Can we make heads or tails of Spanish endocentric compounds? (1).
This paper proposes a unified account of headship assignment in Spanish endocentric compounds with a nominal non-head. It is argued that these compounds follow the universal head-complement word order (Kayne 1994), with the non-head element occupying the rightmost position. This structure surfaces whenever the non-head nominals in the compound are full words, that is, when they appear with their word-class marker (Harris 1991). Surface exceptions result from the insertion of stems instead of words in the non-head position. Nominal stems head a defective phrase that fails to project beyond the NP-level and cannot check features against the missing phrasal projection (ClassP). Therefore, they are forced to move to the left of the head by left-adjunction to circumvent their lack of case. This account, supported with quantitative evidence, is shown to be preferable to previous ones on observational and theoretical grounds. Descriptively, it has no counterexamples in the productive native patterns of Spanish compounding with nominal non-heads. Theoretically, it abides by the principles of procrastination and greed (Chomsky 1995), instead of relying on ad hoc stipulations.
The fact that Spanish compounds can be left- or right-headed and therefore challenge Williams' righthand head rule (1981) is well attested in descriptive and theoretical accounts (Bustos Gisbert 1986; Nunez Cedeno 1991, 1992; Clements 1992; Rainer and Varela 1992; Rainer 1993, among others). How to predict their headedness has proved a recalcitrant problem in Spanish morphology (cf. Nunez Cedeno 1991, 1992; Clements 1992; Piera 1995; Benavides Tarrius 1996, for a number of different proposals). Some of those solutions account for all the compounds of Spanish in a unitary fashion. I contend that this attempt is problematic, given that it does not take into consideration that different internal relations may hold between compound constituents, with possible repercussions on head position. To avoid this problem, this paper focuses exclusively on endocentric compounds, that is, those where the categorial features of one of the constituents are inherited by the compound word, whereas the other constituent is in some way subordinate to the first. The former I will call "head" while the term "non-head" will be reserved for the latter. (2) In other words, exocentric compounds, where neither constituent acts as a head, are excluded (e.g. sacacorchos 'corkscrew', literally 'remove-corks'), as are dual compounds, where both constituents can be said to "share" headedness (e.g. actor-bailarin 'actor-dancer'). Moreover, of various possible types of endocentric compounds, only those with nominal non-heads will be analyzed here, while those with adjectival or adverbial non-heads will be excluded from consideration, for example bajorrelieve 'bas-relief', literally 'low relief', aguardiente 'liquor', literally 'water-burning', malquerido 'unloved', literally 'badly-loved'.
The data in (1) show the productive endocentric left-headed compounding patterns of Spanish that have nominal non-heads ([[N+N].sub.N] and [[N+prep+N].sub.N]). Those in (2) represent fight-headed compounding with nominal non-heads, classified according to the grammatical category of the head as nominal ([[N+N].sub.N]), adjectival ([[N+A].sub.A], [[N+i+A].sub.A]), (3) and verbal ([[N+V].sub.V]) (heads are italicized).
(1) Left-headed endocentric compounds a. [[N+N].sub.N] hombre anuncio, referendum-farsa, coche cama man-poster, referendum-farse, car-bed 'poster man, rigged referendum, sleeper car' b. [[N+prep+N].sub.N] ojo de buey, patas de gallo, libro de cocina eye of ox, feet of rooster, book of cooking 'porthole, crow's feet, cookbook' (2) Right-headed endocentric compounds a. [[N+N].sub.N] baby futbol, euskerahablante, tour-operador, baby soccer, Basque-speaker, tour operator, 'little league soccer, speaker of Basque, tour operator, puticlub hooker-club stripper bar' b. [[N+A].sub.A] (4) drogadicto , telebobo, insulinodependiente drug addicted, tele-silly, insulin-dependent 'drug addicted, couch potato, insulin-dependent' c. [[N+i+A].sub.A] ojialegre, alicaido, boquimuelle eye-happy, wing-fallen, mouth-soft 'of happy eyes, crestfallen, of soft mouth (said of a horse)' d. [[N+V].sub.V] pelechar, aliquebrar hair-lose, wing-break 'molt, break the wings (of a bird)'
The left-headed nominal patterns ([1a]-[1b]) are undoubtedly the most productive endocentric patterns in modern Spanish, followed by ([2b][2c]). The rest are either neologistic (2a) or archaic (2d) (Rainer 1993: 290).
A closer inspection of the data reveals that the distribution of left-and right-headed compounds follows a strikingly regular morphological pattern. In head-initial endocentric compounding, the non-head nominal appears in its full form, including its gender/class inflectional suffix (or word class marker, henceforth WCM, Harris 1991, cf. also Morin 1999): hombr-e anunci-o. In head-final compounds the nominal non-head lacks a WCM: oj-i-alegre (cf. oj-o). (5)
It is argued here that this morphological difference alone can be used to account for the two different head positions in Spanish endocentric compounds with nominal non-heads. The main argument to be presented and supported is that these compounds are head-initial, with the non-head occupying the complement position. This underlying structure surfaces whenever non-head nominals appear with their WCM. Surface exceptions result from the insertion of stems instead of words in the non-head position. That may occur because the non-head is inserted in its bare form (noun minus WCM): drog-adicto, 'drug addict(ed)' (cf. droga). It may also be an unassimilated foreign word lacking a Spanish WCM: tour operador 'tour operator'. For these forms to surface, the non-head must be adjoined to the left of the head as the only way to circumvent the requirement of word class marking for case checking. After adjunction, a linking vowel may be inserted between the stem non-head and the head: oj-i-alegre 'happy of eye', literally 'eye-i-happy', al-i-quebrar 'break the wings (of a bird)', literally 'wing-i-break'. (6)
The present paper is organized as follows: section 2 presents a brief overview of the theoretical assumptions behind the analysis. Section 3 provides quantitative evidence that the position of the head in Spanish endocentric compounds is indeed related to the morphological makeup of the non-head nominal. Section 4 proposes underlying structures for both left- and right-headed compounds. Section 5 provides additional independent support to the proposal. Section 6 deals with what could be considered counterevidence and shows that it does not invalidate the main argument. Section 7 shows the advantages of my account compared to previous ones, and section 8 concludes the paper.
2. Theoretical assumptions
The present analysis is based on the notion that word formation and phrase formation involve essentially the same processes and can be accounted for uniformly (in accordance with Baker 1988; Lieber 1992; Hale and Keyser 1992, 1993, 1997; and Marantz 1997). This view of word formation is particularly apt to account for compounding, which has traditionally been described as the most syntactic of word formation processes, even by authors who strictly separate morphology and syntax (cf. Anderson 1992, who accounts for compounds with syntax-like rules in the morphological component). The present account assumes that compounding involves word formation on the basis of lexical phrases, that is, it includes NPs, APs, PPs, and VPs, but excludes DPs, DegrPs, and IPs (cf. the lexical-phrase/functional-phrase hypothesis in Miller 1993). This accounts for the existence of Spanish compounds such as [[N+prep+N].sub.N] dulce de leche 'caramel paste', literally 'sweet of milk', [[N+A].sub.N] aguardiente 'liquor', literally 'water burning', and the impossibility of parallel cases such as [[N+prep+DP].sub.N] * dulce de esta leche 'sweet of this milk', [[N+DegP].sub.N] * aguamuyardiente 'water-very-burning'. (7) The exclusion of functional phrases accounts for compounds being nortreferential and genetic and is in general agreement with the proposals in Marantz (1997), according to which the difference between morphology and syntax has to do with the categories combined rather than the nature of the combination itself.
The nominal elements involved in compounding, like all Spanish freestanding nouns, may have a complex morphological structure with at least two projections, that is, NP and ClassP (cf. Crisma 1995: 69; Bosque and Picallo 1996:372 for slightly different names for ClassP). This reflects the fact that nouns have a stem element and a WCM which is not present in derivatives: ded-o, ded-ito, ded-al 'finger, finger-dim., thimble' (Harris 1991). When they are compounded, however, nominals may appear in their full form, as nouns with WCM (hombr-e ran-a), or in their bare NP form, that is, without any intermediate class projection (oj-i-alegre).
The data used to quantify endocentric compounds with nominal non-heads are mostly from Bustos Gisbert (1986), a vast, homogeneous, dictionary-based database of nominal and adjectival compounds. Verbal compounds were obtained from the Real Academia Espanola (1992) and Rainer (1993).
All the data were inspected and reclassified whenever necessary. For example, Bustos Gisbert (1986) includes yerba de coral 'scarlet pimpernel, Anagallis arvensis L.', literally 'herb of coral' in the [[N+N].sub.N] section, together with its variant yerba coral 'herb of coral', literally 'herb coral'; the first variant was transferred to the [[N+prep+N].sub.N] class. Data were discarded when their internal structure was not completely transparent or could not be confirmed. For instance, in the noun malguiyu 'evil eye', only the first element is clearly an independent morpheme (real 'evil') in standard Spanish, while the identity of the second element is opaque. Other data were discarded because although formally they belonged in a given category, they had undergone conversion and were therefore exocentric. For example, papialbillo 'genet', literally 'maw-white', is classified by Bustos Gisbert as a [[N+A].sub.N], because of its nominal meaning. However, structurally it is a [[N+A].sub.A], which has undergone A [right arrow] N conversion. In other cases, theoretical considerations led me to discard some items from the database. For example, following the lexical-phrase hypothesis, any item that presented a referential expression, either a determiner phrase or a proper name, was discarded. This excluded [[N+prep+det+N].sub.N] formations such as arbol de la cera 'wax myrtle', literally 'tree of the wax'. In total, 35 such expressions, that is, around 5% of the total data, were discarded, and have not been included on the tables for the sake of clarity.
For each category, the total numbers of left- and right-headed compounds with nominal non-heads were considered first (Table 1). Examination of the data shows that nominal compounding is considerably more productive in contemporary Spanish than any of the other types: 66.8% nominals vs. 30.6% adjectivals, and 2.6% verbals. As for word order, nominal compounds exhibit a marked preference for the left-headed pattern (100% of [[N+prep+N].sub.N] compounds and 83.7% of [[N+N].sub.N]). For the rest of the categories, right-headed compounds constitute 100% of the cases. (8)
The next aspect to consider is whether the non-head element is a stem or a word. For the two left-headed patterns considered, the non-head position is virtually always (99.6% of the time) occupied by a full word, that is, stem and WCM: bocacalle 'intersection', literally 'mouth-street', dulce de leche 'caramel paste', literally 'sweet of milk' (Table 2).9 On the other hand, 94.9% of right-headed compounds have a stem in the non-head position: al-i-cortar 'break the wings (of a bird)', literally 'wing-break', while only 5.1% have a full word: casamuro 'wall, rampart', literally 'housewall' (Table 3).
For the sake of symmetry, I also considered the head element in left-headed compounds, to determine whether loss of the WCM is simply a function of position, that is, whether it affects both heads and non-heads when they appear on the left. Table 4 shows that the vast majority of heads appear in their full form (97.4%), even when they are on the left. Notice, however, that the existence of some preposed heads without WCM does not constitute a problem for this analysis, which is based not on the morphological structure of the head, but on that of the non-head.
In what follows, I propose an analysis that accounts for the two possible positions of the head in Spanish endocentric compounds and links those positions to the morphological form of the nominal non-head. As I stated in section 2 and showed in section 3, when the non-head nominal appears to the right of its head, it has the morphological structure of a freestanding word. When it appears to the left, it is a bare stem. In spite of this difference in position, I propose a single underlying word order in compounds with nominal non-heads, namely, one where this non-head follows the head, in a head-complement configuration. This approach is in line with proposed universals of head position (Kayne 1994), an elegant and economical solution in agreement with Lieber (1992), Miller (1993), and Marantz (1997).
The position that the non-head nominal should be considered a complement of its head deserves some explanation. Justification is straightforward in the case of [[N+V].sub.V] compounds, where the nominal non-head generally acts as its direct object, and is therefore its complement: aliquebrar = quebrar las alas (a un pajaro) 'wing-break = break the wings (of a bird)'. (10) For [[N+A].sub.A] compounds of the type boquiabierto 'open-mouthed', literally 'mouth-opened', the nominal acts as a complement of the adjective (Garcia Lozano 1993:211; Gonzalez Olle and Casado Velarde 1992: 106) in a Greek accusative construction (cariblanco 'white-faced' < blanco de cara 'white of face', i.e., 'white as far as the face is concerned').
Where there seems to be the least agreement about the complement status of the nominal non-head is in the case of compounds with the structure [[N+N].sub.N]. In some approaches, the non-head nominal is considered a modifier of the head, having somehow converted to the category adjective. Thus, if the compound sueldo base 'basic salary', literally 'salary base' can be paraphrased or has an equivalent which is a N+A phrase (sueldo basico 'basic salary', literally 'salary basic') this means that the same type of head-modifier relationship holds between the two (Bartos 1980; Cianca Aguilar 1986). However, this kind of argument is weak because the alleged denominal adjectives have some fundamental differences in behavior from regular adjectival modifiers. Firstly, they lack the number and gender concord of adjectives: sueldo base 'salary-sg base-sg' sueldos base 'salary-pl base-sg' (cf. * sueldos bases 'salary-pl base-pl'). Secondly, unlike most regular adjectival modifiers in Spanish, these non-heads cannot be preposed to their heads: sueldo base vs. * base sueldo. The same kinds of objections could be raised to the proposal that nominal stems (peli- 'hair', cf. pelo 'hair') are in fact adjectival modifiers (Clements 1992: 161), given that they too lack the agreement features of adjectives (* pelisrojos 'red-pl-haired-pl', literally 'haired-pl-red-pl') and their freedom of position (* rojo peli 'red-hair'). (11)
Additionally, [[N+N].sub.N] sequences fail the test of modification presented in Cormack (1998: 111), according to which adjunction gives conjunction. In other words, if a construction AB involves a adjunct/modifier--head relationship, it is possible to say "it is A and it is B," where the most stable property is encoded in a noun and the less stable one in an adjective: a red apple is an apple and is red. In [[N+N].sub.N] this is not possible: a garden chair is not a chair and a garden, a sueldo base is not a sueldo 'salary' and a base 'base'.
If the non-head nominal is not a modifier, that is, not an adjunct in specifier position, the logical alternative is to propose that it is a complement. (12) This means that the non-head noun should follow the head, as per Kayne (1994). (13) For nominal compounds involving nominal non-heads preceded by a preposition (e.g. dulce de leche 'caramel paste', literally 'sweet of milk'), the structure includes a [N.sub.1] head and an [N.sub.2] non-head. The head assigns inherent case to the lower N[P.sub.2] (Chomsky 1995: 114) via an intervening ClassP. The complement checks class features (or raises to pick them up by left adjunction) against the head of ClassP, which in turn checks case features against the head of the higher NP (ClassP of higher NP omitted for brevity). The inherent case surfaces through a dummy preposition de, overtly present in the [[N+prep+N].sub.N] pattern in (3) (for the nonprepositional status of de cf. Bosque 1993). In head-initial [[N+N].sub.N], the head/non-head relationship is still the result of some form of inherent albeit nonovert case assignment (4). (14)
(3) [ILLUSTRATION OMITTED]
(4) [ILLUSTRATION OMITTED]
Let us now consider the cases where the non-head is a bare nominal stem. I propose that bare stems are NPs that lack ClassP and higher projections (following Crisma 1995: 68). Given that the missing ClassP is the location of case checking for N[P.sub.2], the derivation will crash when the lower nominal is incapable of checking its case features. The N in the lower N[P.sub.2] is therefore forced to undergo head-to-head movement via left adjunction to circumvent the visibility condition (Baker 1988; Kayne 1994) (5).
(5) [ILLUSTRATION OMITTED]
This analysis can be extended to compounds where the head is an adjective or a verb assigning case, by simply replacing the higher N by A or V (6). Thus, a single structure can account for the position of the non-head in cases such as [[N+N].sub.N] (puti-club), [[N+A].sub.A] (oji-alegre), and [[N+V].sub.V] (pel-eehar).
(6) [ILLUSTRATION OMITTED]
To summarize, the analysis presented here accounts for the left-headedness of Spanish endocentric compounds with a nominal non-head by positing the left as the unmarked head position, in agreement with universal head-complement order. On the other hand, the occurrence of right-headed cases is driven by morphological requirements of the non-head nominals. When these elements are embedded in a projection without ClassP, they cannot check case features and the derivation will crash unless they undergo head-to-head adjunction to circumvent the case requirements of nominals and ensure visibility. (15)
5. Further evidence: learned compounding and blending
Evidence for the validity of this account can also be obtained by considering two related areas of word formation, namely, learned compounding and blending. Both of these word formation processes are generally distinguished from regular compounding, the first one because it involves manipulation of foreign stems, and the latter because it is irregular in its application. The fact that even marked and sporadic word formation processes like these also follow the principles sketched above is strong evidence of the generality of this account.
Learned compounds are regularly right-headed, including those where both (7a) or just one (7b) of the elements is a Greek or Latin stem with no independent existence in Spanish.
(7) a. psicologia, fotografia, termometro psychology, photography, thermometer (cf. * logia, * grafia, * metro in these senses) b. psicoterapia, fotosensible, termolabil psychotherapy, photosensitive, thermolabile (cf. terapia, sensible, labil)
This type of word formation can be accounted for straightforwardly if we bear in mind that learned stems lack a WCM and the corresponding ClassP projection by definition. This means that they do not project beyond the NP-level. Therefore, even if they underlyingly occupy a complement position to the right, they are incapable of checking case against that projection and they must adjoin if the derivation is not to crash (8). Notice that it is not suggested here that these compounds were created in Spanish following its native rules of compounding, but rather, that their inverted surface order can be accounted for in synchrony exclusively by invoking the morphological makeup of the non-heads involved, without having to posit a different set of compounding rules.
(8) [ILLUSTRATION OMITTED]
This adjunction operation of the non-head element accounts rather neatly for the existence of a compounding pattern that seems so contrary to the word order of Spanish, without any need to posit two different head parameters for native and learned compounding.
Further evidence for the present account comes from blends, that is, witticisms where one or both of the combined morphological words lose some of their phonetic material through overlap (Pharies 1987; Pineros 1998). When this loss affects the end of the head element or the beginning of the non-head, no reordering of constituents ensues: dictadura 'dictatorship' + blanda 'soft' > dictablanda 'permissive dictatorship'. However, when it is the end of the non-head element (including its WCM) that is lost, then it must be placed before the head: centro de servicios 'center of services' > servicentro 'gas station', lentes para pupilas 'lenses for pupils' > pupilentes 'contact lenses', biblioteca de charla 'library of chatting' > charloteca 'library where people chat instead of reading'. Thus, even in an area of word formation that is not as regular as compounding, and which is intentional and conscious in nature, the rule of left-adjoining a non-head that lacks a WCM still holds.
6. Apparent counterevidence
The present analysis predicts that non-head nominals will appear with their WCM when they follow the head, and in their bare stem form when they precede it. Therefore, any instances where non-heads appear without a WCM in word-final position or with it in word-initial position constitute counterevidence. This section presents those cases and shows that they are in fact lexically deviant, that is, either archaic or neologistic, and can be excepted without resorting to ad hoc mechanisms. (16)
The database presents only two cases where the complement appears after the head without its WCM: puercoespin 'porcupine', literally 'porkspine' and its variant cuerpoespin literally 'body-spine'. However, they are completely isolated and have a counterpart puercoespino (Real Academia Espanola 1992), so it is legitimate to discard them as anomalies.
Second, we must consider cases where the non-head appears before the head and has its WCM. These "less known constructions" (Rainer 1993: 261), are not particularly numerous in the database (fifteen out of a total of 296 rightheaded compounds), but they still need to be examined. The list includes [[N+N].sub.N] such as cabrahigo 'wild fig', literally 'goat-fig', casamuro 'rampart', literally 'house-wall', gallocresta 'cockscomb', literally 'rooster-comb'. There is one ambiguous case in the [[N+V].sub.V] pattern arch. salpresar < sal + prensar 'salt down', literally 'salt-press', where the WCM on sal is [empty set] (cf. caramarcar 'put on makeup', literally 'face-mark' in Nunez Cedeno 1992: 140, not found in Real Academia Espanola 1992).
These cases can be explained as archaisms: at least nine are attested since medieval times and several can be shown to be calqued on Latin head-final patterns (cabrahigo < L. CAPRIFICUS, gallocresta < L. GALLICRISTA, both in A.Pal., 1490). (17) Additional evidence of the pattern's archaism is the later reorganization of elements in individual compounds (arch. gallocresta vs. mod. cresta de gallo 'cockscomb'), and sometimes in an entire pattern (cf. the replacement in [[N+A].sub.A] of nominal words by bare stems, e.g. bocarroto 'mouth-broken' > boquirroto 'mouth-broken', Munthe 1889).
Although Bustos Gisbert (1986) has no other exceptions, contemporary Spanish shows some [[N+N].sub.N] head-final neologisms: tour operador 'tour operator', baby alarma 'baby alarm', disco club 'discotheque' (Rainer 1993: 261-262). However, these examples could be excluded from consideration because the non-heads are borrowed and their internal morphological makeup is still opaque to most Spanish native speakers. Occasionally, the right-headed pattern is calqued, and in this case one does find a few examples where the non-head element contains its WCM: libro-forums 'book forums', gitano-soul 'gypsy soul' (Rainer 1993). These calques go against the predictions of this proposal, but it should be noted that for the most part they are transient journalistic creations (cf. Rainer and Varela 1992: 126). (18)
On occasions, the status of the leftmost non-head is ambiguous: dineroterapia 'money-therapy', organoterapia 'organ-therapy'. However, parallelism with other -terapia compounds shows that the -o- element is best interpreted as a linking vowel rather than as a WCM: aeroterapia 'airtherapy', vitaminoterapia 'vitamin-therapy' (where aero- and vitamino- are not free forms). Consequently, the analysis as diner-o-terapia, organ-oterapia is justified. (1)
To summarize, exceptions fall into two general categories: archaisms and occasional neologisms based on foreign patterns. The right-headedness of earlier compounds and of neologisms is often the result of borrowing or calquing from sources that have right-headed compounding (e.g. English) or modeling of native word formation on learned patterns. It is predicted that archaic forms can only be interpreted correctly by piecemeal learning, and neologisms are restricted to certain registers where contact with and borrowing from foreign patterns is frequent. Neither of those two categories of compounds are frequent, and both are unlikely to be encountered by children during acquisition. (20)
7. Previous accounts
Piera (1995), an account of compound headship assignment within the minimalist framework, aims to explain the difference in head position between English and Spanish [[N+N].sub.N] compounding by relating it to the different morphological makeups of the two languages. Like my account, Piera's makes crucial use of the observation that Spanish nouns, unlike those of English, have a derivational stem and an inflectional WCM, yielding a double bracket on the right edge: [[[X].sub.d]Y].sub.i]. He links this structure to compounding via a double bracket restriction (DBR): "[a] double bracket at the edge of a word blocks adjunction of a word" (Piera 1995: 306). Although ambiguous in its formulation, Piera applies his restriction to block the adjunction of a non-head to a complex head with the structure [[[X].sub.d]Y].sub.i], that is, stem+WCM (cf. also Liceras and Diaz Rodriguez 2000, for the same interpretation of Piera's restriction). Consequently, all right-headed compounds are ruled out as ungrammatical in Spanish, because they present the wrong context for adjunction, rather than the wrong adjunction elements.
The solution presented is not satisfactory, however, both on descriptive and theoretical grounds. For one thing, its basic prediction that no right-headed compounds are possible in Spanish, is not borne out, even when we confine ourselves to the [[N+N].sub.N] pattern: gallipuente 'elevated bridge', literally 'rooster-bridge'. The claim is further debilitated when considering other compounding categories, in particular [[N+A].sub.A], [[N+V].sub.V], both of which are exclusively right-headed, as shall be recalled from Table 3.
Additionally, it can be proven that absence of a WCM from the head, which eliminates the double bracket, does not result in left-adjunction of the non-head: cornicabra 'goat-horn', literally 'horn-goat' (cf. cuern-o). If it were true that adjunction to the left was blocked by the head's WCM, then the compound should be * cabracorn, given that the structure of the head is [[X].sub.d] and the DBR would not apply. Finally, the account assumes that head-to-head movement will occur unless it is blocked, in violation of procrastination and greed (Chomsky 1995). Piera fails to present a reason why the non-head needs to move to satisfy its own morphological requirements, which makes it hard to justify that it should move at all.
The account proposed in Kampers-Mahne (2000), which also contrasts Romance and Germanic compounding, is much more in line with the one presented here. Like hers, our analysis is based on the rejection of the righthand head rule (Williams 1981). To justify the two possible orders in compounds, her account also makes crucial use of the complex internal structure of Romance words (with stems and WCMs, which she refers to as roots and inflectional elements). The main differences between her account and mine reside in the relationship assumed between head and non-head element and in whether it is the morphological makeup of the head or of the non-head that triggers the movement (cf.  and  for tree representations).
(9) Kampers-Mahne (2000) (21)
a. Left-headed compounds
b. Right-headed compounds
(10) Present account
a. Left-headed compounds
b. Right-headed compounds
While Kampers-Mahne proposes that the non-head occupies the specifier/ adjunct position and is therefore preposed to the head, I propose that it is a postposed complement. In her account, the head moves leftward to satisfy its morphological requirement of a WCM. In so doing it moves past the non-head and ends in a position to its left. In my account, it is the non-head that moves in the same direction and for the same reason, that is, to satisfy its morphological requirements. It thus raises past the head to a position on its left by head-to-head adjunction.
There are two considerations which make the present analysis superior to Kamper-Mahne's. The first argument is that there are examples where a compound head is preposed to its non-head, and yet it has no WCM (cf. [11a], where the WCM is replaced by a linking vowel, and [11b], where it is absent altogether). In other words, the movement proposed by Kampers-Mahne must have occurred, since the head appears to the left of its non-head, and yet there is no evidence that adjoining to the WCM was the actual motivation for the movement. Although ultimately the analysis could be saved by arguing that a surface phonological operation has applied as a result of lexicalization, as it stands, it is in violation of procrastination and greed, because no apparent motivation can be invoked for a proposed movement operation.
(11) a. pavitonto, boticuero (cf. the full words pavo, bota) turkey-stupid, boot-leather 'stupid, wineskin' b. matalbahaca, sauzgatillo (cf. the full words mata, sauce) bush-basil, willow-acacia flower 'basil plant, chaste tree berry'
The second consideration which makes the present analysis superior to Kampers-Mahne (2000) is that it only allows head movement for both right- and left-headed compounds (cf. [10a] and [10b]). In contrast, her analysis is based on the movement of heads in the case of left-headed compounds but requires the combined movement of heads and maximal projections for the right-headed cases (cf. [9a] and [9b]), in violation of Baker (1988) and Roeper (1999). The type of analysis presented here is therefore preferable because it is simpler and in line with current generative theory.
This paper set out to show that all Spanish endocentric compounds with nominal non-heads share the same underlying representation. They are all left-headed, in accordance with head-complement order. This underlying representation emerges whenever the elements inserted in the non-head node are full nominals. However, when the non-head is a stem, then it lacks a ClassP in its projection and is therefore incapable of checking case. To circumvent this requirement, the non-head must move to the left of the head by move a, leaving behind a properly governed trace. This account is valid for all endocentric compounds with nominal non-heads, regardless of the category of the head.
This solution is superior to previous ones in that it covers all endocentric compounds with nominal non-heads and accounts for virtually all of them, with the exception of some archaic and neologistic patterns. It can be successfully extended to learned compounding and other types of word formation, such as blending, which is evidence of its predictive power. Finally, in line with syntactic and morphological theory, it obeys the principles of procrastination and greed: only those elements move that would not satisfy their own morphological requirements if they stayed in situ. It is a matter of future research whether and how these insights can be extended to other types of endocentric compounds, namely, those with adjectival or adverbial non-heads, where the head/non-head relationship is possible one of head-adjunct.
Table 1. Compounds with a nominal non-head (n = 801) Cpd. type Left-headed (%) Right-headed (%) Total (%) [[N+N].sub.N] 154 (83.7) 30 (16.3) 184 (23.0) [[N+prep+N].sub.N] 351 (100) 0 (0) 351 (43.8) [[N+A].sub.A] 0 (0) 245 (100) 245 (30.6) [[N+V].sub.V] 0 (0) 21 (100) 21 (2.6) Total 505 (63) 296 (37) 801 (100) Table 2. Non-head nominals in left-headed compounds (n = 505) Cpd. type Non-head word (%) Non-head stem (%) [[N+N].sub.N] 152 (98.7) 2 (1.3) [[N+prep+N].sub.N] 351 (100) 0 (0) Total 503 (99.6) 2 (0.4) Table 3. Non-head nominals in right-headed compounds (n = 296) Cpd. type Non-head word (%) Non-head stem (%) [[N+N].sub.N] 14 (46.7) 16 (53.3) [[N+A].sub.A] 0 (0) 245 (100) [[N+V].sub.V] 1 (4.8) 20 (95.2) Total 15 (5.1) 281 (94.9) Table 4. Internal structure of heads in left-headed compounds (n = 505) Cpd. type Head = word (%) Head = stem (%) [[N+N].sub.N] 143 (92.8) 11 (7.1) [[N+prep+N].sub.N] 349 (99.4) 2 (0.6) Total 492 (97.4) 13 (2.6)
(1.) The first version of this article was written while I was still a student at the University of Florida. I am indebted to the members of my dissertation committee, Gary Miller, David Pharies, Ann Wehmeyer, and Shifra Armon, for their comments and suggestions. I also wish to thank the audience of LSRL XXX, held in Gainesville, Florida in 2000, in particular Karlos Arregi, Rafael Nunez Cedeno, and Carlos Eduardo Pineros. Finally, I am grateful to two anonymous reviewers from LSRL XXX and two from Linguistics, who did much to improve the quality of this paper. Any remaining errors are my own. Correspondence address: Department of Spanish and Portuguese Languages and Literatures, San Diego State University, 5500 Campanile Drive, San Diego, CA 92182-7703, USA. E-mail: firstname.lastname@example.org.
(2.) The notion of "head" may be established based on different criteria (cf. Zwicky 1985), some syntactic and some semantic. Here, I adhere to a strictly syntactic definition and view the semantic subset-superset relationship that holds between compound and head as a consequence of syntactic structure.
(3.) There is some confusion in the literature regarding the headedness of these compounds. Thus, for example, Nunez Cedeno (1991) considers compounds of the type caderiancho 'of wide hips', literally 'hip-wide' as exocentric, arguing that the whole compound stands in a non-head relation to the individual to whom the quality of having wide hips is ascribed. However, this is not a result of exocentricity but of the compound being an adjective, modifying an external referent. The endocentric structure proposed here is based on the categorial identity of the compound and its rightmost element ([caderiancho.sub.A], [ancho.sub.A] de caderas), a fact which can be explained if the latter has percolated its features to the former.
(4.) These examples readily nominalize through A [right arrow] N conversion, a very common word creation device in Spanish. The adjectival nature of these compounds can be demonstrated by the possibility of using adverbial modification with them: muy drogadicto 'very addicted to drugs', completamente telebobo 'completely silly due to watching TV', totalmente insulinodependiente 'totally insulin-dependent', and their gender concord: nina teleboba 'girl turned silly from watching TV'.
(5.) One reviewer suggests that the word order may be related to the grammatical category of the head element: when the head is a nominal, the non-head follows it; when it is a verb or an adjective, the non-head precedes it. The existence of nominal compounds of the type in (2a) makes this claim problematic, however. What all three compound types in (2) have in common is that the non-head nominal lacks a WCM and appears to the left. A unified account that takes into consideration both of these communalities has theoretical simplicity and elegance.
(6.) The insertion of the linking vowel shall be considered a PF-readjustment and will not be analyzed any further. The matter of how to determine whether such a vowel will be present and which one will be selected in a given case remains open for future research (cf. Mascaro 1986:77-82 for Catalan).
(7.) Lexical items with internal functional projections (French trompe l'oeil literally 'fool-the-eye', English will-o'-the-wisp, Spanish correveidile 'scandalmonger', literally 'run-go-and-tell-him') do not constitute genuine counterevidence. These forms, called syntactic freezes, are internally complex, but they are structurally isolated, semantically idiosyncratic, and do not result from the application of productive word formation patterns (cf. Miller 1993: 92).
(8.) This difference between compounds with nominal heads, on the one hand, and those with verbal or adjectival heads, on the other, deserves comment. It is not obvious why [[N+N].sub.N] compounds should have the option of having the head on either side, whereas [[N+V].sub.V] and [[N+A].sub.A] should have no [[V+N].sub.V] or [[A+N].sub.A] counterparts. The fact is that those structures do exist, but are not recognized as compounds. The left-headed equivalent of ojialegre 'happy of eye' is alegre de ojos 'happy of eye', with similar semantic and lexical restrictions. Note also, the impossibility of inserting functional elements in the structure: alegre de * los/ * estos/ * sus ojos 'happy of the /these/his eyes'. The left-headed equivalent of [[N+V].sub.V] is constituted by verbs with incorporated nouns which can be shown not to be the complement of the verb, since the latter can still take an independent direct object: dar vuelta la esquina 'to turn the corner', literally 'to give [turn.sub.N] the corner'. For more on the analysis of nominal incorporation to verbs and prepositions, of. Masullo (1996).
(9.) The two exceptions, puercoespin 'porcupine', literally 'pork-spine' and its variant cuerpoespin literally 'body-spine' will be analyzed further in section 6.
(10.) For a different approach to Catalan compounds of a similar structure, cf. Gracia and Fullana (1999).
(11.) Here I assume that the only legitimate cases of N [right arrow] A conversion are those where the non-head nominal has acquired the concord characteristics of adjectives and/or their freedom of position (perro, as in dia perro 'horrible day', literally day-m.sg.-dog-m.sg., vida perra 'horrible life', literally 'life-f.sg dog-f.sg'). In all other instances, the non-head is still a noun.
(12.) The view that the second nominal is in fact a complement is implicit in works such as de Dardel (1999), where NN compounds are called rectionnels, assuming that there is a government relationship between the first and second element, and in Bouvier (2000), where French compounds such as age d'or 'golden years', literally 'years of gold', are identified as instances of complimentation liniaire [linear complementation].
(13.) This position differs from di Sciullo (2000) in that here no difference is assumed to exist between phrase and word level, as far as predicate and argument relations are concerned. At both levels, the non-head noun is configurationally a complement. It is therefore not assumed that the head-adjunct relationship is canonical and overrides head-complement in compound structures. In the present analysis the word order differences found are a consequence of morphological structure alone, not of any morphology/syntax split.
(14.) Evidence for this analysis is the alternation of [[N+N].sub.N] and [[N+prep+N].sub.N] patterns, such as telaraha 'cobweb', literally 'cloth-spider' and tela de arana, 'cloth-spider', literally 'cloth of spider'.
(15.) One reviewer notes that the present analysis accounts for the Spanish data, but not for English and other languages with head-final compounding. This is true, in the sense that if we assume the head of N-N compounds to be universally on the left, then all English N-N compounds would involve movement and adjunction, a situation which I acknowledge is problematic. A possible alternative would be to posit different underlying representations for Spanish and English compounds (i.e., the non-head is a complement in Spanish but a specifier in English). However, this is not warranted by semantic differences between compounds in the two languages. The data from acquisition do not help to settle the matter, either. Evidence from acquisition of Spanish compounds by speakers of a variety of other languages (Liceras and Diaz Rodriguez 2000) suggests that when opting for the N-N pattern, learners have no trouble in correctly producing left-headed compounds. The matter warrants further research.
(16.) Note that the analysis is not invalidated when the head appears in its bare form, because the WCM on the compound HEAD constituent is not at stake. It is assumed here that these WCM losses result from phonological restructuring, that is, the loss of independent prosodic status for the first element in a small number of cases in the [[N+prep+N].sub.N] and [[N+N].sub.N] patterns. Some can be accounted for as the merger of identical phonetic segments: matalbahaca < mata + albahaca 'basil plant', literally 'plant-basil'. Others are reductions of the WCM to a linking vowel: cornicabra < cuerno + cabra 'goathorn', literally 'born-goat'. In other cases, the head WCM is deleted altogether: sauzgatillo < sauce + gatillo 'chaste tree berry, Vitex agnus-castus L.', literally 'willow-acacia flower'. Though not problematic for this account, cases such as these constitute a problem for other analyses (cf. section 7).
(17.) Apart from those mentioned above, the database includes casamuda 'act of moving house', literally 'house-move', maestrepasquin 'tract', literally 'master-tract', varaseto 'cane fence', and zarzamora 'blackberry', literally 'bramble berry'.
(18.) Note that these nominals are still structurally defective in that they lack a number inflection even if clearly plural: libroforum vs. * librosforum 'book forum', where numerous books are likely to have been displayed.
(19.) Examples form specialized jargons such as medicine, with heavy influence from learned and foreign word formation, generally fall into this category: vasoconstriccion/ vasodilatacion 'blood-vessel constriction/dilation' (I have to thank R. Nunez Cedeno for this example). These kinds of compounds are starting to appear in the general lexicon as well. For example, in autorradio 'car radio' and autopiezas 'car parts' the first element is the full word aut-o 'car', in violation of the predictions made here. However, the fact that there is a learned stem auto- (with a different meaning: autoanalisis 'self-analysis'), may have favored the placement of auto 'car' in the same position. In other words, the meaning comes from the word, but the placement comes from the homophonous learned element. It is worth noting that many educated native speakers reject these forms as aberrations.
(20.) It is not implied, however, that changes in the headedness of Spanish compounds are not possible or effectively under way. In fact, because lexical innovation often involves borrowing of compounds and Spanish often borrows from English, whose compounds are right-beaded, it seems highly likely that head position could change over time. This does not invalidate the main claims of this paper, however, which are about current native compounding.
(21.) The nomenclature of the nodes has been slightly modified to bring it in line with that of the present analysis, but the configuration has been kept identical to that in Kampers-Mahne (2000).
Anderson, Stephen (1992). A-Morphous Morphology. Cambridge: Cambridge University Press.
Baker, Mark (1988). Incorporation. Chicago: University of Chicago Press.
Bartos, Lubomir (1980). Acerca de las formaciones N N en el espanol actual. Boletin del Instituto de Filologia de la Universidad de Chile 31(1), 481-487.
Benavides Tarrius, Carlos (1996). Headedness in endocentric Spanish compounds. Torre de Papel 6(2), 29-60.
Bosque, Ignacio (1993). Sobre las diferencias entre los adjetivos relacionales y los calificativos. Revista Argentina de Linguistica 9(1-2), 10-48.
--; and Picalio, Carme (1996). Postnominal adjectives in Spanish DPs. Journal of Linguistics 32(2), 349-385.
Bouvier, Yves-Ferdinand (2000). Definir les composes par opposition aux syntagmes. Generative Grammar in Geneva (GG@G) 1, 165-187.
Bustos Gisbert, Eugenio de (1986). La composicion nominal en espanol. Salamanca: Ediciones Universidad de Salamanca.
Chomsky, Noam (1995). The Minimalist Program. Cambridge, MA: MIT Press.
Cianca Aguilar, Elena (1986). La aposicion sustantivo-sustantivo. Fremdsprachen 30, 246-248.
Clements, Joseph Clancy (1992). Lexical category hierarchy and 'head of compound' in Spanish. In Theoretical Analyses in Romance Linguistics, Christiane Laeufer and Terrell Morgan (eds.), 151-166. Amsterdam and Philadelphia: Benjamins.
Cormack, Annabel (1998). Definitions. Implications for Syntax, Semantics, and the Language of Thought. New York and London: Garland.
Crisma, Paola (1995). On the configurationally nature of adjectival modification. In Grammatical Theory and Romance Languages, Karen Zagnna (ed.), 59-71. Amsterdam and Philadelphia: Benjamins.
de Dardel, Robert (1999). Composes rectionnels nominaux nom + nom en protoroman. Probus 11(2), 177-208.
di Sciullo, Anna Maria (2000). Asymmetries: consequences for morphological configurations and paradigms. Acta Linguistica Hungarica 47(1-4), 81-101.
Garcia Lozano, Francisco (1993). Los compuestos de sustantivo + adjetivo del tipo pelirrojo. In La formacion de palabras, Soledad Varela (ed.), 205-214. Madrid: Taurus Universitaria.
Gonzalez Olle, Fernando; and Casado Velarde, Manuel (1992). Spanisch: Wortbildungslehre. In Lexikon der Romanistischen Linguistik. Aragonesich/Navarresisch, Spanisch, Asturianisch/Leonesisch. Aragones/Navarro, Espanol, Asturiano, Leones, Gunter Holtus, Michael Metzeltin, and Christian Schmitt (eds.), 91-109. Tubingen: Niemeyer.
Gracia, Lluisa; and Fullana, Olga (1999). On Catalan verbal compounds. Probus 11(2), 239-261.
Hale, Kenneth; and Keyser, S. Jay (1992). The syntactic character of thematic structure. In Thematic Structure. Its Role in Grammar, Iggy Roca (ed.), 107-143. New York: Foris.
--; and Keyser, S. Jay (1993). On argument structure and the lexical expression of syntactic relations. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Kenneth Hale and S. Jay Keyser (eds.), 53-109. Cambridge, MA: MIT Press.
--; and Keyser, S. Jay (1997). On the complex nature of simple predicators. In Complex Predicates, Alex Alsina, Joan Bresnan, and Peter Sells (eds.), 29-65. Stanford, CA: Center for the Study of Language and Information.
Harris, James (1991). The exponence of gender in Spanish. Linguistic Inquiry 22(1), 27-62.
Kampers-Mahne, Brigitte (2000). La composition en francais et en neerlandais. Leuvense-Bijdragen 89(1 2), 157-172.
Kayne, Richard (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Liceras, Juana; and Diaz Rodriguez, Lourdes (2000). Aspects of the acquisition of grammar: triggers in L2 acquisition: the case of Spanish N-N compounds. Studia Linguistica 54(2), 197-211.
Lieber, Rochelle (1992). Deconstructing Morphology. Chicago: University of Chicago Press.
Marantz, Alec (1997). No escape from syntax: don't try morphological analysis in the privacy of your own lexicon. U. Penn Working Papers in Linguistics 4(2), 201-225.
Mascaro, Joan (1986). Morfologia. Barcelona. Biblioteca Universitaria. Enciclopedia Catalana.
Masullo, Pascual J. (1996). Los sintagmas nominales sin determinante: Una propuesta incorporacionista. In El sustantivo sin determinacion: La ausencia de determinante en la lengua espagola, Ignacio Bosque (ed.), 169-200. Madrid: Visor.
Miller, D. Gary (1993). Complex Verb Formation. Amsterdam and Philadelphia: Benjamins.
Morin, Regina (1999). Spanish substantives: how many classes? In Advances in Hispanic Linguistics: Papers from the 2nd Hispanic Linguistics Symposium, Javier Gutierrez-Rexach and Fernando Martinez-Gil (eds.), 214-230. Somerville, MA: Cascadilla.
Munthe, Ake (1889). Observations sur les composes espagnols du type aliabierto. Recueil de memoires philologiques presente a M. Gaston Paris par ses eleves suedois, 31-56. Stockholm: Central-Tryckereit.
Nunez Cedeno, Rafael (1991). Headship assignment resolution in Spanish compounds. In Current Studies in Spanish Linguistics, Hector Campos and Fernando Martinez-Gil (eds.), 573-598. Washington, D.C.: Georgetown University Press.
--(1992). Headship assignment in Spanish compounds. In Theoretical Analyses in Romance Linguistics, Christiane Laeufer and Terrell A. Morgan (eds.), 131-149. Amsterdam and Philadelphia: Benjamins.
Pharies, David (1987). Blending in Spanish word formation. Romanistisches Jahrbuch 38, 271-289.
Piera, Carlos (1995). On compounding in English and Spanish. In Evolution and Revolution in Linguistic Theory, Hector Campos and Paula Kempchinsky (eds.), 302-315. Washington, D.C.: Georgetown University Press.
Pineros, Carlos Eduardo (1998). Prosodic morphology in Spanish: constraint interaction in word-formation. Unpublished doctoral dissertation, Ohio State University.
Rainer, Franz (1993). Spanische Wortbildungslehre. Tubingen: Niemeyer.
--; and Varela, Soledad (1992). Compounding in Spanish. Rivista di Linguistica 4(1), 117-142.
Real Academia Espanola (1992). Diccionario de la lengua espanola. Madrid: Espasa Calpe.
Roeper, Thomas (1999). Leftward Movement in Morphology. MIT Working Papers in Linguistics 34, 35-66.
Williams, Edwin (1981). On the notions 'lexically related' and 'head of a word.' Linguistic Inquiry 12(2), 245-274.
Zwicky, Arnold M. (1985). Heads. Journal of Linguistics 21(1), 1-29.
San Diego State University
Received 13 August 2002
Revised version received
14 March 2003
|Printer friendly Cite/link Email Feedback|
|Author:||Moyna, Maria Irene|
|Publication:||Linguistics: an interdisciplinary journal of the language sciences|
|Date:||May 1, 2004|
|Previous Article:||Light and heavy reflexives (1).|
|Next Article:||Shouldn't it be breakfunch? A quantitative analysis of blend structure in English *.|