Word and Learner Factors in Guessing Unknown Words from Context in a Second Language Meral Ozturk Uludag University, Turkey Ebru Ertarman Senyigit Uludag University, Turkey Abstract This study investigates five factors that might be effective in guessing word meanings from context in a second language. Two of these are related to the word to be guessed: word class and morphological complexity. Guessing of 40 English words from the four major word classes (nouns, verbs, adjectives, and adverbs) were tested. Half of these words had a morphologically complex structure and the other half were simple. The other variables were related to the guesser: L2 proficiency, L2 vocabulary size, and cognitive style. The subjects were university-level EFL learners. The data were gathered by means of a guessing test, a vocabulary checklist, a cloze test, and the Group-Embedded Figures Test. The results suggested that both word factors were effective in guessing, but the study provided no evidence for the effect of learner variables except having a combined effect in the case of complex words. Guessing from context while reading is claimed, by some researchers (Nagy, 1997, p. 64; Nation, 1990, p. 160, Nation, 2001, p.232), to be a useful strategy for learning word meanings in a second language. One advantage is that guessing can be used by the learner independently of the classroom, the teacher and of any other material used for teaching. To learn a new word’s meaning the learner only needs the context in which the word has appeared. The practical implication of this is that the learner will be able to learn words 2 which are not covered in the limited classroom time or classroom material and thus will be able to build a large vocabulary on her own and one consistent with her needs. Moreover, the context often contains information about other aspects of words besides their form and meaning such as collocations, syntactic behaviour, formality, register and even frequency. With repeated exposure to a word in rich contexts, it should be possible for a learner to pick these up along with form and meaning. Studies on the use of vocabulary learning strategies have shown that guessing is very popular among L2 learners. In Fraser (1999), intermediate ESL learners preferred the guessing strategy more often than consulting a dictionary or ignoring the word altogether when they encountered unfamiliar words in reading. Harley and Hart (2000) investigated, in a questionnaire, the vocabulary learning strategies of thirty-four learners of French as a second language in Canada. Guessing from context was reported as being one of the most frequently used strategies of word learning. In a large-scale survey of the word learning strategies used by Japanese EFL learners, Schmitt (1997) has found guessing to be one of the most frequent strategies as well as being perceived as most helpful. In spite of this favourable attitude of both researchers and learners towards guessing, research on the incidental learning of foreign language words through reading did not strengthen the case for guessing as a useful word learning strategy. Incidental vocabulary learning occurs during a language activity in which the focus is not on learning words such as when reading a 3 book, watching TV, listening to radio, or having a conversation, and it “primarily occurs through the process of inferring [i.e., guessing] word meaning” (Fraser, 1999, p. 226). These studies are mostly replications (Pitts, White, and Krashen, 1989; Day, Omura, and Hiramatsu, 1991; Hulstijn, 1992; Dupuy and Krashen, 1993; Horst, Cobb, and Meara, 1998) of the study by Saragi, Nation, and Meister (1978), which investigated the incidental learning of foreign words by adult English native speakers through reading a novel in their native language. The results of these studies, though, were not as dramatic as those in Saragi, Nation, and Meister (1978). While 75 % of the items in Saragi, Nation, and Meister (1978) were learnt as revealed by a multiple-choice test following the reading treatment, the highest amount of words retained in its replications, which also used a multiple-choice format to measure vocabulary gain with the exception of Hulstjin (1992), was 22 % (Horst, Cobb, and Meara, 1998; Dupuy and Krashen, 1993). These results highlight a difference between L1 and L2 vocabulary learning. While most vocabulary in L1 is learnt incidentally from context (Nagy, 1997, p.75; Nation, 2001, p.232), interacting with a context does not seem to facilitate vocabulary learning in a second language to the same degree. Of course, learning words through reading is a process involving several steps, and guessing may not be, by any means, the only step in the process where learners fail (see Laufer, 2003): a learner might fail to notice an unknown word (i.e., noticing); she might notice the word but decide not to make an attempt to guess it (i.e., decision); she might make an unsuccessful guess due to a lack of sufficient contextual clues or to shortcomings in her own 4 abilities (i.e., guessing); or she might make a successful guess but may not remember it later (i.e., retention). Even if the learner successfully retains words from guessing, the gain may not be reflected in her overall vocabulary size if the words learnt are small in number. The present study will investigate how successful learners are at guessing word meanings from clues available in context as well as the word itself. A number of factors will contribute to successful guessing in context. Mondria and Wit-de Boer (1991) identify three groups of factors that could affect successful guessing. These are contextual factors, word factors and learner factors. Contextual factors are aspects of the context in which the new word appears, and among these are the “pregnancy” of context (i.e., amount of contextual clues available) as well as the frequency with which the new word is used in the context. Word factors are the characteristics of the word to be guessed such as morphological complexity, morphological transparency, parts of speech, or degree of abstractness. Finally, learner / reader factors are those that relate to the learner’s abilities which include L2 vocabulary size, general L2 proficiency, or ability to guess. While contextual factors (Mondria and Wit-de Boer, 1991; Horst, Cobb, and Meara, 1998) and learner factors (Hulstijn, 1993) have received some attention in L2 vocabulary acquisition research on guessing, word factors have been largely ignored. The present study will investigate the effect of two word factors on guessing: word class and morphological complexity. In addition, three learner variables will be studied: vocabulary size, general L2 proficiency 5 and field independence. This study will be the first to investigate field independence and L2 proficiency in relation to guessing. Word Class The grammatical category of the word to be guessed is a potential factor that could affect the difficulty with which it is guessed. Nation (1990, p.162; 2001, p.257) considers the identification of the parts of speech of a new word in a text as the very first step in successful guessing. Words belonging to different parts of speech are likely to present learners with different levels of difficulty because different word types require different types of contextual clues to be guessed. Research has shown that SL learners are sensitive to word class differences in their use of context clues. Paribakht and Wesche (1999) identified various knowledge sources used by ESL learners to deal with unknown vocabulary in reading. They found that learners appealed to what they call “sentence-level grammatical knowledge” (i.e., knowledge of the grammatical position of the word in the sentence and its relation to other words in the same sentence, which depends heavily on parts of speech knowledge) more often to guess the meanings of adjectives. Clues provided by word morphology was used more with verbs, which often carried information about the tense and aspect of the sentence through the inflectional suffixes attached to them. Punctuation was more important to the guessing of nouns: capitalization signalled proper nouns and commas separated nouns in a series. 6 As there is no research to our knowledge that investigates the effect of word class on guessing L2 words from context, evidence will be brought forward from the L2 acquisition studies on word class to predict differences between grammatical classes in guessing. Research into the acquisition of word class knowledge in a second language suggests that learners’ knowledge of word class information about the words they know is far from complete. In Odlin and Natalicio (1982), intermediate and advanced ESL students were able to identify the word class of high frequency words only 75% of the time. Word class knowledge was lacking in one-fourth of the words which should be well-known to these learners at their current level of proficiency. Schmitt (1998, 1999) has found that his advanced level EFL subjects knew two word class forms on average out of four in a word family. In only a few cases did they know all four word class forms (noun, verb, adjective, and adverb). In contrast to the findings of these studies, Bensoussan (1992) reports only five parts of speech errors (1%) in a pool of 361 mistranslations to Hebrew of English words contained in a reading passage by advanced EFL learners. Strategy research also indicates that using word-class information to identify word meaning is not very popular among L2 learners. Kanatlar (1995) used think-aloud and retrospective protocols to investigate the strategies used by beginning and upper-intermediate Turkish EFL learners to identify the meanings of nonsense words in a reading text. She reports only 18 instances of the use of part of speech information in guessing words from context out of 7 a total of 561 cases. With this rate, part of speech was the fifth most frequent strategy among the ten strategies identified as being used by these learners. Schmitt (1997) investigated the use of a rather extensive list of vocabulary learning strategies (40 in total) in a large-scale survey of Japanese EFL learners. Results have shown that analysis of the part of speech of the target word was not a particularly popular discovery strategy. Only 32% of the learners reported using this strategy to identify the meaning of a new word although a majority of the learners (75%) thought the strategy to be helpful. As the learners progressed through the school system, however, the use of the strategy increased from 20% to 43% and perceptions of helpfulness increased from 64% to 87%. Other studies have concentrated on the relative difference in difficulty between nouns and verbs. Ellis and Beaton (1996) have found that concrete nouns were easier to learn in word lists than verbs. In an error analysis with ESL learners, Lennon (1996) has found verbs to be the most frequent source of error. In Kallkvist (1998), inappropriate use of verbs in a free composition task in English was twice as high as for nouns. While there is a clear difference between nouns and verbs in favour of nouns, these two categories are often believed to be easier than either of the adjective and adverb categories. In a study cited earlier, Schmitt (1998) has shown that in a word family learners were more likely to know the nouns and the verbs than the adjectives or the adverbs. Laufer (1990, 1997b), on the other hand, reviews the limited SL research on word class and argues that the differences observed among words of different grammatical categories in these studies might have resulted 8 from phonological or morphological factors and not from their part of speech as such. She concludes that there is no research evidence for the effect of word class on the ease or difficulty of learning L2 words. While the research cited above on word class differences relate to learning rather than guessing words in an L2, Nation (1990, p. 48) suggests that these findings “partly agree with experience in guessing from context” and that “nouns and verbs are easier to guess than adjectives and adverbs”. In the present study, we seek to substantiate this claim. Morphological Analysis Clues to the meaning of an unknown word may also be found in the word itself. In the case of morphologically complex words (i.e., words consisting of a base and one or more derivational affixes), the meaning can be worked out by analysing the meanings of parts that the word consists of. The meaning of the word homeless, for example, can be worked out by analysing it into its separate morphemes as home and -less. Given that the meaning of the base (i.e., home meaning “a place to live”) and that of the suffix (i.e., –less meaning “without”) are known it should not be so very difficult for a learner who was unfamiliar with the word before to figure out that it refers to someone with nowhere to live. Although this strategy can be very effective in guessing word meaning if used in combination with contextual guessing, research reports indicate limited use of this strategy by second language learners. In Harley and Hart (2000), 9 referred to earlier, the strategy of morphological analysis was one of the least popular of the discovery strategies used to learn new L2 words. In Kanatlar (1995), word analysis was the sixth most frequent strategy out of ten strategies used for guessing words from context. In Schmitt (1997), analysis of roots and affixes was used by only 15% of the learners surveyed even though the learners found the strategy generally helpful for discovery purposes. The perceived usefulness of the strategy increased from 52% to 79% as the learners progressed to higher grades although there was not any significant increase in the frequency of use. A number of explanations might be offered for the reluctance of learners to use this strategy to identify meanings of new words. First, the learners, especially at an advanced level of proficiency, might have become aware of the non-compositionality of meanings of many morphologically complex words through previous unsuccessful analyses of word structure. Many words in English are “deceptively transparent” (Laufer, 1989; 1997a) and their meanings are not the sum of the meanings of their parts. Laufer (1997a, p. 146) provides examples of incorrect guesses by EFL learners based on morphological analysis: the interpretation of dis- in the word discourse as “incorrect” and of course as “way” led to the incorrect guess “wrong way”; the analysis of falsities into fall and cities led to the incorrect guess “falling cities”; and a morphological analysis on outline produced an “out-of- line” interpretation of the word. Laufer (1997a, p. 153) urges teachers to “warn the learners not to rely on word morphology too much and not to draw conclusions about sentence meaning on the basis of individual words, as some of them 10 may be “pseudofamiliar”, that is, they appear to be familiar though they are not”. In a similar vein, Clark and Nation (1980, pp. 173-174) and Nation (1990, p. 168) suggest this strategy be used as part of a checking procedure after a guess has been made on the basis of contextual clues. Laufer (1997a, p. 153) also cautions that the meaning worked out from morphological analysis should be checked against the context in which the word is encountered. Therefore, it is only sensible for learners not to risk the outcome by applying a strategy which is not guaranteed to be successful. An alternative explanation might be that learners might have deficiencies in their knowledge of derivational morphology of the L2, which prevents them from analysing morphological structure of words. There is some research evidence for this inadequacy of second language learners regarding derivational morphology. Schmitt and Meara (1997) investigated the suffix knowledge of intermediate Japanese learners and found that they were able to produce only 15% of possible derivative forms of base words. Their recognition knowledge of targeted derivational suffixes was better, but they failed to recognise more than 50% of the legal derivations as being allowable in English. Morin (2003) trained L2 Spanish learners on derivational morphological analysis, but found no effect of training on learners’ vocabulary knowledge of the L2, although the study “indicated a positive trend in the effectiveness of morphological knowledge as a tool for building vocabulary knowledge” (p. 215). 11 It is also possible to explain the limited use of morphological analysis in terms of avoidance. Osburne and Mulling (2001) found that knowledge of morphology was not lacking in the advanced ESL learners who participated in their study as shown by a forced-choice task where the learners had to decide between a morphologically correct and a morphologically incorrect non-word. In about half of the cases, the learners chose the morphologically correct option. This was offered as an explanation of the results of an earlier study by the same authors, which had shown that, when other word recognition strategies were available like using cognates or context, learners avoided using a morphology strategy. The authors conclude that learners avoid using morphological cues not because they do not know them but because they find morphological analysis difficult, as evidenced by such learner comments from their study as the following: It’s too hard or You have to think a lot. Blum and Levenston (1979) were first to document this avoidance phenomenon with morphologically complex words. In a sentence completion task, more than 75% of the learners of Hebrew failed to use the required complex forms of verbs. They opted for non-complex synonyms or other acceptable answers. The present study will investigate if learners are able to make use of their knowledge of derivational affixes in inferring the meaning of unfamiliar words encountered in a sentence context rich in contextual clues. Paribakht and Wesche (1999) reported that learners often used multiple knowledge sources when inferring the meaning of new words when reading in the L2. We ask here if learners are able to compound contextual clues with morphological 12 clues, or whether the presence of context clues renders appeal to morphological clues unnecessary. Learner Characteristics Vocabulary size has also been claimed to be important for guessing (Laufer, 1997a; Nation, 1990). Learners with larger vocabularies should be able to guess better as they will know more of the words in the context of the unknown word and thus will be able to utilize more of the contextual clues available than those with smaller vocabularies. Laufer (1997a) suggests that 95% of the words in the context of the unknown word should be known for successful guessing to occur. Hirsh and Nation (1992) suggest that the learner should know 98% of the words in a text to be able to guess the meaning of an unknown word successfully. The research evidence for the relation between guessing and vocabulary size is sparse. While Hulstijn (1993) found a significant correlation between vocabulary scores as measured by the Levels Test (Nation, 1983) and inferring ability scores, in Horst, Cobb, and Meara (1998) the correlation between vocabulary size and the number of words learned incidentally from reading (i.e., by guessing) was weak. The present study will attempt to substantiate this relation. General proficiency in the second language can also be important for successful guessing. More proficient learners will be able to understand the 13 surrounding context better as they will know more words of the text; they will understand more of the grammatical structures used in the text; they will be able to better follow the discourse and will be more experienced readers and word guessers. Therefore, they should be able to use the context better to make successful guesses at the word meaning. There is, however, a lack of research into this relation between proficiency and guessing. Finally, degree of field-independence of the learner is suggested, in the present study, as a factor that might be effective in guessing. Fieldindependent individuals are better able to identify figures embedded in more complex figures and are not distracted by the irrelevant details surrounding the figure in comparison to field-dependent individuals. Field independence is not restricted to the domain of visual perception, but is believed to be a basic cognitive style, which also determines how an individual approaches a learning task. When applied to guessing, one would expect field independent learners to guess more accurately as they will not be misled by the irrelevant information contained in the context and will be better able to pinpoint the appropriate contextual clues. More specifically, the following research questions will be sought answers to: 1. Are there any differences in guessing between nouns, verbs, adjectives and adverbs? 2. Do learners use their knowledge of affixes to guess the meanings of morphologically complex words that are previously unfamiliar to them? 14 3. Do learners with larger vocabularies guess more successfully than learners with smaller vocabularies? 4. Do more proficient learners guess more successfully than less proficient learners? 5. Do more field-independent learners guess more successfully than less field-independent learners? Method The participants were 64 EFL learners (18 males and 46 females) between the ages of 18-21 in four intact groups in a teacher-training college in Turkey. Two of the groups were chosen from among the first year classes and two from the fourth-year classes so that a pool of subjects from the two ends of the EFL proficiency continuum at the institution, which roughly corresponded to intermediate and advanced levels of English proficiency, could be selected. Subjects were assigned later to a proficiency group on the basis of scores on a cloze test (see below). All participants spoke Turkish as their L1. The participants answered four tests measuring the various variables of the study: a guessing test, a cloze test, a vocabulary size checklist and the group-embedded-figures test. The guessing test (see Appendix A) measured the subjects’ success in guessing 40 target words in English. The target words were those in Oxford Advanced Learners’ Dictionary (1989) that were judged to be unknown to the 15 subjects by the two researchers who have taught these learners over two semesters. Half of the targets were morphologically simple and half were complex. In each category, there were five words that belonged to each of the four major word classes (i.e., nouns, verbs, adjectives, and adverbs) (see Appendix B for a list of the test words). The selection of the complex words was guided by the prefixes in these words. All pages with a given prefix were scanned for appropriate words. The simple words were selected randomly in accordance with any criteria set for a given word class category. The simple words in the test had a simple morphological structure and contained no affixes apart from the adverbs which all inevitably ended in –ly. All complex words contained prefixes and some had suffixes that indicated their word class category like –able in adjectives (e.g., irrefutable) or –er in nouns (e.g., misnomer). The complexity of these words was thought to have brought about by the prefixes rather than the suffixes as prefixes change the meaning of the base in more significant ways than suffixes. The prefixes in the target words were common prefixes in English (i.e., un-, dis-, re-, pre-, im-, in-, ir-, mis-), which are usually introduced early to language learners through frequent vocabulary like misunderstand, disadvantage, irregular, etc. The bases for all the prefixes in complex words were unfamiliar to the learners as in irrefutable, misconstrue, or dismantle. The reason for this was to force the learners to use their knowledge of prefixes in arriving at a guess. They would be expected to do better on the complex words than the simple words if they used clues to meaning offered by these prefixes compounded with contextual clues available in both cases. 16 All verbs in the test were transitive (e.g., enhance, exert, etc.). Transitive verbs were believed to offer a richer (i.e., more pregnant) context in that additional contextual clues could be provided through the direct object (e.g., enhance one’s chance of getting a job). The adjectives in the test were all attributive. While a predicative adjective in a sentence may be quite apart from the noun it modifies, an attributive adjective is adjacent to the noun, which may lead to a more direct and unambiguous interpretation of the clues provided by the noun. Other factors that might have an effect on the difficulty of a word like word length, phoneme-grapheme correspondence, frequency, or the size of the word family a word belongs to have not been controlled for in this study because these seem to be more important for long-term retention of words than for guessing their meaning. However, given the lack of research on the effect of these factors on guessing, they should ideally be controlled for in future research. The target words were tested in sentences rich in contextual clues. The following sentence for the complex verb uncoiled, for instance, contains a number of clues to the meaning. The gardener patiently uncoiled all 100 meters of the long pipe to water the plants at the end of the garden. 17 In this sentence, the phrases the gardener, the long pipe, and to water the plants lead to the inference that uncoiling is something done to pipes by gardeners to water plants and also that it will affect a long stretch of the pipe if the plants are far from the water source and is likely to take a long time and therefore, it is something that has to be done patiently. The contextual richness of the sentences was checked for with 19 native speakers of English who were EFL teachers in various schools in the city. Initially, 64 sentences each with a different target word (i.e., 8 words for each category) were pilot-tested for the guessability of the target words on the basis of the contextual clues provided in the sentence. The native speaker informants were given the sentences with the target words missing and they were asked to supply the missing word. The sentences for which more than 50% of the informants provided the original word or a synonym (e.g., continuously for incessantly) were counted as sufficiently rich in contextual clues. The top 5 items in each category which matched this criterion (i.e., 40 items altogether) were used in the final version of the test. For the main data collection, 20 distractor sentences which tested grammar words from minor word classes (e.g., with, unless, much, although, etc.) were added and randomised with the target items. Two versions of the test were developed by swapping the first half of the test with the second. The target words were written in bold and underlined. The subjects were asked to give an L1 equivalent for the target words. Two points were granted for an exact equivalent, 1 point for a synonym of the exact equivalent (e.g., 18 anlaşmazlık meaning ‘disagreement’ for feud) and no points for an unrelated word (e.g., hastalık meaning ‘sickness’ for feud). Subjects’ vocabulary sizes were measured by means of a checklist (see Appendix C) following Meara and Buxton (1987). The words in the checklist were selected randomly from Longman Dictionary of Contemporary English (1995), which contains 56000 headwords. The sixth headword on every twenty pages of the dictionary was picked up, which yielded 60 words with a pick-up rate of 1 in 933. Certain types of words were avoided like proper nouns, abbreviated forms, words that belong to specific registers (i.e., religion, medicine, etc.), words belonging to other languages, slang words, grammar words or past and past participle forms of irregular verbs. These real words were mixed with 20 non-words some of which were chosen from introductory linguistics coursebooks and others were invented by one of the present researchers. The words were randomised and two different orders of the test were developed. The words were presented in eight blocks of ten words in order to relieve the monotony and inattention that might result from a single unbroken list. The subjects were asked to check the words they knew. In the calculation of vocabulary size, the number of words in the checklist known to a subject was taken to be the number of real words checked by the learner, which was then adjusted according to the number of non-words checked. Adjustment was necessary, as some of the real words might have been checked as known when they were not. The number of nonwords was one third of the number of the real words in the checklist. Therefore, the number of non-words checked by a learner was multiplied by 19 three and the result was subtracted from the number of real words checked. For example, a subject who checked 50 words 48 of which were real words and 2 non-words might have checked, with this rate, 6 of the real words inaccurately as known. The number of words known to the subject is the number of real words checked minus the corrected number of non-words checked (48-6=42). The result was multiplied by 933 (i.e., the pick-up rate) to estimate the number of words known in the dictionary (42x933=39186 words out of 56000). These scores were then used to divide the subjects into two groups as those with a larger vocabulary size and those with a smaller vocabulary size. Subjects who scored above the mean (i.e., 25 words on the test corresponding roughly to 23000 words) were considered to have a larger vocabulary while those who scored below the mean had a smaller vocabulary. The somewhat high mean vocabulary size might be attributable to the fact that the scores represent lexical items rather than word families in that LDOCE chooses to present words belonging to different parts of speech of the same word family as different headwords with the result that the number of words in the dictionary as well as the number of words known to the learners are increased. Large vocabulary sizes are also expected of this group of students as they are already at a pretty advanced stage of proficiency with substantial vocabulary in English. The purpose of the test was merely to draw a rough distinction between those who acquired a larger vocabulary and those with a relatively smaller vocabulary in the group, and the test seemed adequate towards that end. 20 Field-independence was measured by the Turkish version of Witkin’s (1976) The Group Embedded Figures Test (Okman-Fisek and Maktav-Yildirim, 1983). The subjects were given scores relative to the difficulty of items (i.e., 2, 4, and 8 points respectively for each item in the three sections). The highest score a subject could get was 122, which represented extreme field independence. The mean score of the group (i.e., 85) was taken as the cut-off point to assign subjects into one of two groups. Those who scored 85 or more were classified as more field-independent and those who scored below 85 were categorized as less field-independent. Proficiency was measured by means of a cloze test (see Appendix D). The standard cloze is often used as a measure of overall ability in language testing and suggested, by Hughes (1989, p.62-3), to be an economical way of obtaining an approximate idea of someone’s ability, which was also our main motive in using the cloze. In making the test, the blanks were created using the fixed-ratio deletion technique. Every seventh word of the text was deleted with the exception of the first and last sentences, which yielded 50 blanks. Subjects were instructed to fill in the blanks with one word only and guess if unsure. The answers were scored using the exact word technique where the only correct answer was the original word itself. Each blank was worth one point. The maximum score that could be obtained was 50. The mean score of the group with first and fourth year students combined was 17 and subjects who scored above this were in the higher proficiency group and those who scored below were in the lower 21 proficiency group. The fact that the mean score turned out to be somewhat low for this group of learners is probably the result of the scoring technique used. With this technique, synonyms of the original word which would be perfectly acceptable in the given context had to be rejected. Oller (1979, p. 367) provides evidence from research that compare the exact word technique and the synonym technique, and argues that while the synonym technique yields higher scores than the exact-word technique the distribution of scores remains the same. Since the purpose of using the cloze-test in the present study was to divide the subjects into two groups that differed in proficiency relative to one another rather than to determine their exact level of proficiency, the distribution of scores was considered more important than the actual values they represented. The exact-word technique seemed both adequate and practical in this respect. The subjects answered the tests in their usual class hours over two weeks. In the first week, they answered the cloze test and the checklist in one sitting and on a different day they did the group embedded figures test. In the following week, they answered the guessing test. All instructions were given in the L1 both in written and spoken form to avoid any misunderstanding. Data Analysis The data were analysed by a 5-way ANOVA on Statistica 4.5 for Windows (see Appendix E for the ANOVA results). The two within-subjects variables were morphological complexity and word class. The betweensubjects variables were proficiency, vocabulary size and field independence. The main effects of both within-subjects variables were statistically significant. 22 There were no significant main effects, however, for the between-subjects variables. Only two of the interactions were significant: the two-way interaction between morphological complexity and word class and the four-way interaction between proficiency, vocabulary size, field independence and morphological complexity. Results and Discussion The participants in the present study guessed about half of the test words correctly. On the average, they scored 43.4 points over the possible 80 points (54%) in the guessing test (see Table 1 below). Overall, they were not as successful in guessing as would be expected of them given the richness of clues in the surrounding context and the tightness of focus on guessing words. This result could be offered as an explanation for the discouraging findings of earlier studies on learning words from reading L2 texts (essentially through word guessing), which reported up to 22 % gains for non-native speakers. Gains in these studies often represent the retention rate of correctly guessed words. If learners guess only about half of the words correctly, as this study indicates, there will not be many words available for correct retention. It would follow from this that low learning rate from reading is largely the result of incorrect guesses. However, the discrepancy between the rate of correct guessing in this study and the highest learning rate in incidental vocabulary learning research (54% vs. 22% ) suggest that other factors are also at work. One of these could be that when guessing learners attend to the meaning whereas learning requires attention to both meaning and form. Because form 23 is arbitrary it is likely to be forgotten very quickly and require repeated encounters to be acquired in association with the corresponding meaning. Table 1 The lack of success in almost half of the items also raises concerns as to the adequacy of single-sentence contexts in providing the contextual richness necessary for successful guessing. It is all too possible that a single sentence, however rich in contextual clues, does not match the detail provided by a whole text. The present study revealed interesting results with respect to morphological complexity. On the whole, simple word scores were higher than complex word scores (see Table 1 for the means and standard deviations of the main variables of the study). The difference was found to be statistically significant (F.05(1, 56)=123.31, p<.000). Thus, the learners in the study guessed the meanings of simple words more correctly than those of complex words although one would expect them to do better with the complex words if they used clues provided by the target word. Two explanations can be offered for this result: it is either that the learners did not use morphological clues at all and relied on contextual clues only or that they used morphological clues but not very effectively. The former explanation is less likely to be true, in which case no significant difference would have been observed between simple and complex word scores as the learners would have treated complex words as single units like simple words and rely solely on context for guessing their meaning. 24 The lower success scores with complex words, on the other hand, suggests that learners used morphological clues, but were not able to work out the meaning quite well. The difficulty of the complex words probably resulted from the unfamiliarity of the bases for the prefixes rather than from avoidance of using prefix knowledge. It seems that learners were aware of the clues offered by the prefixes, but they were confused rather than helped by them when the base was unfamiliar. In a similar study, Manga (2003) compared the guessing of simple words with that of complex words using single-sentence contexts. The results revealed significant differences in favour of complex words. The complex words in her study contained familiar affixes and bases and thus, the learners were able to perform morphological analysis on these words. On the other hand, the unknown bases in the present study probably rendered them semantically opaque for the subjects. For successful guessing to take place, it is, therefore, important that the word to be guessed should be subjectively transparent (i.e., that all parts are known to learners) as well as being descriptively transparent (i.e., the meaning of the word could be inferred from the meanings of the parts). Significant differences were also found among word class categories (F.05(3,168)=10.75, p<.000). Verbs and adjectives were guessed more successfully than nouns and adverbs. However, the difference between verbs and adjectives, on the one hand, and nouns and adverbs, on the other, was not significant at .05 level according to Tukey HSD Test (see Figure 1). Figure 1 25 Although previous research into the learning of word class categories in a SL indicated nouns to be the easiest category, nouns in the present study turned out to be as difficult to guess as adverbs and more so than verbs and adjectives. Two explanations might be offered for the lower success rate of the learners with nouns. First, the nature of the nouns used in the study might be partly responsible for this result. In our attempt to find test words that would be unknown to our subjects of advanced EFL proficiency, we opted for lowfrequency words in the case of nouns as high-frequency nouns tend to be generally known at this proficiency level. As low-frequency nouns are often abstract in meaning, the noun category consisted mainly of abstract nouns. Abstract nouns are generally considered to be more difficult to learn than other types of nouns. Referring to the categorization of nouns proposed by Lyons (1977), Hatch and Brown (1995, p. 220) attribute the difficulty of abstract nouns to the fact that they are less typical. Lyons (1977, Vol. 2, pp. 422-466) identifies three categories of nouns. First-order nouns are concrete objects that exist in time and space, second-order nouns are states, processes and events and as such they take place in time and finally third-order nouns denote abstract entities which have no relation to time and space. Hatch and Brown (1995, p. 220) claim that first-order nouns, being more corelike (i.e., typical), should be easier to learn than second-order nouns and second-order nouns easier than third-order nouns. Being the least corelike of the noun category, abstract nouns must be the most difficult. 26 Indeed, research shows that learners have more difficulty acquiring abstract nouns than concrete nouns. In Kallkvist (1998), referred to earlier, 91% of the inappropriate use of nouns involved abstract nouns. If this is the case, the statement that nouns are the easiest category does not apply to nouns universally. Nouns might be easier to learn and to guess than other categories when they are concrete, but may not be so when they are abstract. However, hard research evidence as to the effect of abstractness in word learning is lacking. Laufer (1997b), in her review of the factors that affect word learning, concludes that concreteness does not have a clear effect. Alternatively, nouns may be easy to teach and learn, but may not be easy to guess. While the other categories have relational meanings (i.e., adjectives are related to nouns, adverbs to verbs, and verbs to nouns in the sentence), nouns have largely independent meanings. Aitchison (1987, p. 102) argues that nouns “are relatively free of syntactic restrictions. Verbs, on the other hand, are intrinsically entangled with the syntax of the sentence”. While nouns may provide good contextual clues to the meaning of words from other categories, they may not, in turn, receive the same contextual support. This is a matter for future research. The non-significant difference between verbs and adjectives suggested that adjectives are not more difficult to guess than verbs. The fact that the adjectives in the study were all attributive suggested close proximity of contextual clues: the noun modified by the adjective provided immediate clues 27 to the meaning of the adjective. Thus, contextual clues could be interpreted directly and unambiguously, which might have led to increased accuracy in guessing. The same explanation might be offered for Paribakht and Wesche’s (1999) finding regarding greater frequency of use of sentence-level grammatical knowledge as a knowledge source for guessing adjectives. It is likely that learners experience little difficulty if at all in relating the adjective to the noun it modifies when it immediately precedes it, which helps to materialize the meaning of the adjective. While, for example, the meaning of the word ardent is more elusive and abstract in isolation, it becomes rather more specific when used to modify the supporters of a football team as in the following sentence from the guessing test: He is one of the most ardent supporters of Galatasaray I’ve ever seen. The results suggested that adverbs are the most difficult class of words to guess, which was hardly unexpected. In the data with native English speakers, the blanks for adverbs often elicited prepositional phrases with quite different meanings from the target adverbs but which were also appropriate in the given context. Also, several non-synonymous adverbs were appropriate in a given context. This versatility might be a factor in causing adverbs to be difficult to guess. The interaction between morphological complexity and word class was revealed to be significant by ANOVA (F.05(3, 168)=5.000, p<.002). An examination of the means in Figure 2 has shown that the differences among word class categories were more marked in the case of simple words. Tukey 28 HSD Test on simple word means indicated that simple verbs and adjectives were more successfully guessed than simple nouns and adverbs whereas there was no statistically significant difference between verbs and adjectives and between nouns and adverbs. These differences observed among word classes in simple words seem to mirror the overall differences among word class categories, which suggested that overall differences were in fact a reflection of the differences in simple words only, as the group mean scores in complex words were stable across word class categories. Figure 2 In the case of simple words, the only type of clue that was available to aid guessing was contextual and the fact that different word categories were guessed with different success rates suggests that the degree of contextual support for some categories must be different from the others. While verbs and adjectives seem to be well-supported by the context, nouns and adverbs seem to be at a disadvantage in that respect. In complex words, on the other hand, there were no differences among word categories, which suggests that the contextual clues did not work differently for different types of words. This might be taken to mean that the context did not play an important role in the guessing of complex words. Still, the learners guessed about 9 of the 20 complex words (17.5 points out of 40, cf. Table 1) successfully. If this cannot be explained by the use of specific contextual clues in the sentence, it will have to be explained with the use of 29 prefix knowledge. However, prefix knowledge alone may not be considered sufficient to explain this result given the unfamiliarity of the bases to the learners. We would suggest here that learners made intelligent guesses with the complex words by using their prefix knowledge in combination with a global understanding of the sentence meaning and knowledge of the world concerning the situations described in the sentences. None of the main effects involving the subject variables of the study reached statistical significance: proficiency (F.05(1, 56)=0.25, p<.617), vocabulary size (F.05(1, 56)=2.54, p<.116) and field independence (F.05(1, 56)=.032, p<.856). Subjects with a higher level of L2 proficiency did not guess significantly more correctly than those with lower L2 proficiency. Subjects with larger vocabularies did not guess better than those with smaller vocabularies, and more field independent learners were not more successful than less field independent learners. Although subject variables were not effective individually, together they had a significant effect on guessing in combination with morphological complexity (F.05(1, 56)=6.564, p<.013). The combined effect of subject variables was examined by Tukey-HSD post-hoc comparisons separately for simple and complex words. There was only one statistically significant difference, which related to the guessing of complex words (see Table 2). Higher proficiency subjects with a larger vocabulary size and a more fieldindependent learning style guessed complex words more successfully (mean=21pts) than those who were highly proficient and more field- 30 independent but had a smaller vocabulary size (mean=13.2pts). This suggests that a larger vocabulary facilitates the guessing of complex words if the subject has a more field-independent style and is more proficient in the target language. Table 2 One might offer the following explanation for the facilitative effect of vocabulary size in combination with proficiency and field-independence on the guessing of complex words. Higher proficiency in the L2 and a more fieldindependent style will lead to a better processing of the surrounding context. When this is compounded by more experience with complex words that might be afforded by a larger L2 vocabulary guessing will be facilitated. In other words, these individuals will be able to use contextual and morphological clues in combination when guessing complex words and thus, be more successful at it. The insignificant effect of the subject variables, on the one hand, and the significant effect of these in combination on the guessing of complex words on the other suggest that the subject variables had an effect on guessing; but, for some reason, these were too weak to reach statistical significance individually. Only in combination did they make a strong enough effect to be statistically significant. The fact that only two of the 72 possible differences among the nine three-way combinations of the subject variables on complex 31 word scores were significantly different further suggests that the effect was weak. Manga (2003) explains similar findings by the strict control of vocabulary and structure in the context of the target word. As in her study, all the words in the context of a target word were known to the subjects in the present study, and the structure of the sentence was equally comprehensible. Therefore, more proficient learners and / or those with greater vocabulary sizes did not have an advantage over the less proficient learners or those with smaller vocabularies, as the latter did not experience any more difficulty with the surrounding context than the former. In further studies, if the context around the target word is varied between difficult and easy structure and / or vocabulary, more significant effects of proficiency and vocabulary size variables could be obtained. Conclusion In general, this study suggested that the nature of the word to be guessed affects the correct guessing of its meaning from context in significant ways. Different word class categories seem to present learners with different levels of difficulty. There is some indication, however, that the difficulty of guessing words in a major word class may not be uniform within the class. 32 Sub-classes of major categories might pose different challenges to the learner. For instance, concrete nouns may not be as difficult to guess as abstract nouns, or attributive adjectives as predicative adjectives. It seems necessary, therefore, to be more precise when using the major word class category labels. When one says noun, for example, does it refer to all types of nouns or only to the most typical type, i.e., concrete nouns? Word class labels are often used in the latter sense, but findings are generalised to the category as a whole, which may or may not be generalisable. The morphological make-up of a word also seemed to have an effect on guessing. Morphologically complex words seem to be difficult to guess when a part of the word is unknown to the learner. The subjective transparency of the complex words seems to be as important to correct guessing as the descriptive transparency. The three learner factors investigated, namely second language proficiency, vocabulary size and field-independence, were only effective in combination and then only in complex words. With larger groups and with greater differences among the groups with respect to a given construct, more significant differences might be obtained. REFERENCES Aitchison, J. (1987). Words in the Mind: An Introduction to the Mental Lexicon. Oxford: Blackwell. 33 Bensoussan, M. (1992). Learners’ spontaneous translations in an L2 reading comprehension task: Vocabulary knowledge and use of schemata. In Vocabulary and Applied Linguistics, P.J.L. Arnaud & H. Bejoint (eds.), 102-112. London: MacMillan. Blum, S. & Levenston, E. (1979). Lexical simplification in second-language acquisition. Studies in Second Language Acquisition 2 (2): 43-63. Clark, D.F. & Nation, I.S.P. (1980). Guessing the meanings of words from context: Strategy and techniques. System 8 (3): 211-220. Day, R.R., Omura, C., & Hiramatsu, M. (1991). Incidental EFL vocabulary learning and reading. Reading in a Foreign Language 7 (2): 541-551. Dupuy, B. & Krashen, S. (1993). Incidental vocabulary acquisition in French as a foreign language. Applied Language Learning 4 (1&2): 55-63. Ellis, N.C. & Beaton, A. (1995). Psycholinguistic determinants of foreign language vocabulary learning. In Lexical Issues in Language Learning, B. Harley (ed.), 107-165. Ann Arbour: John Benjamins. Fraser, C.A. (1999). Lexical processing strategy use and vocabulary learning through reading. Studies in Second Language Acquisition 21: 225-241. 34 Hatch, E. & Brown, C. (1995). Vocabulary, Semantics, and Language Education. Cambridge: Cambridge University Press. Harley, B. & Hart, D. (2000). Vocabulary learning in the content-oriented second language classroom: student perceptions and proficiency. Language Awareness 9 (2): 78-96. Hirsh, D. & Nation, P. (1992). What vocabulary size is needed to read unsimplified texts for pleasure. Reading in a Foreign Language 8 (2): 689-696. Horst, M., Cobb, T. & Meara, P. (1998). Beyond a Clockwork Orange: acquiring second language vocabulary through reading. Reading in a Foreign Language 11 (2): 207-223. Hughes, A. (1989). Testing for Language Teachers. Cambridge: Cambridge University Press. Hulstijn, J.H. (1992). Retention of inferred and given word meanings: Experiments in incidental vocabulary learning. In Vocabulary and Applied Linguistics, D.J.L. Arnaud & H. Bejoint (eds.), 113-125. London: MacMillan. 35 - (1993). When do foreign language readers look up the meaning of unfamiliar words? The influence of task and learner variables. The Modern Language Journal 77 (2): 139-147. Kallkvist, M. (1998). Lexical infelicity in English: the case of nouns and verbs. In Perspectives on Lexical Acquisition in a Second Language, K. Haastrup & A. Viberg (eds.), 149-174. Sweden: Lund University Press. Kanatlar, M. (1995). Guessing Words-in-Context Strategies Used by Beginning and Upper-Intermediate Learners. Unpublished MA dissertation, Bilkent University, Ankara. Laufer, B. (1989). A factor of difficulty in vocabulary learning: Deceptive transparency. AILA Review 6: 10-20. - (1990). Why are some words more difficult than others?-Some intralexical factors that affect the learning of words. IRAL 28 (4): 293-307. - (1997a). The lexical plight in second language reading. In Second Language Vocabulary Acquisition: A Rationale for Pedagogy, J. Coady & T. Huckin (eds.), 20-34. New York: Cambridge University Press. - (1997b). What’s in a word that makes it hard or easy: Some intralexical factors that affect the learning of words. In Vocabulary: Description, 36 Acquisition and Pedagogy, N.Schmitt & M.McCarthy (eds.), 140-155. Cambridge: Cambridge University Press. - (2003). Vocabulary acquisition in a second language: Do learners really acquire most vocabulary by reading? Some empirical evidence. The Canadian Modern Language Review 59 (4): 567-587. Lennon, P. (1996). Getting ‘easy’ verbs wrong at the advanced level. IRAL 34: 23-36. Longman Dictionary of Contemporary English (1995) Lyons, J. (1977). Semantics. Vol.2. Cambridge: Cambridge University Press. Manga, B. (2003). Word, Learner and Contextual Factors in Guessing Word Meaning from Context in a Second Language. Unpublished MA dissertation, Uludağ University, Bursa. Meara, P. & Buxton, B. (1987). An alternative to multiple choice vocabulary tests. Language Testing 4 (2): 142-154. Mondria, J.A. & Witt-de Boer, M. (1991). The effects of contextual richness on the guessability and the retention of words in a foreign language. Applied Linguistics 12 (3): 249-267. 37 Morin, R. (2003). Derivational morphological analysis as a strategy for vocabulary acquisition in Spanish. The Modern Language Journal 87 (2): 200-221. Nation, I.S.P. (1983). Testing and Teaching Vocabulary. Guidelines5, 1: 12-25. - I.S.P. (1990). Teaching and Learning Vocabulary. Boston: Heinle & Heinle. -I.S.P. (2001). Learning Vocabulary in Another Language. Cambridge: Cambridge University Press. Odlin, T., & Natalicio, D. (1982). Some characteristics of word classification in a second language. Modern Language Journal 66: 34-38. Okman-Fisek, G. & Maktav-Yildirim, S. (1983). Cocuk Gelisimi. Istanbul: Milli Egitim Bakanligi. Oller, J.W. (1979). Language Tests at School. London: Longman. Osburne, A.G. & Mulling, S.S. (2001). Use of morphological analysis by Spanish L1 ESOL learners. IRAL 39: 153-159. Oxford Advanced Learner’s Dictionary (1989) Paribakht, T.S. & Wesche, M. (1997). Vocabulary enhancement activities and reading for meaning in second language vocabulary acquisition. In 38 Second Language Vocabulary Acquisition: A Rationale for Pedagogy, J. Coady & T. Huckin (eds.), 174-200. New York: Cambridge University Press. - (1997). Reading and incidental L2 vocabulary acquisition. Studies in Second Language Acquisition 21: 195-224. Pitts, M., White, H. & Krashen, S. (1989). Acquiring second language vocabulary through reading: a replication of the Clockwork Orange study using second language acquirers. Reading in a Foreign Language 5 (2): 271-275. Saragi, T., Nation, I.S.P., & Meister, G.F. (1978). Vocabulary learning and reading. System 6: 72-78. Schmitt, N. (1997). Vocabulary learning strategies. In Vocabulary: Description, Acquisition, and Pedagogy, N. Schmitt & M. Mc-Carthy (eds.), 199-227. Cambridge: Cambridge University Press. - (1998). Tracking the incremental acquisition of second language vocabulary: A longitudinal study. Language Learning 48 (2): 281-317. - (1999). The relationship between TOEFL vocabulary items and meaning, association, collocation and word class knowledge. Language Testing 16 (2): 189-216. Schmitt, N. & Meara, P. (1997). Researching vocabulary through a word knowledge framework: Word associations and verbal suffixes. Studies in Second Language Acquisition 19 (1): 17-36. 39 Witkin, H.A. (1976). Cognitive Styles in Personal and Cultural Adaptation. Worchester: Clark University Press.