show off, jumped he dangerously this oes off and hurt his leg. (While showing off, he dangerously jumped off this oes and hurt his leg.) 6. Om niemand te kwetsen heeft hij behoedzaam dit tung uitgesproken. In order to nobody offend has he cautiously this tung expressed. (In order to prevent offence he has cautiously expressed this tung.) 7. Hij nodigde erg begaan dit pung uit bij hem thuis. He invited very sympathetically this pung at his home. (He very sympathetically invited this pung at his home.) 8. Joris zal even haarfijn dit ulg uitleggen, zodat we ermee kunnen beginnen. Joris will now minutely this tulg explain, so we on it can start. (Now Joris will minutely explain this tulg, so we can start on it.) 9. Moeder komt verrukt dit tarp binnen en gaat alles bekijken. Mother comes enrapturedly this tarp into and begins to everything examine. (Mother enters this tarp in an enraptured state and begins to examine everthing.) 10. Alle toeschouwers renden ontzet dit kijp uit. All spectators ran aghast this kijp out of. (All spectators were aghast and ran out of this kijp.) 11. Mijn buurvrouw bracht attent dit ers weer terug. My neighbour brought considerately this ers again back. (My neighbour considerately brought back this ers again.) 12. Voor zijn afstuderen heeft Michel enthousiast dit ift ontworpen. For his graduation has Michel enthusiastically this ift designed. (Michel has enthusiastically designed this ift for his graduation.) 162 Appendix 4 INSTRUCTIES Je gaat een experiment doen waarin je kennis van de Nederlandse taal getest wordt. Lees de instructies goed door en vraag om mondelinge toelichting als iets niet duidelijk is. 1 Boven in beeld staat een nonsenswoord, in dit geval fleng. Je krijgt een Nederlandse zin te horen waarin dit woord verwerkt is. In de gele blokjes staan de Nederlandse lidwoorden de en het. Klik met de muis aan welk lidwoord je gehoord hebt vóór het nonsenswoord. Er zijn ook zinnen waarin een aanwijzend voornaamwoord (demonstrative) wordt gebruikt in plaats van een lidwoord. Kies hier het aanwijzend voornaamwoord dat je gehoord hebt (deze vs. dit of die vs. dat). 163 2 In dit voorbeeld is gekozen voor het. Het vakje met het wordt rood. Onderin wordt een balk geel. Geef op een schaal van 1 (onzeker) tot 5 (zeker) aan hoe zeker je bent van je antwoord. Hierna ga je naar het volgende scherm, waarin je weer hetzelfde moet doen. Het is belangrijk dat je zo snel mogelijk klikt nadat je het lidwoord hebt gehoord, ook al is de zin nog bezig. Als je de instructies begrepen hebt, kun je beginnen met de oefening. 164 Appendix 5 Datum: … / … / 2007 Questionnaire A. Algemene Informatie Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Geboren in (land) Wanneer ben je naar Nederland verhuisd? Maand: Jaar: Opleiding (Geef hoogste niveau aan) B. Taalvaardigheid Moedertaal (L1) Andere talen (L2s) 1. Leeftijd: … jaar (Hoogste niveau eerst) 2. Leeftijd: … jaar Geef aan op welke leeftijd je bent 3. Leeftijd: … jaar begonnen met het leren van de L2. 4. Leeftijd: … jaar 5. Leeftijd: … jaar Heb je cursussen Nederlands gevolgd? Ja / Nee Hoogste cursusniveau: Beginner / Gemiddeld / Gevorderd Geef jezelf een cijfer van 1 (laagst) tot 10 (hoogst) Hoe is je lezen in het Nederlands? Cijfer: Hoe is je spreken in het Nederlands? Cijfer: Hoe is je luisteren in het Nederlands? Cijfer: Hoe is je schrijven in het Nederlands? Cijfer: Zie volgende bladzijde. 165 C. Taalgebruik Welke taal gebruik je: Altijd Engels Meestal Engels, soms Nederlands Engels en Nederlands Meestal Nederlands soms Engels Altijd Nederlands N/A Thuis? Als je TV kijkt? Als je de krant leest? Als je boeken leest? Als je naar de radio luistert? Op je werk? Op school? Met je vrienden? In een winkel? Met iemand die je voor het eerst ontmoet? Zie volgende bladzijde. 166 D. Lidwoordentest Hieronder staan 60 Nederlandse zelfstandige naamwoorden. Geef aan welk lidwoord (de of het) bij het woord hoort. We willen graag dat je spontaan antwoord geeft, dus twijfel niet te lang. Als je echt niet weet of een woord de of het moet hebben, geef dit dan aan met ?. Als je het woord niet kent, geef dit dan aan met X. 1. …… stilte 21. …… activiteit 41. …… houding 2. …… hoofd 22. …… kruid 42. …… bedrijf 3. …… boom 23. …… haan 43. …… visioen 4. …… plaats 24. …… naam 44. …… ritme 5. …… laken 25. …… gebied 45. …… scène 6. …… vuur 26. …… beslissing 46. …… functie 7. …… antwoord 27. …… gesprek 47. …… schema 8. …… kennis 28. …… reden 48. …… ontwerp 9. …… vliegveld 29. …… contact 49. …… casino 10. …… informatie 30. …… beweging 50. …… kamer 11. …… gevoel 31. …… infectie 51. …… wijk 12. …… koninkrijk 32. …… pistool 52. …… transport 13. …… systeem 33. …… verjaardag 53. …… feit 14. …… resultaat 34. …… voorbeeld 54. …… inspecteur 15. …… jaloezie 35. …… suggestie 55. …… schip 16. …… lever 36. …… piano 56. …… oever 17. …… water 37. …… podium 57. …… avontuur 18. …… snor 38. …… dier 58. …… volgorde 19. …… mengsel 39. …… konijn 59. …… kliniek 20. …… manier 40. …… hertog 60. …… patiënt Zie volgende bladzijde. 167 E. Vul het woorddeel in Op de volgende bladzijden staan 3 korte Nederlandstalige tekstjes. In de teksten zijn gaten gevallen. Het zijn geen hele woorden die zijn weggelaten, maar delen van woorden. Het is de bedoeling dat je uit het zinsverband probeert af te leiden welk woorddeel op de puntjes zou kunnen staan. De eerste zin is steeds helemaal intact gelaten om je een beetje op weg te helpen. Je hebt maximaal 5 minuten de tijd per tekst. Zie volgende bladzijde. 168 Tekst 1 Ik houd van Nederland en niet zo’n beetje ook. Waarom ik van het land houd is niet alleen omdat velen van wie ik houd hier leven, nee, het is me___________________ dan groo___________________ reden d___________________. De v___________________ mijn lie___________________ voor het land ko___________________ voort u___________________ het feit dat al___________________ zo geor___________________ en syste___________________ is. Er i___________________ een systeem e___________________ het wer___________________. Je kan, ni___________________ zonder twi___________________, dic___________________ maar ogen to___________________ er___________________ met uitgaan d___________________ het recht zege___________________. Zie volgende bladzijde. 169 Tekst 2 Openlijke narcisten zijn mensen met een opgeblazen gevoel over zichzelf. Ze ei___________________ vaak ande___________________ aandacht o___________________ en ko___________________ charmant ov___________________, ond___________________ het feit d___________________ ze wei___________________ besef he___________________ van de beho___________________ van anderen. Verb___________________ narcisten zijn weli___________________ net z___________________ hevig met zichzelf arrogant be___________________ en a___________________ ma___________________ ze ev___________________ openlijke narcisten, do___________________ dit o___________________ een subti___________________ manier. Zie volgende bladzijde. 170 Tekst 3 Het internationaal perscentrum Nieuwspoort discussieert weer eens over de code. De Haagse soci___________________ waar h___________________ journaille en de poli___________________ in een onged___________________ samenzijn verk___________________, hanteert si___________________ jaar en d___________________ de ongesc___________________ d___________________ wat er gez___________________ bui___________________ wordt mag regel t___________________ plekke ni___________________ naar ko___________________. Alt___________________: niet her___________________ mag worden tot de betre___________________ persoon en pla___________________. Voorzitter van het bestuur van Nieuwspoort Max de Bok maa___________________ onlangs pla___________________ voor Casper Becx, maar het beleid bleef ongewijzigd. Bedankt voor je medewerking! 171 Appendix 6 Questionnaire Nederlandse Controls Datum: … / … / 2007 1. Datum: ... / ... / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 2. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 3. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 4. Datum: … / … / 2007 Naam Geslacht Man / Vrouw 172 Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 5. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 6. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 7. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 8. Datum: … / … / 2007 Naam 173 Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 9. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 10. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 11. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 12. Datum: … / … / 2007 174 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 13. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 14. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 15. Datum: … / … / 2007 Naam Geslacht Man / Vrouw Geboortedatum … / … / 19 … Moedertaal Opleiding (Geef hoogste niveau aan) 175