Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 436 28.8.2008 4:08pm chapter 23 .................................................................................................................................................... I E , RO M A N C E : S PA N I S H .................................................................................................................................................... laura malena kornfeld 23.1 Introduction ......................................................................................................................................................................................... The present chapter oVers a short review of the characteristics of compounding in Spanish, presenting not only the main patterns of compounds but also the empirical and theoretical problems they pose for the literature on Spanish morphology. The main types of Spanish compounds are presented here with observations on their internal structure and analysis of the formal relations between their constituents. I intend to lay out the problems that compounds would cause to a strict Lexicalist view of grammar (such as DiSciullo and Williams 1987), which has prevailed in the description of Spanish morphology. In my view, the problems in the analysis of many properties of compounds have their origin in the attempt to keep some premises of the Lexicalist approach ‘untouched’. I claim that these premises, being false, condition a wrong analysis of the phenomena they are meant to describe. The present chapter intends to change this tendency by proposing a new direction to re-examine the data from a critical point of view. Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 437 28.8.2008 4:08pm laura malena kornfeld 437 23.2 Types of Spanish compounds ......................................................................................................................................................................................... There are some topics concerning the description of compounding that can be repeatedly found in the generative literature on Spanish morphology. One of these is the traditional distinction between ‘proper’ and ‘improper’ compounds or, in other words, between lexical compounds (that result from a ‘process of compounding that operates on words’) and syntactic compounds (that imply a ‘process of reinterpretation that operates on syntactic constructions’), following Val Alvaro (1999: 4760). Val Alvaro assumes that prototypical lexical compounds have phonological amalgamation of their constituents and morphological unity. The Wrst characteristic implies one main stress and morphological juncture, by means of elision of phonological material or insertion of linking elements and phonological endings; thus, telaraña ‘net-spider’ (‘cobweb’) has phonological amalgamation (by means of elision of phonological material), while tela de araña ‘net of spider’, with the same meaning, does not. On the other hand, morphological unity implies marginal and unique inXection in the second constituent: compare the plural telarañas ‘net-spider(pl)’ with telas de araña ‘net(pl) of spider’, which does not have marginal inXection. However, a quick review of the diVerent types of compounds makes it possible to corroborate that the behaviour of compounds with the same pattern or internal structure is not uniform. Thus, almost all the patterns of compounding present some variation in the formal properties of compounds. Morphophonological (and consequently orthographical) variations can be found even in the same word, as in the oscillation between the forms with and without phonological amalgamation and morphological unity medialunas ‘half(sg)moons’ / medias lunas ‘half(pl) moons’ (‘croissants’), or casaquintas ‘house(sg)villas’/ casas quintas ‘houses(pl) villas’ (‘country houses’). For this reason, the majority of compounding patterns in Spanish have been considered by some authors as improper or syntactic compounds and by other authors as prototypical or lexical compounds (see, for instance, Varela 1990 vs. Rainer and Varela 1992 on P þ N compounds; Rainer and Varela 1992 vs. DiSciullo and Williams 1987 on N þ N and N þ de þ N compounds), perhaps with the sole exception of V þ N compounds, uniformly deWned as lexical in the literature. Another repeated topic in the literature is the distinction between endocentric and exocentric compounds, which refers to the possibility of recognizing one constituent as the head of the unit.1 Basically, some compounds permit us to identify a head constituent, while others do not. Since phrases are, in principle, always endocentric, exocentricity is considered as a proof of the non-syntactic 1 The notion of head employed here is the usual one in the literature on morphology, according to which the head of a morphologically complex word would be a constituent with the same category as the whole, and, on a semantic level, a kind of hypernym of the complex. Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof 438 page 438 28.8.2008 4:08pm ie, romance: spanish (i.e. lexical/morphological) nature of the compound. This distinction has also given rise to some puzzles, to which we will come back in the third section of the chapter. I will now review brieXy the main types of Spanish compounds. The presentation is organized observing the internal structure of the compounds, and not their output lexical category (since – as we will see – the same compound can function alternatively as an adjective or as a noun). I will only mention the productive patterns, which can be used in order to create new words, putting aside all obsolete types (as in sopicaldo ‘soup-i-stock’ ¼ ‘poor quality soup’ or altiplanicie ‘high-i-plain’ ¼ ‘high plain’) or the patterns that create a very reduced set of words (for instance, the pattern Adv þ V in malvender ‘badly-sell’ ‘sell oV cheap’ or malgastar ‘badly-spend’ ¼ ‘waste’), as well as compounds formed with Greek or Latin stems, which present other peculiarities and pose other theoretical problems. This review is restricted, then, to the patterns that permit a productive creation of new compounds constituted by Spanish free morphemes.2 23.2.1 V þ N V þ N is a compounding pattern common to several Romance languages, although relatively rare in the rest of the Indo-European languages. The pattern implies the relation between a transitive verb and a noun that, in syntax, would be a direct object complement:3 in this way, in espantapájaros ‘scares-birds’ [scarecrow], pájaros ‘birds’ is a plausible direct object for the verb espantar ‘to scare’. The same could be said about abrelatas ‘opens-cans’ (‘can opener’), abrepuertas ‘opens-doors’ (‘door opener’), ayudamemoria ‘helps-memory’ (‘memory aid’), chupamedias ‘sucks-stockings’ (‘boot-licker’), comeniños ‘eats-children’ (‘child eater’), cortacésped ‘cuts-lawn’ (‘lawnmower’), cubrecama ‘covers-bed’ (‘bedspread’), cuentakilómetros ‘counts-kilometres’ (‘mileometer’), rompecabezas ‘breaks-heads’ (‘puzzle’).4 The examples in (1) illustrate the parallelism between the internal structure of a compound V þ N and the one in an analogous VP: 2 The examples presented correspond to different Spanish varieties, particularly Argentine Spanish. We can find very few exceptions to this generalization, as the cases of girasol, ‘turns-sun’ (‘sunflower’) (verb þ complement of a PP) or pagadios, ‘pays-God’ (verb þ subject). Related cases are the coordinated verbs in correveidile, ‘run-go-i-tell-him/her’ (‘gossip’) or subibaja ‘goes-up-i-goesdown’ (‘seesaw’), that Rainer and Varela (1992: 127) consider to be a different pattern. 4 There is a debate in the traditional and generative literature about the nature of the verbal constituent in V þ N compounds. Some authors assume that the V is a stem (root þ thematic vowel), while others claim it is the third person singular of the present tense (i.e. the Spanish non-marked verbal form). As seen in the glosses, we adopt the second view. 3 Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 439 28.8.2008 4:08pm laura malena kornfeld (1) 439 a. En el campo hay un [X0 espantapájaros]. in the Weld there-is a scares-birds ‘There is a scarecrow in the Weld.’ b. Ese muñeco this puppet [SV espanta [pájaros]]. scares birds. In general, the V þ N pattern is very productive in the creation of instrumental nouns: abrelatas ‘opens-cans’ (‘can opener’), lavaplatos ‘washes-dishes’ (‘dish-washer’); and agentive nouns: matasanos, ‘kills-healthy (people)’ (‘physician’), cuidacoches ‘carescars’ (‘car watcher’). It has rightly been noted that, although they often function as nouns, they can also be adjectives (cf. Rainer and Varela 1992: 129; Val Alvaro 1999: 4793), as in una tarjeta abrepuertas ‘a card opens-doors’ (‘a key-card’) or un alumno chupamedias ‘a student sucks-stockings’ (‘a student who is a boot-licker’). A problem concerning the analysis of V þ N compounds is their exocentricity, that is, the fact that none of the constituents may apparently function as the head (see note 1). It seems evident that the head of espantapájaros is neither the verb espanta (since it does not have the category of the whole) nor the noun pájaros (which has the ‘right’ category, but lacks all semantic relation with the meaning of the compound). In any case, there exist several ‘endocentric’ analyses of V þ N compounds, that propose, for instance, the reanalysis of the (originally inXectional) Wnal -a/-e of the Wrst constituent as a derivational suYx (cf. Varela 1990) or the existence of a derivational zero-aYx that subcategorizes a VP and determines the nominal category of the compound (cf. Lieber 1992a for French V þ N compounds). 23.2.2 P þ N The nature of the P þ N pattern has been intensely discussed in literature on Spanish grammar, where we can Wnd very diVerent analyses. In traditional grammars, P þ N compounds such as sinvergüenza ‘without-shame’ (‘shameless’) or contraluz ‘against-light’ (‘back light’) are considered lexical or proper because of their frequent phonological amalgamation (e.g. RAE 1931), but some morphologists put them among syntactic compounds given their similarity with PPs (e.g. Rainer and Varela 1992). There are also analyses suggesting that a process of preWxation is involved, in which case these forms would be derived words and not compounds (e.g. Varela 1990 or Varela and Garcı́a 1999). By including them in the present chapter, we reject the latter approach, since the elements appearing in the Wrst position (sin ‘without’, contra ‘against’, entre ‘between’, sobre ‘over’) are Spanish prepositions that have the same meaning in compounds as they usually have in syntax. Therefore, the P þ N pattern can be simply described as the combination of free morphemes. P þ N compounds are never formed with the so-called ‘light’ prepositions (mainly de ‘of’, en ‘in’, and a ‘to’), which produce, instead, idioms that can be eventually lexicalized Lieber and Stekauer / The Oxford Handbook of Compounding 440 23-Lieber and Stekauer-chap23 Page Proof page 440 28.8.2008 4:09pm ie, romance: spanish later on (such as enseguida ‘in-consecutive’ ¼ ‘immediately’ or acerca ‘to-near’ ¼ ‘about’, cf. Masullo 1996).5 The Wrst constituents are, then, ‘full’ prepositions from a semantic point of view, such as contra (e.g. contraargumento ‘against-argument’ ¼ ‘counterargument’, contraluz ‘against-light’ ¼ ‘back light’, contraindicación ‘against-indication’ ¼ ‘contraindication’, contracultura ‘against- culture’ ¼ ‘counterculture’), entre (e.g. entretiempo ‘between-time’ ¼ ‘break’, entrepierna ‘between-leg’ ¼ ‘crotch’, entretelones ‘between-curtains’ ¼ ‘behind-the-scenes’), sin (e.g. sin techo ‘without-roof’ ¼ ‘homeless’, sin papeles ‘without-documents’ ¼ ‘undocumented’, sinvergüenza ‘without-shame’ ¼ ‘shameless’, sinrazón ‘without-reason’ ¼ ‘mistake’, sinsentido ‘without-sense’ ¼ ‘nonsense’), or sobre (e.g. sobrepeso ‘overweight’, sobreprecio ‘over price’ ¼ ‘surcharge’, sobretecho ‘over roof’ ¼ ‘outer tent’). However, notice that other ‘full’ prepositions, such as con ‘with’, desde ‘from’, and hacia ‘towards’, cannot appear in this kind of compound. Like V þ N compounds, P þ N compounds have a predictable syntactic form and are perfectly parallel to PPs, as (2) shows: (2) a. Carlos es un [X0 sinvergüenza]. Carlos is a without-shame ‘Carlos is a shameless (man)’ b. Carlos es un Carlos is a polı́tico [PP sin politician without [vergüenza]] shame P þ N compounds have also been considered exocentric units, like the V þ N ones, since they are often used as nouns (a fact that ‘preWxal’ analyses claim as a proof of their derivational nature). It is possible to try a slightly diVerent analysis: the combination of a preposition with its nominal complement may function in syntax as a PP (i.e. with an adjectival distribution) and in fact P þ N compounds have this function in sequences such as carrera contrarreloj ‘race against-clock’ (‘timed race’) or even polı́tico sin vergüenza/ sinvergüenza ‘politician without shame/ shameless’, which actually may be alternatively understood as a PP, as suggested in (2b), or as a compound. PPs and adjectives are liable to be recategorized as nouns, as Rainer and Varela (1992: 121) have remarked, and, when seen in this way, the nominal category of P þ N compounds is fully predictable. 23.2.3 N þ N N þ N compounds are not so common or productive in Romance as they are in Germanic languages, but all the same there are a number of neological forms, 5 We are not considering here the forms created from Greek or Latin prepositions such as anti, pro, pre, post, which are very productive in word formation, since these forms cannot be free morphemes in Spanish syntax (but see note 13). Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 441 28.8.2008 4:09pm laura malena kornfeld 441 which is a sign of their relative productivity. Some N þ N compounds exhibit phonological amalgamation and marginal inXection, as aguanieve ‘water-snow’ (‘sleet’), casaquinta ‘house-villa’ (‘country house’) or bocacalle ‘mouth-street’ (‘street intersection’), sometimes with a shortening of the Wrst element, as cantautor ‘sing-author’ (‘singer-songwriter’). In the majority of N þ N compounds, however, the nominal constituents are morphophonologically independent: arco iris ‘bow iris’ [rainbow], bar restaurante ‘bar restaurant’, barco tanque ‘ship tank’, buque escuela ‘ship school’ ( ¼ ‘training ship’), carta bomba ‘letter bomb’, casa cuartel ‘house barrack’, célula madre ‘cell mother’ (‘stem cell’), ciudad dormitorio ‘town dormitory’ (‘bedroom-community’), centro-izquierda ‘centre-left’ (political party). They are all classiWed as endocentric compounds, in which the Wrst noun is the head of the compound, although two subtypes are usually distinguished (cf. Rainer and Varela 1992: 125–6). In the so-called ‘coordinative’ N þ N compounds, both nouns seem to have the same hierarchy from a semantic point of view, which could be paraphrased by ‘an object that is both N1 and N2’ (for instance, casaquinta ‘house-villa’ ¼ ‘country house’, pollera pantalón ‘skirt trousers’ ¼ ‘trouser-skirt’). On the other hand, in ‘subordinative’ N þ N compounds, the second noun indicates a property of the class referred to by the Wrst noun, which could be paraphrased by ‘an N1 that is like a N2’, as in célula madre ‘cell mother’ (‘stem cell’) or perro policı́a ‘dog police’ (‘German shepherd dog’).6 With respect to the characterization as proper or improper, in Spanish grammars N þ N compounds are usually considered not as ‘syntactic’ as the N þ de þ N or N þ A ones. Rainer and Varela (1992) and Val Alvaro (1999), who assume the traditional characterization, recognize, however, that there is a ‘fuzzy border’ between N þ N compounds and appositions. Indeed, there exists a signiWcant parallelism between N þ N compounds and the so-called restrictive apposition, as illustrated in (3) (and in fact DiSciullo and Williams 1987: 81–2 consider the French equivalent to N þ N compounds as lexicalized phrases that keep their phrasal nature): (3) a. cantante autor singer author ‘singer-songwriter’ b. un clérigo poeta autor de miles de versos a priest poet author of thousands of verses deleznables futile (from Rainer & Varela 1992: 119) 6 In our view, some cases presented in the literature as examples of N þ N compounds, such as viaje relámpago ‘trip lightning’ (‘flying trip’) or periodista estrella ‘journalist star’ (‘star journalist’), correspond actually to the resemanticzation of the second noun; because of that, it would be more appropriate to define these nouns as ‘new adjectives’ that produce systematic and transparent paradigms (e.g. visita/ vuelo/ guerra/ ataque relámpago ‘lightning visit/ flight/ war/ attack’, redactor/ futbolista/ empleado/ alumno/ profesor estrella ‘star writer/ soccer-player/ employee/ student/ professor’), as Val Alvaro (1999: 4785) suggests. Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof 442 page 442 28.8.2008 4:09pm ie, romance: spanish 23.2.4 A þ A A þ A compounds are very similar to N þ N compounds, since in both cases the compounds show a relation of coordination/juxtaposition between elements of the same category. A þ A compounds present diVerent levels of formal ‘cohesion’ with respect to the processes of phonological amalgamation and marginal inXection indicated by Val Alvaro (1999). Thus, there are some compounds with an interWxal i (such as agridulce ‘sour-i-sweet’ ¼ ‘sweet and sour’, blanquiceleste ‘white-i-skyblue’) or shortening of the Wrst element (e.g. semio-lingüı́stico ‘semiolinguistic’, socioeconómico ‘socioeconomic’) – in both cases with obligatory marginal inXection – as well as compounds with less or null phonological amalgamation and with or without marginal inXection: árabe-israelı́ ‘Arab-Israeli’, sordomudo ‘deaf-mute’, (clase) media alta ‘(class) middle-high’ (‘upper middle (class)’). Morphophonological variations can be found even in the same word: semánticos-pragmáticos ‘semantic(pl)-pragmatic(pl)’ / semántico-pragmáticos ‘semantic(sg)-pragmatic (pl)’or verdes-azules ‘green(pl)-blue(pl)’ / verde-azules ‘green(sg)-blue(pl)’. 23.2.5 N þ de þ N N þ de þ N compounds are constituted by two nouns related by the Spanish ‘default’ preposition in the nominal domain, de,7 as in agente de seguridad ‘agent of security’ (‘security oYcer’), barco de vapor ‘ship of steam’ (‘steamboat’), bautismo de fuego ‘baptism of Wre’, bicicleta de montaña ‘bike of mountain’ (‘mountain bike’), bodas de plata ‘wedding of silver’ (‘silver wedding’), botas de lluvia ‘boots of rain’ (‘rubber boots’), caja de música ‘box of music’ (‘music box’), calidad de vida ‘quality of life’, canción de cuna ‘song of cot’ (‘lullaby’), cara de bebé ‘face of baby’ (‘baby face’), casa de campo ‘house of country’ (‘country house’), cerebro de mosquito ‘brain of mosquito’ (‘pea brain’), diente de leche ‘tooth of milk’ (‘milktooth’), ojo de buey ‘eye of bull’ (‘porthole’), patas de rana ‘legs of frog’ (‘Wns’), torre de marWl, ‘tower of ivory’ (‘ivory tower’). The formal parallelism between N þ de þ N compounds and phrases is evident and, in fact, these compounds can never produce morphophonological words (i.e. with phonological amalgamation and morphological unity). For these reasons, N þ de þ N compounds are placed within the set of syntactic or improper 7 The default nature of de is supported by the multiplicity of the semantic relations between the two nouns that this preposition can express (‘possession’, ‘substance’, ‘part–whole’, ‘origin’, ‘function’, ‘domain’, etc.), by its functioning as a mere grammatical marking in the relation between deverbal nouns and their arguments (e.g. la contaminación del agua ‘the pollution of the water’, la pesca de ballenas ‘the fishing of whales’, la decisión del presidente ‘the decision of the president’), and by the fact that it is the sole Spanish preposition that can appear as a sentinel in nominal ellipsis and other anaphoric phenomena (e.g. el hijo de Juan y el e de Pedro, ‘the son of Juan and the e of Pedro’, cf. examples (12) and Brucart 1987). Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 443 28.8.2008 4:09pm laura malena kornfeld 443 compounds in the literature on Spanish morphology (cf., for instance, Rainer and Varela 1992, Val Alvaro 1999). However, some morphologists have suggested that N þ de þ N compounds should be considered as ‘real’ compounds – with the same status as V þ N compounds, for instance – given that they imply a conceptual unit and that it is impossible to make syntactic operations within the unit (cf. DiSciullo and Williams 1987), as shown in examples (6)–(7) below. On the other hand, from the point of view of acquisition, N þ de þ N compounds have been compared with English N þ N compounds because of their great productivity in Romance languages (Snyder and Chen 1997). 23.2.6 N þ A / A þ N These compounds are usually grouped under the label of improper compounds, because of the obvious syntactic nature of the relation between nouns and adjectives (cf. Rainer and Varela 1992, Val Alvaro 1999). However, there is a restricted subset of forms with phonological amalgamation and marginal inXection, including forms with the linking vowel i (e.g. pelirrojo ‘hair-i-red’ ¼ ‘red-haired’, manicorto ‘hand-i-short’ ¼ ‘stingy’) and other words whose formal unity seems to be the result of accidental phonological or metrical factors: thus, two-syllables words are ‘better candidates’ than longer words for phonological amalgamation and marginal inXection (e.g. aguafuerte ‘water strong’ ¼ ‘etching’ vs. agua mineral ‘water mineral’ ¼ ‘mineral water’), but other diVerences seem to be attributable to purely ‘euphonic’ eVects, e.g. malapata ‘bad-leg’ ¼ ‘hard luck’ vs. mala suerte ‘bad luck’). Compounds of the type of pelirrojo or manicorto are considered adjectives, although they may often function as nouns (see Val Alvaro 1999). Among the adjectives that constitute these compounds, there are qualifying adjectives in prenominal position (e.g. malapata ‘bad-leg’ ¼ ‘hard luck’, altavoz ‘high-voice’ ¼ ‘loudspeaker’, malasangre ‘badblood’ ¼ ‘heartache’, librepensador ‘freethinker’) or postnominal position (marea roja ‘tide red’ ¼ ‘red tide’, Webre amarilla ‘fever yellow’ ¼ ‘yellow fever’, caja negra ‘box black’ ¼ ‘black box’, aguardiente ‘water-burning’ ¼ ‘liquor’, aguafuerte ‘water strong’ ¼ ‘etching’, bicho raro ‘bug strange’ ¼ ‘weirdo’), as well as relational adjectives (pastor alemán ‘shepherd German’ ¼ ‘German shepherd dog’, escalera mecánica ‘ladder mechanic’ ¼ ‘escalator’, portero eléctrico ‘porter electric’ ¼ ‘door intercom’). Because of the usual classiWcation in Spanish traditional grammar of quantiWers as prenominal adjectives, cases such as medianoche ‘middle night’ (‘midnight’), medialuna ‘half moon’ (‘croissant’), ciempiés ‘one hundred feet’ (‘centipede’), milhojas ‘one thousand leaves’ (‘puV pastry’) (in which the Wrst element is a quantiWer) are generally included in the set of A þ N compounds. With respect to the determination of the head, some N þ A compounds are considered exocentric, since the noun cannot be a semantic hypernym of the Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof 444 page 444 28.8.2008 4:09pm ie, romance: spanish whole, e.g. piel roja ‘skin red’ (‘redskin’), carapintada ‘face-painted’ (‘special forces soldier’), caradura ‘face-hard’ (‘cheeky devil’). Nonetheless, most N þ A compounds are endocentric: thus, Webre amarilla ‘fever yellow’ (yellow fever) is a kind of fever, escalera mecánica, ‘ladder mechanic’ (‘escalator’) a kind of ladder, etc., although the meaning of the compound is often metaphoric, as in malapata ‘bad-leg’ (‘hard luck’) or aguardiente ‘water-burning’ (‘liquor’). BrieXy, following the mainstream literature on morphology, Spanish compounds are mostly nominal, although A þ A compounds (such as agridulce ‘sweet and sour’) are always adjectives and some N þ A, V þ N, and P þ N compounds can also function as adjectives (e.g. muchacho pelirrojo, ‘boy hair-ired’ ¼ ‘red-haired boy’, tarjeta abrepuertas ‘card opens-doors’ ¼ ‘key-card’, carrera contrarreloj ‘race against-clock’ ¼ ‘timed race’, respectively). The relations established between the constituents of the compound are invariably syntactic and can be reduced to two cases: either it is the relation expected between a head and its complement (as in P þ N, A þ N, N þ N, N þ de þ N compounds) or a relation of coordination/ juxtaposition (as in A þ A and most N þ N compounds). 23.3 Discussion ......................................................................................................................................................................................... In the previous section, I have presented the diVerent types of Spanish compounds in the way they have usually been described in the relevant literature. In the present section I will explain some aspects that might challenge the division of labour among the diVerent grammatical components if a strict separation of morphology and syntax (i.e. a real Lexicalist view of grammar) is assumed, as occurs, in fact, in the most part of the descriptive literature of Spanish (see Varela 1990, Rainer and Varela 1992, Varela and Garcı́a 1999, Val Alvaro 1999, Varela 2005, among others). The Wrst aspect has to do with the fact that the relation between the constituents of compounds can always be explained by syntactic properties of Spanish. That is to say, the combination is always predictable from a syntactico-semantic point of view. Thus, as has already been remarked, V þ N compounds consist of a transitive verb and a noun that might be its direct object in syntax; neither intransitive verbs nor nouns that could not constitute a plausible direct object from a semantic point of view can be found as part of these forms. As (4) shows, a compound combining a verb with a non-plausible direct object is equally odd in the equivalent sentence: (4) a. Ese es un this is a ??espantamesas. ??scares-tables b. Ese es un this is a muñeco puppet que espanta ??mesas. that scares ??tables Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 445 28.8.2008 4:09pm laura malena kornfeld 445 Moreover, the more subtle formal properties of compounds are a direct reXex of sentential syntax: thus, the noun is usually plural if it is countable (cf. (5a)) and singular if it is uncountable (cf. (5b)), just as we would expect within a VP:8 (5) a. espantapájaros / cuentakilómetros / cascanueces scares-birds / counts-kilometres / cracks-nuts ‘scarecrow / mileometer / nutcracker / levantavidrios / raises-glasses / glass-raiser’ b. guardapolvo shelters-dust ‘overalls / tragaluz / swallows-light / skylight / abrepuertas / opens-doors / door-opener / quitaesmalte / ayudamemoria / removes-polish / helps-memory / polish remover / memory aid’ Something similar can be said about the rest of Spanish compounds, which also have predictable properties from the syntactico-semantic point of view, as has already been remarked. The perfect parallelism between the properties of compounds and phrases in Romance languages has led scholars to question whether it makes any sense to duplicate the rules of syntax in morphology, since what determines the relation between espanta and pájaros in espantapájaros is, in short, the syntactic component (DiSciullo and Williams 1987: 79–83). For these authors, assuming that morphology contains rules related to the combination of a transitive verb with a direct object would result in a general lack of economy for grammar. But then, if compounds result from the simple combination of words following syntactic rules, what is ‘special’ about compounds with respect to phrases? Or, in other words: why are they are regarded as the product of morphology and not of syntax? The justiWcation for regarding compounds as words (and not phrases) is based on evidence showing that they are syntactic atoms in the sense of DiSciullo and Williams (1987), which additionally implies that they exhibit a number of formal properties. Thus, their constituents cannot be separated by determiners or quantiWers (6) nor be modiWed in an independent way (7): (6) a. Encontró en la orilla una estrella found in the shore a star mar [in the relevant interpretation] sea ‘She found a starWsh on the shore.’ de (*l) of (*the) 8 In fact, the presence of inflectional marks in countable nouns is one of the most solid proofs that constituents of compounds should be regarded as words rather than stems (despite the ambiguity in the case of the verbal constituent). Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof 446 page 446 28.8.2008 4:09pm ie, romance: spanish b. Es un abre (*muchas) latas. is a opens (*many) cans ‘It is a can-opener.’ (7) a. Es un abre [latas (*grandes)]. is a opens cans (*big) ‘It is a can-opener.’ b. La habitación tiene un the room has a ‘The room has a porthole.’ ojo (*grande) eye (*big) c. una mujer (??excelente) a woman (??excellent) ‘a police woman’ policı́a police de buey. of bull d. Tiene un perro policı́ai (*, que proi es una profesión terrible). profession terrible) has a dog policei (*, that proi is a ‘He has a policedog.’ As an additional proof of the syntactic opacity of compounds, notice that, if some kind of lexical substitution is made, the referential value of the compound is lost (cf. chupamedias ‘sucks-stockings’ / ??chupazoquetes ‘sucks-socks’, espantapájaros ‘scares-birds’ / ??espantaves ‘scares-Xying animals’). From data similar to (6)–(7), DiSciullo and Williams (1987) propose the notion of syntactic word: some compounds in Romance languages must be produced by the rules of syntax and then relabelled as words by means of a peripheral morphological rule which enables them to occupy a head position (X0). In other words, syntactic words are created by means of a (morphological) reanalysis rule applied to phrases (XP!X0). In this sense, Romance compounds should be distinguished from English compounds, which, following DiSciullo and Williams, are produced by speciWc morphological rules of stem combination. DiSciullo and Williams assume, then, that Romance languages lack morphological compounding and have only lexicalized phrases (i.e. syntactic objects listed in the lexicon) and syntactic words, created by the XP!X0 rule. Setting aside the theoretical problems that this view poses for a Lexicalist framework, I can provide empirical arguments that indicate that the notion of syntactic word is, at least, debatable. The Wrst empirical argument is that compounds can contain phrases (in a way similar to English, Dutch, and Afrikaans compounds, as described by Lieber 1992a). This possibility makes it unlikely that they could really occupy an X0 position, which is a fundamental characteristic of syntactic words. For instance, V þ N compounds – i.e. the prototypical proper compounds in Spanish (and whose status as compounds has never been challenged in the literature), – can be formed by two coordinated verbs (see examples in (8a)) Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 447 28.8.2008 4:09pm laura malena kornfeld 447 or by a verb and a direct object expanded by means of coordination (8b) or of a nominal complement (8c): (8) a. un [[ lava y limpia] faros] / [[ lava y seca] ropas] a washes and cleans headlights / washes and dries clothes ‘an [artefact that] washes and cleans headlights / washes and dries clothes’ b. un [cubre [piletas y carpas]] / [ cubre a covers swimming-pools and tents / covers [piletas y carritos]] swimming-pools and trolleys ‘a [fabric that] covers swimming-pools and tents / swimming-pools and trolleys’ c. proyectiles de tipo [‘perfora [chalecos’ antibalas]] projectiles of kind perforates- vest anti-bullets 9 ‘projectiles that perforate bullet-proof vest’ These data may be contrasted with the examples usually discussed in the literature insisting on the opacity of compounds (see examples in (6)–(7) and also Varela 2005: 75, who presents data incompatible with (8a/b)). And, crucially, there is no coherent way to account for data such as (8) from a Lexicalist perspective. Any possible explanation in terms of either coordination, ellipsis (cf. Varela 2005: 63–4), or complementation would mean that syntactic operations can apply within syntactic atoms, in direct contradiction of the strict Lexicalist principles assuming the total opacity of these units. Parallel data to (8) exist for P þ N and N þ de þ N compounds: (9) a. un sin techo ni hogar / un sin-techo a without roof nor home / a without-roof (ni paredes, ni ventanas ni puertas) (nor walls nor windows nor doors) ‘a homeless homeless’ / ‘a homeless wall-less, window-less, door-less’ b. dientes permanentes y teeth permanent and ‘permanent and milk teeth’ de leche10 of milk Interestingly, the cases in (8) and (9) are much more complex from a theoretical point of view than the examples analysed by DiSciullo and Williams (1987), which only show the existence of syntactic relations between heads and complements 9 All the controversial data in this chapter (cf. specially examples in (8) and (9)) may be checked in Google. 10 Obviously, coordination is impossible in the case of metaphorical units, such as ojo de buey ‘eye of bull’ (‘porthole’) or luna de miel ‘moon of honey’ (‘honeymoon’). Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof 448 page 448 28.8.2008 4:09pm ie, romance: spanish within compounds (see (6)–(7)). Moreover, examples in (8) and (9) are in radical contrast with the syntactic opacity that is characteristic of all inXectional and derivative Spanish suYxes (with the exception of -mente) and a substantial part of Spanish preWxes: (10) a. *[am *[lov y odi]ó, and hat]ed, b. *in [ útil *un [ useful y and *[nominaliza *[nominaliz y verbaliza]ción and verbaliz]ation capaz], *re [hacer capable], *re [make y escribir] and write] It is obvious that the probability of phrases appearing within a compound as in examples (8)–(9) is very low, but, evidently, it is above zero. Therefore, these data allow us to question the assumption of an essential distinction between words and phrases (or, in other words, that there exists a clear-cut separation between morphology and syntax). Nonetheless, the prohibition on functional elements such as determiners or quantiWers appearing within compounds, illustrated in (6), is strictly respected (in other words, determiners and quantiWers do seem to be limited to sentential syntax). On the other side, we have already remarked that a property that the literature on Spanish compounds usually takes as a sign of the lexical or morphological nature of certain compounds is exocentricity (i.e. apparent absence of a head), while syntactic constructions are always endocentric. This is the case with V þ N or P þ N compounds, in which – as has been observed – it is not easy to identify a constituent that functions as the ‘head’ of the construction (see footnote 1). However, the existence of exocentric compounds in Spanish seems to be dependent on a much more general property: as Kornfeld and Saab (2003, 2005) have noted, apparent ‘exocentricity’ can be found also in ‘normal’ syntactic constructions with nominal anaphora,11 which do not involve word formation at all: (11) a. Habı́a dos vestidos: el e más elegante era there-were two dresses: the e most elegant was azul (e ¼ ‘vestido’) blue (e ¼ ‘dress’) ‘There were two dresses: the most elegant one was blue.’ b. ¿Te gusta el e más elegante o el e más you like the e most elegant or the e most raro? (e ¼ ‘humano’) odd (e ¼ ‘human’) ‘Do you like the most elegant one or the oddest one?’ 11 Anaphora is here used in the traditional sense of Hankamer and Sag (1976), involving merely an empty (or pronominal) category that is syntactically or pragmatically controlled. Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 449 28.8.2008 4:09pm laura malena kornfeld 449 (12) a. Hay dos vestidos: el e de arriba me gusta there-are two dresses: the e of above me like más (e ¼ ‘vestido’) most (e ¼ ‘dress’) ‘There are two dresses: I like the one above better.’ b. El e de arriba me tiene bronca (e ¼ ‘humano’) the e of upstairs me has hate (e ¼ ‘human’) ‘My upstairs neighbour hates me.’ (13) a. Habı́a dos vestidos: compré el e que te there-were two dresses: bought-I the e that you gustó (e ¼ ‘vestido’) liked (e ¼ ‘dress’) ‘There were two dresses: I bought the one you liked.’ b. Saludé al e que te gusta (e ¼ ‘humano’) said-hello-I the e that you like (e ¼ ‘human’) ‘I said hello to the one you like.’ The nature of this kind of construction has been extensively discussed in the literature. Kornfeld and Saab (2005) observe that in examples such as (11)–(13) there are actually two diVerent phenomena. In one case there is nominal ellipsis, in which the semantic interpretation of the omitted noun is identical to an antecedent (vestido, in the (a) examples); in the other case, the empty noun is interpreted, by default, as ‘human’ (see the (b) examples).12 The relevant literature has often emphasized one or the other case in order to propose unitary explanations for the examples (cf. for instance Brucart 1987 or Bosque 1990), although neither of the phenomena can be clearly explained in terms of the other. It sounds reasonable to believe that the existence of anaphoric phenomena such as these also inXuences diVerent processes of word formation, for instance conversion from A to N, as in el tonto ‘the fool [one]’ or elimpermeable ‘the waterproof ’ [raincoat], a process that is much more frequent in Spanish than in languages that lack nominal ellipsis, as English. In the speciWc case of compounds, anaphoras are involved in the apparent ‘exocentricity’ of diVerent patterns. Thus, the ‘normal’ category for a P þ N compound is the one corresponding to a PP (with an equivalent distribution to an adjective in the nominal domain). Its use in contexts of omission of the noun as in (14) is what leads to its ‘recategorization’:13 12 There is as well a third possibility: the factive interpretation, as in El que hayas venido me conmovió ‘The [fact] that you have come moved me’. 13 This analysis can also be applied to the cases of word formation involving Greek/ Latin ‘prepositions’, such as anti in antimisil ‘against missile’ (‘antimissile’) or inter in interclubes ‘among clubs’ (‘interclub’). The established literature assumes that anti or inter should be prefixes and involve derivational processes, since they cannot function as free morphemes in Spanish syntax (see, for instance, Varela and Garcı́a 1999 and Varela 2005). The problem with this idea is that the resulting Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof 450 (14) page 450 28.8.2008 4:09pm ie, romance: spanish a. un polı́tico sin vergüenza a politician without shame ! un sinvergüenza ! a shameless [person] b. un juicio sin sentido a statement without sense ! un sinsentido ! a nonsense ! un ! a ! ! un e a e e sinvergüenza e without shame sinsentido without sense A similar explanation can be proposed for the so called ‘exocentric’ N þ A compounds, such as piel roja ‘skin red’ (‘redskin’) or carapintada ‘face-painted’ (‘soldier’): they are originally adjectives, their exocentricity being the superWcial consequence of the presence of a nominal anaphora with the interpretation ‘human’, as can be observed in the paraphrases in (15): (15) a. un hombre piel roja ! un e piel roja ! un piel roja a man skin red ! a e skin red ! a red skin b. un militar cara pintada ! un e carapintada ! un a soldier face painted ! a e face-painted ! a carapintada face-painted [soldier] The case of V þ N compounds seems less obvious; nonetheless, in these compounds also, a predicative construction is often reanalysed as a noun by virtue of the anaphoric processes already described. If anaphoras are involved in the category of V þ N compounds, it is easy to explain why V þ N compounds can be alternatively nouns or adjectives, as examples in (16) show, and also why the gender of V þ N compounds sometimes depends on an elided noun (16a):14 (16) a. una máquina a machine tragamonedas swallows-coins (‘una (‘a máquina machine que that traga swallows words do not seem to have the expected category for the combination of a prefix with a noun: they often function as As (as in escudo antimisil ‘shield antimissile’ or crema antiarrugas, ‘cream antiwrinkle’) and not as Ns. In order to solve this problem, it is usually assumed that antimisil or antiarrugas are actually nouns in apposition. However, it seems much more plausible to assume that the complex functions in a similar way to sinvergüenza ‘without-shame’ (‘shameless’) (that is, as PP/adjective), given that the normal syntactic function for antiarrugas ‘anti-wrinkle’ is modifying a noun. Therefore, the prepositional character of anti, inter, pro, etc. is kept and they are more like compounds than derivational words. 14 For a complete analysis, see Kornfeld (2005), where V þ N compounds are considered as complex heads created in syntax that can include some phrasal projections, as the ones illustrated in (8), but, crucially, never involve certain functional categories such as I or D. In this analysis, V þ N compounds function originally as adjectives, but can be recategorized as nouns by the anaphoric processes already mentioned. It is worth noticing that analyses of V þ N compounds that include null categories have been previously proposed in generative literature, for instance in Contreras (1985) and DiSciullo (1992b) for Italian V þ N compounds. Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof page 451 28.8.2008 4:10pm laura malena kornfeld monedas’) coins’) ! ! 451 una tragamonedas a(fem) slot machine b. un/a chico/a cuidacoches (‘un/a chico/a a boy/girl cares-cars (‘a boy/girl coches’) ! un/a cuidacoches cars’) ! a(masc/fem) car-watcher que cuida that cares To sum up, it can be remarked that, on the one hand, the case of compounds with an expanded or coordinated constituent shows that establishing a clear-cut boundary between compounds and phrases is almost impossible. This explains why intermediate categories such as syntactic words, syntactic compounds, lexicalized phrases, etc. are so common in the literature, as well as the debates and controversies about the lexical/ morphological or syntactic nature of particular patterns of compounds. On the other hand, the parallelism between the exocentricity of certain compounds and the phenomena of nominal ellipsis and other anaphoric elements suggest that the study of the general syntactic properties of Spanish should say more about compounds than the proposal of speciWc ad hoc properties. And, additionally, this would make it possible to explain several properties of Spanish (nominal ellipsis, categorial conversion, ‘exocentric’ compounds, Greek/ Latin compounding, etc.) with only a few rules, which is an epistemological desideratum. 23.4 Conclusions ......................................................................................................................................................................................... The brief review in this chapter suggests that in the case of Spanish compounds it is not possible to separate morphology from syntax. From an empirical point of view, one of the most important conclusions is that it is diYcult to obtain interesting generalizations by focusing on the patterns of internal structure of Spanish compounds, since compounds with the same internal structure behave in syntax in radically diVerent ways. On the other hand, although morphophonology supplies more transparent and less debatable criteria, it leads also to a great degree of arbitrariness, since the same pattern alternatively can or cannot produce unitary words, depending on random or non-grammatical factors such as the number of syllables or metrical reasons. Another relevant result is that compounds present more syntactic transparency than would be expected. On the one hand, their constituents may be phrases, as we have seen with the case of V þ N compounds, the most prototypical examples of Spanish compounds. This implies that the constituents of compounds (and not Lieber and Stekauer / The Oxford Handbook of Compounding 23-Lieber and Stekauer-chap23 Page Proof 452 page 452 28.8.2008 4:10pm ie, romance: spanish compounds themselves) are syntactic atoms (i.e. minimal elements manipulated by syntax). On the other hand, a property considered as typical of certain compounds (exocentricity, as opposed to the obligatory endocentricity of phrases) has proved to be a consequence of a normal syntactic phenomenon in Spanish (nominal anaphoras). Both empirical arguments explain why there is so much controversy in the literature about the syntactic or the lexical nature of this or that compound, and also call into question the importance of these debates. From the theoretical point of view, the data strongly indicate that the simplistic Lexicalist view that morphology produces words and syntax produces phrases (predominant in Spanish literature) cannot be maintained. In spite of their greater adequacy to the data, reanalysis rules of the kind proposed by DiSciullo and Williams (1987) should also be abandoned, at least if these rules are assumed to apply before syntax (see Lieber 1992a for a similar criticism of the strict Lexicalist views based on data from diVerent languages). Additionally, this permits us to question wordhood as a prime or real principle of grammar, something that the theoretical frameworks proposing a complete ‘deconstruction’ of grammar (such as Distributed Morphology, cf. Halle and Marantz 1993) have repeatedly suggested. Therefore, given that compounds can include certain syntactic structures, it seems more plausible to conceive of them as the product of a ‘reduced syntax’ (or microsyntax, in Benveniste’s 1966 proposal), in which certain properties of ‘real’ syntax are available while other properties are not.