Phrasal Spell Out: An Argument From Haplology Antonio Fábregas University of Tromsø Abstract. One of the ways in which theories that use Late Insertion differ is with respect to whether exponents can be associated with complex syntactic phrases (Phrasal Spell Out) or not. In this article we provide evidence for Phrasal Spell Out taken from morphological haplology phenomena in Spanish. We argue that some nominal affixes, like -idad ‘-ity’, spell out a phrase [FP [NP]]. The adjectival version of the structure involves projecting AP between FP and NP ([FP[AP[NP]]]). In this structure, an exponent like -idad cannot be used because the features it spells out do not form a single constituent; another exponent, -ist(a) ‘ist’, has to be used. The surface result is that -idad is haplologised when -ist(a) is introduced, even though the morphosyntactic features of -idad are still present in the structure. 1. Spell out: a controversy In neo-constructionist theories (Halle & Marantz 1993; Marantz 1997; Julien 2002; Borer 2005a, 2005b, 2013; Ramchand 2008; Embick 2010; Panagiotidis 2011; Arregi & Nevins 2012) the lexicon is a repository of morphological exponents associated with morphosyntactic features (1). These exponents encode morphophonological information, and the morphosyntactic features they are associated with determine their context of insertion. Such theories presuppose some form of Late Insertion (see Marantz 2013: 154); once the computational system has defined a particular structure, a post-syntactic component—where the lexicon is placed—scans the syntactic information and selects specific exponents from its repertoire. (1) /EXPONENT/ <—>[X, Y] Beyond these common assumptions, there are lots of different theories about the particular format lexical entries should have. One Linguistic Analysis, 39, 1-2 2014 Linguistic Analysis P.O. Box 2418, Vashon, WA 98070 Antonio Fábregas 2 of the points of disagreement, and the one we will concentrate on in this article, is whether an exponent has to be linked to a terminal node—a head—or is it possible to link it also to a complex syntactic constituent—a phrase–, in what has been called ‘Phrasal spell out’ (eg., Caha 2009)? In other words: is a lexical entry like (2), where a single exponent spells out a whole phrase, possible? (2) /EXPONENT/ <—> XP XY Distributed Morphology (Bonet 1991, Halle & Marantz 1993, Noyer 1997) assumes (2) is not possible. When a single exponent produces features that are independently shown to belong to different syntactic heads—cumulative exponence (Stump 1998), a postsyntactic operation called Fusion is assumed. That operation maps two syntactic heads (X0) into one single morphological node (M0). In contrast, other theories—significantly Nanosyntax (Svenonius, Ramchand, Starke & Taraldsen 2009)—accept an entry like (2), where the insertion of the exponent in the non-terminal node XP spells out the features contained under that node, which in (2) are X and Y. This procedure, with some technical differences we will remain neutral about, has been used by Weerman & Evers-Vermeul (2002) to analyze the expression of case in pronouns, by Neeleman & Szendrői (2007) to explain the lack of competition between pro and an overt pronoun, by Ramchand (2008) to account for the different aspectual behaviour of verbs, and by Caha (2009) to restrict the syncretism patterns in case exponence. It is extremely difficult to argue in favour or against Phrasal Spell Out on theoretical grounds. The idiosyncratic nature of the lexicon makes it difficult to impose plausible restrictions on the shape of lexical entries; claims about whether it is more or less parsimonious to stipulate a post-syntactic level with operations like Fusion or whether it is better to introduce a complication in the format of lexical entries generally come hand in hand with different assumptions about modularity and complexity which are sometimes difficult to compare with each other. For this reason, the domain where we will try to find arguments to support one theory over the other is the set of empirical phenomena predicted by each one of them. Phrasal Spell Out: An Argument from Haplology 3 1.1. A distinct empirical prediction In this article we will focus on one domain where the two theories make a different prediction: what are the possible outcomes when another head is projected between two heads that are spelled out by a single exponent? Imagine a situation like (3a), and assume that there is one single exponent, /blah/, that lexicalises both X and Y—either by Phrasal Spell Out or by Fusion. Then consider (3b). Both theories predict that the same exponent should not be used in (3b). In Phrasal Spell Out, (3b) cannot be lexicalised by /blah/ because there is no single node that contains both X and Y without containing H (Caha 2009). With Fusion, on the other hand, /blah/ is equally unavailable, but the reason is that X and Y cannot undergo Fusion because X and Y are not sisters (Bobaljik 1994). (3)a.XPb.XP X Y X HP HY Where the two theories differ is in what the spell out of (3b) should be. In Phrasal Spell Out, given that lexical items can be associated with complex constituents, there could be a single exponent /don/ (4a) that spells out the tree in (3b). If (3a) and (3b) correspond to affixes, the surface result would be a seemingly paradoxical one: even though (3b) has the morphosyntactic features of /blah/, and there has been an increase in the structural complexity of the syntactic tree, there will not be an increase in the complexity of the morphological marking. /blah/ will not emerge, and it will be replaced by the single exponent /don/ (4b). In other words: the morphosyntactic features associated with /blah/ are still in the tree and will be interpreted by syntax and semantics, but /blah/ will not be spelled out in the phonology. (4)a. /don/<—>XP X HP HY Antonio Fábregas 4 b. [[base]-blah] + H —> [[base]-don] An account where Fusion is necessary to allow the spell out of features distributed among different heads makes a different prediction. Given that Fusion is not recursive,1 adding another head will necessarily imply an increase in the morphological complexity of the word. Several specific outcomes are possible, depending on the fusion operations one assumes are triggered by H and whether one assumes that the exponent /blah/ can be inserted in heads that carry a subset of its features. If H and X fuse as /don/ and /blah/ is introduced in Y, (5a) emerges. If H and Y fuse as /don/ and /blah/ is introduced in X, (5b) emerges. If there is no fusion, one expects each head to be spelled out by a different exponent, whether /blah/ is one of them (5c) or not (5d). In all cases, adding a new head implies an increase in morphological marking with respect to the version that lacked that head. (5) a.base-blah-don b.base-don-blah c. base-bluh-don-blih d. base-blah-don-blih In this article, we will examine haplology—the situation where seemingly a whole exponent is cancelled when the word is further derived—in Spanish. Building on this empirical phenomenon, our main goal will be to argue that one kind of haplology in Spanish illustrates the situation described in (4), and therefore provides evidence that supports Phrasal Spell Out. The secondary goal of the discussion is to show that there are no reasons to treat haplology as a purely morphological phenomenon, and that a syntactic explanation can successfully account for a wide set of patterns which are not necessarily covered by lexicalist analyses. 1 Fusion cannot be recursive for the following reason: it is an operation that takes two syntactic sisters (X0) depending from the same category node and maps them into a single position of morphological exponence (M0) (cf. Halle & Marantz 1993, Noyer 1997). Thus, once two heads have been fused together, we obtain one M0, not another X0. If we try to fuse a third head, the operation would involve one X0 and one M0; as X and M are objects belonging to different levels of the grammar, and modularity states that no operation can refer to both at the same time (see Scheer 2011). See also §5 for a discussion of why an account that uses head movement and impoverishment / obliteraton misses some crucial generalisations. Phrasal Spell Out: An Argument from Haplology 5 The article is structured as follows. In §2 we will introduce the relevant data, which have to do with two haplologies caused by the affix -ist(a). In §3, we will address one kind of haplology which is successfully analysable as an alternation of two derivations from the same root (Marantz 2001), and we will explain why the other kind noted in §2 is not amenable to that analysis. In §4 we analyse this second haplology, arguing for a Phrasal Spell Out analysis of the affixes involved. In §5 we present our conclusions. 2. Haplology in Spanish There is a relatively abundant literature on haplology2 (Aronoff 1976, Wurzel 1976, Dressler 1977, Stemberger 1981, Plénat 1991, Corbin & Plénat 1992, Plag 1998, de Lacy 1999, Ortmann & Popescu 2001, McBride 2004, Volpe 2005, Fradin, Montermini & Plénat 2007, Xu 2007). In Spanish, the phenomenon is not so widely studied. Here we will concentrate specifically on a case of haplology which involves the suffix -idad, roughly translated as ‘-ity’ in English, in contexts where -ist(a),3 roughly equivalent to English ‘-ist’, is inserted.4 2 We leave aside the so-called syntactic haplology effects or OCP-in-syntax effects, situations where derivations involving identical heads or heads containing quasi-identical sets of features are banned. See Neeleman & Van de Koot (2006), Van Riemsdijk (2008), Richards (2010) or Grohmann (2011) for such cases. We remain grateful to an anonymous reviewer for pointing this out to us. Note that the Phrasal spell out-related haplology is not equivalent to these syntactic cases, as it does not involve truncation (or any other modification, for that matter) of the syntactic structure underlying the exponents. 3 When representing affixes, we show between parentheses ‘(...)’ the noun marker, which, as we will see, is always absent from the morphophonological representation when the word is further derived. 4 In part of the literature (e.g., Aronoff 1976, and Marantz 2001 when discussing Aronoff’s analysis) this situation is referred to as “truncation.” We will use the term “haplology” in order to avoid confusion with the phonological operation that takes a whole word and cancels one of its segments without contact with another affix, which is more commonly referred to as truncation in the literature. Note that whether the word is morphologically complex (ia) or not does not play a role for this operation. (i) a. universidad > uni ‘university’ b. Federico > Fede ‘Frederick, proper name’ c. profesor > profe ‘teacher’ This phenomenon has been analysed mainly from a phonological perspective; see Roca & Felíu (2003), Piñeros (2000) and Alber & Arndt-Lappe (2012) for some recent proposals. Antonio Fábregas 6 2.1. Haplologies triggered by -ist(a). The suffix -ist(a) triggers two kinds of haplology that we will analyse differently. The first kind of haplology is the one that it triggers when the base, a noun, ends in the sequence -is (6). When a word in -ist(a) is derived from it, the expected sequence /isis/ is simplified to /is/. (6)a.tenis> ten-ist(a)*tenis-ist(a) ‘tennis’ ‘tennis player’ b.tesis>tes-ist(a)*tesis-ist(a) ‘dissertation’ ‘graduate student’ c. mímesis>mimet-ist(a)*mimesis-ist(a) ‘mimesis’ ‘mimetist’ In one case there is a “double” haplology of sorts: the base noun contains a sequence of two /is/, and both disappear in contact with -ist(a). (7) análisis>anal-ist(a) *analisisista, *analisista ‘analysis’ ‘analist’ Although we will not adopt this analysis (cf. §3), it is conceivable that this particular haplology is phonologically-induced; given the phonological identity between the last segment of the base and the first segment of the affix that is adjacent to it, the sequence gets simplified and the repetition is avoided. This would imply a phonological analysis of haplology, one that in Nevins’ (2012) classification would be a “vocabulary-insertion-level dissimilation” (Nevins 2012:105) which is sensitive to (a) adjacency and (b) segmental non-distinctness at a phonological level. As such, this haplology would have the same status as the examples in (8), where haplology is not sensitive to whether the segment that gets simplified is a morphological unit or just a phonological sequence inside the stem. Note that sometimes the result is ineffability (8a, b), sometimes the sequence gets simplified (8c, d, e), and sometimes two versions, one with haplology and one without it, coexist (8f). Phrasal Spell Out: An Argument from Haplology 7 (8)a. Honig + ig>*honig-ig honey + ish ‘honey-ish’ [German, Dressler 1977] b. fish + ish > ??fish-ish [English, Nevins 2012] c. nutri + trix> nutrix*nutritrix feed + er ‘woman that feeds’ [Latin, Dressler 1977] d. edel + eling > edeling *edeleling noble + man nobleman [Dutch, de Lacy 1999] e. trágico + cómico > tragicómico *tragicocómico tragic + comic tragicomic [Spanish, Corbin & Plénat 1992] f. morpho + phonologie > morphonologie morphophonologie morpho + phonology morphonology morphophonology [French, Corbin & Plénat 1992] The cases illustrated in (8) have received a phonological analysis which concentrates on how to explain the impossibility of repeating a sequence in these combinations while at the same time not predicting that in general languages will reject any case of phonological repetition under adjacency, as it happens internally to some roots (cf. French venénéux ‘poisonous’) and generally in cases of reduplication. Phonological analyses typically discuss whether this haplology is a case of deletion or coalescence, with the current consensus being more inclined to the latter (de Lacy 1999, McBride 2004, Xu 2007). Note however that if the phonological analysis of truncation in the above cases is potentially tenable, the following case of truncation is not easily amenable to a phonological analysis. In the following cases, a word ending in the suffix -idad in Spanish is further derived with the suffix -ist(a). In these cases, only -ist(a) emerges, although—as witnessed by the gloss—the meaning of the complex word presupposes the meaning of the word derived by -idad. This phenomenon is noted in some descriptive grammars, such as RAE & ASALE (2009: 8 Antonio Fábregas §6.9e); the examples in (9) are taken from this source. (9)a. public-idad> public-ist(a)*public-idad-ist(a) publicity, ad-man advertisement b. electr-ic-idad> electr-ic-ist(a) *electr-ic-idad-ist(a) electricityelectrician The semantics of the form shows us that in these cases we have real haplology and not simply an alternation between two affixes taking the same base. In (9a) the base public- can be used as a noun in Spanish: it means ‘audience,’ and as an adjective ‘public.’ At least in the first meaning, it would not be impossible for it to combine directly with the affix -ist(a), and we can even imagine the meaning the word would have: ‘someone that works with an audience.’ However, this is not the meaning that publicista has. Its meaning ‘someone that works in the advertising industry’ is built over a non-compositional meaning that the word public-idad has, but not its base públic(o): ‘advertising industry.’ Thus, the semantic interpretation shows us that at a structural level we do not have an alternation between two morphemes, which clashes with the surface result, where the two exponents alternate.5 As always when there are few examples documented in dictionaries and grammars, one wonders whether this haplology is a lexical accident or a really systematic phenomenon that speakers have intuitions about. In the first case, when confronted with new lexical formations that are not established in the standard Spanish varieties, we expect speakers not to have haplology, and will allow forms which contain the sequence -idad-ist(a). In the second case, 5 One anonymous reviewer notes that there is a second interpretation of these data that should be discussed: assume a theory like Arad (2003), where roots can get special meanings under lexical categorisers. It could be argued that public- gets assigned a special meaning under -idad, and the same special meaning would be assigned under -ist(a), making haplology unnecessary. However, there are reasons to reject this account. Note, first, that in the words used in the experiment reported in (10) the base is not a root, so Arad’s theory could not be used here in its literal form. It is true that a version of the analysis like Borer’s (2012), where non compositional meanings can be assigned provided no functional head intervenes, could still be used. However, the crucial problem would be that in this case we would have to guarantee that two unrelated derivations, public-idad and public-ist(a), get assigned the same non compositional meaning, which is unexpected to the extent that non compositional meanings are unpredictable. Phrasal Spell Out: An Argument from Haplology 9 we expect that, even if they have not heard the word before, speakers will reject that sequence. In order to determine which status this haplology has, I conducted an experiment. 25 native speakers of European Spanish received a questionnaire, and were asked to give their judgements about some new words in Spanish. In each question I described the situation where a new formation that involved the suffixes -idad and -ist(a) could be used, and then gave the speaker three options. (10) is an example of the kind of question posed, with the options. (10) A group of fools creates a new political movement called pasividad [‘passivity’], which claims that all problems in the modern world are caused by people who try to be proactive, and recommends that its followers approach all possible problems by not acting, and by letting things happen. The followers of this movement are called: a) pasivistas [‘passivists’, with haplology] b) pasividadistas [‘passivitists’, without haplology] c) pasistas [‘passists’] Ten questions had this shape and were oriented towards determining whether native speakers allow -idad-ist(a) sequences. The options were randomized, so that the version with haplology did not always appear in the first place. An overwhelming majority of answers chose the version with haplology, and no speaker ever chose the version where the sequence -idad-ist(a) was documented. In total, out of 250 answers, we obtained 243 where the version with haplology was preferred, which is an unusually high result.6 Consider now why a phonological analysis of the haplology is unlikely. It could not be the kind of phonological haplology that is sensitive to the shape of the vocabulary items introduced in spell out, simply because there is no phonological identity between the two affixes, -idad and -ist(a). However, it could still be some kind 6 The unexpected answers always involved the choice of the third, shorter, form of the base. For instance, in one of the questions, the movement was called caloricidad ‘caloricity,’ and the proponents recommended consuming highly caloric food. Although 23 speakers opted for the expected form calor-ic-ist(a)s to name the followers, two speakers preferred the form calor-ist(a)s. These speakers did not show a systematic preference for the shorter forms in their answers. 10 Antonio Fábregas of phonologically-based haplology that takes place through prosodic phrasing (Nevins 2012: 94), and which is sensitive to more abstract notions, such as stress placement. It could also be that haplology is the result of a phonological process that, even without identity, cancels part of the sequence for other phonological reasons that are sensitive to the segmental content of the exponents. In the next few paragraphs I will show why this kind of analysis is unlikely. The sequence that we would obtain if both affixes co-occurred is shown in (11). (11)-idadista Note that the syllabification of this sequence is unproblematic, and almost perfectly complies with a CV pattern where the only coda consonant is a coronal. Syllabification is not a likely candidate to trigger the haplology, i.e.: (12)i.da.dis.ta Considering now stress, the affix -idad is oxytone. In Spanish, when the base is oxytone, adding to it the suffix -ist(a) does not generally imply cancelling (part of) the syllable that carries stress. Note in (13) that this is so even if the non-cancellation implies having a syllable without onset. (13)maná>mana-íst(a) *man-ist(a) ‘manna’ Finally, if we consider metrical feet, note that each affix—before and after resyllabification—contains two syllables, so no degenerate foot will arise either. (14) a.(i.dad) b.(is.ta) There is a second affix that has been systematically claimed to be cancelled by -ist(a), and that is -ism(o), which is historically related to it. Consider the example in (15). (15) andaluc-ism(o)> andaluc-ist(a) *andaluc-ism-ist(a) Andalusian-ism ‘supporter of Andalusism’ Phrasal Spell Out: An Argument from Haplology 11 Although subtle, the interpretation of andalucista shows that the word presupposes the meaning of the word with -ism(o). If andalucista was directly derived from Andalucía, a region in Spain, the meaning that we would expect is ‘supporter of Andalusia’ or ‘person that works in something related to Andalusia.’7 This is not the meaning of the word; it means ‘supporter of Andalucism.’ Andalucism is a specific political movement which is exclusive to Andalusia, but someone that supports Andalusia is not necessarily an andalucista. Although conceptually more difficult, it is also possible to conceive of a person who is andalucista, supports the movement, but is not a supporter of Andalusia. That person might think, for instance, that the political leaders of Andalusism deserve respect, while Andalusia itself does not. One could also ask how it is possible to know that -ist(a) triggers haplology of -ism(o), and not vice versa. That is, how do we know that the ordering of the derivation is not the one in (16), where -ism(o) would trigger haplology of -ist(a)? (16)Andalucía >andaluc-ist(a) > andaluc-(ist)-ism(o) ‘Andalusia’ Several arguments show that this derivation is wrong. The first has just been given: if (16) were right, we would expect andalucista to mean ‘supporter of Andalusiam,’ which is wrong. The second is that there is independent evidence that when -ism(o) combines with a word derived by -ist(a) there is no haplology. For the acceptance of -ist-ism(o) sequences, it seems to be necessary that the movement described is not widely known, and therefore that the words are not used frequently. For instance, in CREA—the corpus of contemporary Spanish hosted by the Royal Academy—we find (17). 7 Indeed, this meaning is available for some speakers, independently of whether they also have the one presented in (15). In fact, as an anonymous reviewer notes, in the Dictionary of the Royal Academy (21st edition), the first entry of andalucista is ‘specialist in topics related to Andalucía.’ This is not problematic for our account if we assume that andalucista, on the surface, can correspond to two different derivations: the one with haplology, where the base is andaluc-ismo, and a second derivation where the base is Andaluc-ía ‘Andalusia’ and there is no haplology of -ismo, corresponding to the meaning of the first entry. Anticipating the analysis a bit, in this second derivation the base would have the structure [NP -ía [AP andaluz ‘Andalusian’]]. Andalucista would be obtained by adding over it AP and FP layers, as in [FP [AP [NP -ía [AP andaluz]]]]. -ista would lexicalise the chunk between FP and NP, making -ía unavailable for spell out. The surface result would be identical, but the derivational history of the word would be different. Antonio Fábregas 12 (17) ...caer en el lud-ist-ism(o) médico ...fall in the ludd-ist-ism medical ‘...to fall into medical Luddistism’ [e-mail, 1999] Here we see that the speaker forms the word lud-ist(a) ‘Luddite,’ to describe a supporter of Ned Ludd, the fictional hero who argued against the use of machines to replace people in industrial production. The movement related to these followers is lud-ist-ism(o), where both affixes appear in a sequence and neither one haplologises the other. Along the same lines, we contrasted data with native speakers to see if they would accept the sequence -ist-ism(o) when the base is a non-existent item. In order to do that, we made up the name of several fictional historical figures who advocated for different ideals. Speakers accepted to a high degree formations like (18). (18)gronf-ist-ism(o) Gronf-ist-ism A third and final affix which has been claimed to haplologise in presence of -idad is -í(a). The examples in (19) are taken from RAE & ASALE (2009: §6.9e). (19) a. utop-í(a)> utop-ist(a)%utop-i-ísta ‘utopy’ ‘supporter of utopy’ b. regal-í(a)> regal-ist(a)%regal-i-ísta ‘royalty’ ‘supporter of the payment of royalties’ As in the previous cases, we conducted an experiment, and here the results where more mixed than in the case of -idad. Out of 25 speakers, 7 systematically preferred the forms where -í(a) did not undergo haplology in the presence of -ist(a), while 14 of respondents preferred the form with haplology. When confronted with a movement called entrop-ía ‘entropy,’ the first group preferred the form in (20a), without haplology, while the second group preferred the form with haplology (20b). (20) a.entrop-i-íst(a) entrop-y-ist Phrasal Spell Out: An Argument from Haplology 13 b.entrop-ist(a) entrop-ist The remaining speakers gave mixed judgements. If we focus now on the group of speakers that have haplology with -í(a), note that the reason for the haplology does not seem to be phonological either. A sequence -ií- can be created by combining -ist(a) with a base. All speakers consulted accepted the derived forms in (21), where the sequence is present, coming from bases ending in tonic /i/. None accepted the form without the diphthong. (21) a. PRI>pri-íst(a)*pr-ist(a) ‘certain political party’ ‘supporter of PRI’ b. Martí>marti-íst(a)*mart-ist(a) ‘certain soccer player’ ‘supporter of Martí’ The situation with the sequence -ería ‘-ery’ is similar. Although treated as a single morpheme by some authors (e.g., RAE & ASALE 2009: §6.3m-6.3p), other authors have suggested that at least in one of its meanings—the place where a particular action is performed, as in (22)—it should be understood as the combination of the suffix -er(o) ‘-er, maker’ and the suffix -í(a) (see Bustos & Santiago 1999: 4551 for discussion). (22)zapat-er-ía shoe- er-y ‘shoe shop’ This second segmentation is consistent with the results obtained in our study. Speakers where asked to name the person whose job is to work in some shop—for example, a helad-er-ía ‘ice cream shop’— as opposed to the person whose job it is to make the merchandise that is sold in that shop—in our example, helad-er(o) ‘ice-cream maker.’ Speakers divided almost perfectly in the two classes noted before: one group used the word with haplology in (23a), and the other group, the word with haplology of -í(a) (23b). No speaker went for the form where also the -er(o) part of the sequence is haplologised (23c). Antonio Fábregas 14 (23) a. helad-er-i-íst(a) b. helad-er-ist(a) c.*helad-ist(a) The following points summarise the empirical results: i. ii. iii. Speakers systematically reject sequences -idad-ist(a) Speakers systematically reject sequences -ism-ist(a) (but not ist-ism(o)) S peakers divide in two groups with respect to -i-ísta sequences: some haplologise systematically, some do not haplologise, also in a systematic way. The generalisation is not that -ist(a) systematically triggers the haplology of adjacent nominal affixes. (24) shows that other affixes different from -idad, -ism(o) (and -í(a)) are perfectly fine combined with -ist(a).8 (24) a. cens-ur-ist(a) censor-ship-ist ‘related to censorship’ 8 The exponent -idad apparently undergoes a second kind of haplology in Spanish, in contact with -eño. (i) Navidad > Navid-eño *Navidad-eño ‘Christmas’ ‘Christmas-like’ However, there are reasons to think that this is not a systematic haplology that has to be somehow captured through operations of any kind. First, note that it is not the affix, but a segment of it (/ad/) what is cancelled in contact with -eñ(o). Second, the word Navidad is very lexicalised, and if it is indeed derived with -idad in contemporary Spanish, it is not easy to recognise the base (Nav-, historically related to Latin nascior ‘to be born’). Second, combinations of -idad (or -id-) with—eñ(o) are otherwise absent from dictionaries. Third, speakers consulted do not accept -eñ(o) after -idad forms when the base is clearly derived, as in sant-idad ‘sanct-ity.’ Fourth, we ran a test where we offered speakers the choice between sets of words like (ii). (ii) a. orfan-dad-eño [without haplology] orphan-ity-ish b. orfan-d-eño [with partial haplology] c. orfan-eño [with total haplology] Speakers did not show any systematic preference for any of these forms. Some would pick the form with haplology in one case (iic), but with comparable examples they would pick either the one without haplology (iia) or the one with partial haplology (iib). What these seemingly chaotic results suggest is precisely what is expected if the form in (i) is not the result of any grammatical rule—phonological or otherwise— but a stored irregular form. Phrasal Spell Out: An Argument from Haplology 15 b. adop-cion-ist(a) adopt-ion-ist ‘supporter of adopting children’ c. arma-ment-ist(a) weapon-ment-ist ‘supporter of weapons’ 3. Marantzian haplology In this section we will show that the type of haplology that involves -idad and -ism(o) in combination with -ist(a) is different in significant respects from the one discussed first by Aronoff (1976) and later analysed, from a syntactic perspective, in Marantz (2001) and Volpe (2005). The kind of haplology studied by Aronoff (1976:88-98) refers to cases like (25)—for expository convenience, we segment the internal structure of the word, although obviously Aronoff does not do so in the original. (25) a. releg-ate >releg-able *releg-at-able b. nomin-ate>nomin-ee*nomin-at-ee Marantz (2001) discusses these cases, and notes that a haplology analysis only holds on one assumption: affixes that trigger haplology must attach to constituents that have a full grammatical category, in this particular case “verbs.” If we give up this assumption, there is a second, unproblematic analysis: that these words in -able and -ee are not derived from verbs in -ate, but from roots. Then the haplology stops being necessary; in the first forms of each pair, the root is categorised as a verb by a projection which is spelled out as -ate. In the second forms, the roots are nominalised or adjectivised by a different projection. Departing from Marantz’ notation, and siding with Borer’s (2012), we will use the labels V, N and A for the heads that categorise the root. Antonio Fábregas 16 (26) a.VPb.APc.NP V √ A √ N √ -ate releg- -able releg- -ee nomin nominThe procedure used to analyse these cases is similar to the one used in conversion, the situation where the same base, without any phonological change, manifests in two distinct grammatical categories (e.g., (to) dance and (a) dance); the two forms are not derived one from another, but both of them are derived from a common source, a root without grammatical category. The difference with this kind of haplology is that in cases of conversion, the head that categorises the root is null. (27) a.VPb.AP V √ A ø clear ø √ clear If this analysis of haplology is right, we expect that in at least some cases haplology should show non-directionality problems, like conversion does (Clark & Clark 1979)—which word is the basic form and which one is derived from it should not be clear. Indeed, as Marantz (2001) shows, this is borne out. Of the two words in (28), which one is derived from the other? We can interpret (28b) as the name that we give to the property in (28a), or (28a) as the property of causing the abstract concept in (28b). The problem, as in conversion, is solved if we assume that both words are derived from the same root (29). (28) a.atroc-ious b.atroc-ity (29) a. APb.NP A -ious √ N √ atroc- -ity atroc- See Volpe (2005: 26-33) for further evidence of this analysis. Phrasal Spell Out: An Argument from Haplology 17 3.1. Why this analysis does not work for our cases Note, however, that there are three reasons why the haplology of -idad by -ist(a) does not comply with what we have called Marantzian haplology. Marantzian haplology reduces to affix alternation from the same root. Real affix alternation implies that the morphosyntactic features of one affix should not be present when the other one appears, as there is no directionality between the two complex forms. However, in the case of -idad, we have seen evidence that there is directionality; in a form like public-ist(a) ‘advertising agent,’ the meaning of the base is the one denoted by public-idad ‘advertisement,’ not públic-o ‘audience.’ Secondly, Marantzian haplology predicts that in some cases the haplology will not be compulsory: the same affix, like -ble, might be able to combine both with a root and with a complex base. Aronoff (1976) noted minimal pairs like navig-at-able vs. navig-able, where Volpe (2005) argues that the first is a case of derivation from a verb, while the second is a case of derivation from a root. We see, however, no cases where -idad remains under -ist(a). This haplology is obligatory. Third, Marantzian haplology crucially requires that the base with which the relevant affixes combine be a root, as the effect is claimed to be an epiphenomenon of the fact that a root can alternatively combine with different categorisers. However, in the case of -idad, the haplology is not restricted to cases where the base is a root. Several of the examples that we provided in the questionnaire involved morphologically complex bases to which -idad attached (30). Also in these cases, speakers haplologised -idad in the presence of -ist(a). (30) a. tono > tón-ico > ton-ic-idad > ton-ic-ist(a) *ton-ic-idad-ist(a) tone tonic tonicity tonicist b. calor > calór-ico > calor-ic-idad > calor-ic-ist(a) *calor-ic-idad-ist(a) heat caloric caloricity caloricist We conclude, therefore, that Marantz’ analysis cannot be adopted for these examples. Antonio Fábregas 18 3.2. Haplology of -is Recall that -ist(a) intervenes in a second kind of haplology, namely the one which apparently deletes a segment -is which is part of the root (tenis ‘tennis’ > ten-ist(a) ‘tennis player’, *tenis-ist(a); mímesis ‘mimesis’ > mimet-ist(a) ‘mimetist’, *mimesis-ist(a)). In this section we will show that rather than being phonologically induced, this case is likely to be an instance of Marantzian haplology. We will shortly argue that the ending -is is not part of the root but an idiosyncratic nominal marker (NM), along the lines originally proposed in Harris (1991). As is the case with all nominal markers, it has to be cancelled when the noun is used in derivation as the base for another category. (31)[[ten]N is]NM Consider the evidence that -is can be analysed as a noun class marker. This means that -is forms a class with -a (cas-a, ‘house’), -o (perr-o, ‘dog’), -e (clas-e, ‘class’), and -ø (papel-ø, ‘paper’), which are the more regular markers of noun class in Spanish. Like these other four markers, -is is compulsorily unstressed (32). (32) a. c/á/s-a b. p/é/rr-o c. cl/á/s-e d. t/é/n-is vs. vs. vs. vs. *cas-/á/ *perr-/ó/ *clas-/é/ *ten-/í/s In fact, the stress pattern of words ending in -is can be explained following Oltra-Massuet & Arregi’s (2005) analysis of the mapping between prosody and morphological structure. These authors argue that in the nominal domain stress is assigned defining the right boundary of an iamb to the right of the highest nominal head (2005: 65-66), which in their analysis is n. Technical details aside, this always leaves noun markers to the right of the iamb, and therefore unstressed. This is evidence that -is is an idiosyncratic noun marker which is selected by the root. Words that exhibit this marker are generally nouns of Greek origin, occasionally Latin (33). (33) acrópol-is ‘acropolis’, apocalips-is ‘apocalypse’, apódos-is ‘apodosis’, bil-is ‘bile’, cox-is ‘coccyx’, cut-is ‘skin’, éxtasis ‘ecstasy’, epiderm-is ‘epidermis’, fimos-is ‘fimo’, narcos-is Phrasal Spell Out: An Argument from Haplology 19 ‘narcosis’, parális-is ‘paralysis’, perífras-is ‘periphrasis’, pub-is ‘pubis’, sintax-is ‘syntax’ Note that as expected from an exponent that is idiosyncratically selected by a set of roots, the noun marker must be immediately adjacent to the root. All the words in (36) show that the marker -is combines with bases that are roots, and no overt nominalisers—like -ción ‘-ation’ or -ment- ‘-ment’—can be present. Idiosyncratic insertion of vocabulary items can only take place when there is no other exponent intervening between the selector and the selectee (Embick 2010). If -is is an idiosyncratic noun marker selected by the root, we expect that in the presence of the diminutive affix, -is should be impossible. The diminutive affix is a constituent that is merged between the root and the noun marker (see Harris 1991, Eguren 2001, Bermúdez-Otero 2007, Fábregas 2013), so its presence would block the idiosyncratic selection between the root and the marker. This prediction is borne out. The diminutive of the noun in (34) cannot contain -is. The exponent -is is substituted by a noun marker -a, reflecting feminine gender. (34) tes-is>tes-it-a*tes-it-is thesis-NMthesis-dim-NM Note that this change in the exponent would be unexpected if -is were part of the root. When the root finishes in an unstressed syllable and mere concatenation would give odd phonological results, the diminutive acts as an infix (Bermúdez-Otero 2007). (35) Carlos>Carl-it-os CharlesCharles-dim Given an analysis where -is is a noun marker, its disappearance in front of -ist(a) is expected. Noun markers are systematically absent when the noun produces another grammatical category. Consider (36): the noun marker disappears when the noun is turned into a verb, and also when it is turned into an adjective. (36) a. tabac-o>tabac-al*tabac-o-al tobacco-NMtobacco-adj ‘tobacco related’ Antonio Fábregas 20 b. clas-e > clas-ifica *clas-e-ifica class-NM class-ify The explanation of why noun markers must be absent from the base when the noun is further derived is orthogonal to our purposes; we can provisionally suggest a principle like the one in (37). Given a hierarchy as the one in (38), this would mean that whenever the word changes grammatical category, the noun marker is not included in the base because operations that introduce a lexical categoriser like A or V always target a node below the noun marker position (represented in 38 as NMP, “Noun marker phrase”). We will not attempt to find a deeper reason why this should be so in this article. (37) Lexical categorisers always select a constituent lower than the noun marker. (38)NMP NM DimP Dim NP N √ Once we adopt a principle like (37) the explanation of why -is disappears in the presence of -ist(a) does not need to be phonological: -ist(a) combines with a constituent that excludes the noun marker. If this explanation is on the right track and the reason for the haplology is not phonological, then we expect affixes with a different phonological shape to produce the same haplology with -is. This prediction is borne out. (39) a. pub-is>púb-ico *pub-ís-ico pubis-NMpub-ic b. apocalips-is>apocalípt-ico *aplocalips-ís-ico apocalypse-NM apocalipt-ic c. éxtas-is>extát-ico*extas-ís-ico ecstasy-NM ecstat-ic Phrasal Spell Out: An Argument from Haplology 21 Thus, in such cases an exponent disappears because the head it is associated with (NM) is absent from the derived word. Let us now move to the haplology of -idad, which we will argue shows evidence for Phrasal Spell Out, because the exponent disappears even though the syntactic structure that underlies it is still included in the derived word. 4. Haplology and phrasal spell out What we will argue for in this section is the following. The nominaliser -idad is an exponent that spells out a phrasal constituent that includes a hierarchically low head (N) and a high head (which we will label FP9), as shown in (40). When the computational system generates a complex syntactic constituent where FP directly dominates NP, -idad is introduced to spell out that phrase. (40)FP FNP N -idad ...√ The adjectival version of (40), however, does not imply merging A over (40). Given the hierarchy of domains in the functional sequence (Cinque 2005, 2010), A is introduced above NP. Thus, to obtain a structure where a noun is adjectivalised and FP is present, the computational system generates a structure like (41). 9 The proper characterisation of the functional head here identified as FP remains to be developed. The evidence gathered here suggests that it is a head which directly relates to the activation of a rich qualia structure in a nominal domain, defining the situations which can be associated with the denotation of the noun; as such, it could be related to the structures that define the syncategorematic readings of some adjectives, like easy, and could even be considered parallel to the heads that define situations in the verbal domain. Antonio Fábregas 22 (41) FP F AP ANP N √ Here, although the morphosyntactic features associated with the exponent -idad are present (N and F), -idad cannot be inserted, because N and F do not form a single syntactic constituent to the exclusion of AP (no node contains both F and N without dominating A). Instead, -ist(a) is introduced, because -ist(a) is the spell out of a complex constituent [F [A [N]]]. Haplology follows, thus, as an expected effect of Phrasal spell out: if the morphosyntactic features associated with an exponent are present, but do not form a syntactic constituent, another exponent must be used (42). (42) FP F AP A NP -ist(a) N √ *-idad 4.1. Motivating the projections and their hierarchy One important part of our analysis is that -idad and -ism(o) (and, for some speakers, -í(a)) spell out two heads, one that is below A and another one that is above it. The evidence that these affixes spell out N can be provided easily: data like (43) show that it produces words belonging to the noun category, as witnessed by the fact that they can combine with determiners of all kinds. (43) a. la tranquil-idad the tranquil-ity Phrasal Spell Out: An Argument from Haplology 23 b. esta opac-idad this opac-ity Evidence that the designated position for adjectives inside the nominal domain is higher than nouns has been provided elsewhere. Cinque (2005) shows that the universally attested orderings of demonstratives, numerals, adjectives and nouns follow if one assumes that adjectives are introduced as specifiers of functional heads which are higher than N and lower than numerals. Cinque (2010) further elaborates the proposal, showing that the relative ordering of adjectives and the semantic interpretations that come associated with each adjectival position can be explained on the assumption that adjectives are base generated in positions higher than N. (44) XP APX X NP There is evidence that X must enter into a checking relation with its specifier. First, X is sensitive to the lexical category hosted by its specifier. For this reason, prepositional modifiers must be external to adjectival modifiers. See (45) for an illustration in Spanish. (45) a. la casa blanca enorme de ladrillo the house white huge of brick ‘the huge white brick house’ b. ??la casa de ladrillo blanca enorme the house of brick white huge The semantic ordering of a series of adjectives (e.g., Sproat & Shih 1991), which implies that X selects a class of adjectives, is also evidence for a checking relation. We assume, thus, that X gets adjectival features in the course of the derivation. Given this scenario, the assumption that A dominates N also in cases of category derivation is the default one. The difference with the cases in which the adjective is introduced as a modifier is that in those cases, the functional projections of the adjective (significantly, degree) are introduced in Antonio Fábregas 24 the specifier of the head (46a). When the noun is turned into an adjective, A is the head position and—assuming the lexical derivation stops there—Degree and other functional heads dominate N (46b). (46) a.XPb.DegP DegP XDeg AP Deg ...AP X NPA NP From here, the differences between a noun modified by an adjective and a noun turned into an adjective follow naturally. In (46a), the noun is not in a gradable constituent, and the rest of the nominal functional projections (Num and D) can be projected above XP. In (46b), the noun is in a gradable structure, and Num and D cannot be projected because Deg, an adjectival functional head, has been projected. For this reason, the syntactic distribution of (47a), corresponding to (46a), is different from (47b), which corresponds to (46b). In the first, a DP can be merged above XP because the adjectival lexical head did not dominate NP; in the second, AP and DegP dominate NP, so DP cannot be merged. (47) a. the very big house b. (*the) very child-ish Let us see now evidence for the existence of a high F projection which dominates both A and N. We would like to suggest that this functional head is the locus, among other things, of the expression of Qualia semantic information (Pustejovsky 1995). Inside Pustejovsky’s theory, a quale is a conceptual notion that gives the relational force of an item (1995:76)—it provides information about the constituent parts of an entity (Constitutive quale), its function (Telic quale), the conditions about how it was created (Agentive quale), or the properties that distinguish it inside a wider domain (Formal quale). The notion of quale has proved descriptively useful to deal with cases of coercion and semantic type shifts. For instance, a phase verb like begin must take as direct object constituents that denote events. Object-denoting nouns are only allowed if they have an agentive or a telic quale which relates the origin or the function of the object Phrasal Spell Out: An Argument from Haplology 25 to some event (48). In the first interpretation (48a), the reading is licensed by the agentive quale, which relates the book with the action of writing that creates it. In the second interpretation (48b), the telic quale is relevant, relating the book to its function (reading). Nouns which do not contain telic or agentive quale produce unacceptable results in this context (49). (48) I began the book. a. ‘I began writing the book’ b. ‘I began reading the book’ (49) #I began the grass. Even though in its original formulation qualia structure is a lexical notion, there are reasons to think that it can be at least partially syntactised and codified as a syntactic head that dominates the nominal constituent. First, there is evidence that some exponents spell out this F layer expressing quale. Johnston & Busa (1999) make a convincing case that prepositions like da and a in Italian play this role. Da (50) is used to denote the telic quale, and a (51) identifies the constitutive quale. Note that the noun they introduce can be modified by adjectives. (50) a. coltello da pane knife DA bread ‘bread knife’ b. bicchiere da vino (rosso) glass DA wine red ‘(red) wine glass’ (51) a. porta a vetri (rossi) door A glass ‘(red) glass door’ b. seni al silicone breasts A-the silicon ‘silicon breasts’ See also Lin & Liu (2005) for the idea that qualia structure can be expressed syntactically. Antonio Fábregas 26 There is also evidence that qualia cannot be defined at a lexical level, because one noun might lack qualia, while a complex constituent including the same noun and an adjective has it. Consider (52). Normally, nouns denoting natural objects lack telic and agentive quale, which makes them weird as a complement of a phase verb like begin. (52) a. #Empecé el agua. I.began the water b. #Empecé la piedra. I.began the stone However, some sets of N+A with the same nouns do have qualia structure and are thus acceptable in the same context. For instance, agua angélica, lit. ‘angelic water’ is a type of concoction, and piedra afiladera, lit. ‘sharpening stone,’ ‘whetstone,’ is a mixture of rocks and other substances used in construction. These sets of N+A contain agentive quale, and as such they can be used as direct objects of begin. (53) a. Empecé el agua angélica. I.began the water angelic, ‘I began the concoction’ b. Empecé la piedra afiladera. I.began the stone sharpening, ‘I began the whetstone’ Similar examples are arena ‘sand’ vs. arena de moldeo ‘moulding sand,’ aire ‘air’ vs. aire acondicionado ‘air conditioning,’ or tierra ‘ground’ vs. tierra de Siena, lit. ‘ground of Siena,’ a type of clay treated with rust used for dying clothes. These contrasts are consistent with a theory where qualia is not defined lexically in individual items, but has as its locus a syntactic projection that can dominate complex syntactic constituents, including at least the noun with adjectives and prepositional modifiers. 4.2. Evidence that -idad contains qualia Now we will show that there is evidence that -idad should be treated as a nominaliser that contains this FP layer which codifies qualia structure. Phrasal Spell Out: An Argument from Haplology 27 The traditional description of -idad (e.g., Alemany Bolúfer 1920: 40) is that it is a quality nominaliser, making abstract nouns from adjectives, and as such it would form a natural class with affixes like -ur(a), -ez(a) or -itud, which are used to create quality nominalisations. However, this description encounters one problem: -idad is also used to make collective nouns. The nouns in (54) are ambiguous between two readings: in one, they denote the quality expressed by the base adjective. In the second, they denote the group of all people characterised by the property denoted by the adjective. (54) a.cristian-dad Christian-ity ‘being Christian’ or ‘group of all Christians’ b.vecin-dad neighbour-ity ‘being neighbour to something / someone’ or ‘group of all neighbours’ c.human-idad human-ity ‘being human’ or ‘group of all humans’ d.herman-dad brother-ity ‘being brotherly’ or ‘group of all brothers or brethren’ The affixes -ur(a), -ez(a) or -itud never have this collective meaning. The problem caused by this ambiguity is not likely to emerge from accidental homophony; note that the collective reading emerges with both the allomorphs -idad and -dad, which means that the two homophonous affixes would have exactly the same allomorphs, something completely unexpected. Instead, if we assume that -idad contains a layer where constitutive quale can be expressed, the ambiguity can be explained. In both uses, the affix states that one must form a concept which denotes something whose constituent parts are expressed by the base. In the collective reading, the parts are the individuals that are characterised by a property. In the apparent quality nominalisation reading, in contrast, the constituent element is the property itself. Antonio Fábregas 28 For explicitness, the two readings can be differentiated by the presence of a [mass] or [count] feature in the N layer of the affix.10 The quality reading is associated to a mass noun (55a), and the concept expressed is simply the kind defined by the property of the adjectival base, without any individuation; the structure would be the one in (55b). In contrast, the collective reading is a count noun (56a), and the constituent parts would be the individuals defined by that property, with a structure like (56b). (55) a. *la-s hermandad-es entre Juan y María the-pl brotherhood-pl between Juan and María b.FP FNP N ...√ [mass] (56) a. las hermandad-es religiosa-s de España the-pl brotherhood-pl religious-pl of Spain ‘the religious brotherhoods in Spain’ b. FP F NP N ...√ [count] If this analysis is on the right line, we are claiming that while -ur(a), -ez(a) or -itud spell out single N heads, that of -idad includes a high functional projection, as shown in (57). 10 Alternatively, following Borer (2005a), we could assume a layer DivP over NP for count nouns. DivP, in Borer’s system, is a nominal functional head responsible for turning a mass—denoted by N or the root—into a bounded entity that can be combined with numerals; its presence is associated, then, with count readings. Our explanation is orthogonal to this aspect. Phrasal Spell Out: An Argument from Haplology 29 (57) a. NPb.FP N ... F NP ura -idad N... A straightforward prediction about affix ordering emerges from here. As -idad involves a high functional head, but -ur(a) does not, we expect that there will be a whole set of affixes that can follow -ur(a) but will be unable to follow -idad. These affixes would be those whose position in the functional hierarchy is higher than NP but lower than FP—for instance, adjectival affixes. This prediction is borne out. In (58) we illustrate it with four of the most productive adjectival affixes in contemporary Spanish. (58) a. cult-ur-alvs.*herman-dad-al cult-ur-al brother-ity-al b. escrit-ur-ariovs.*herman-dad-ario write-ing-ary brother-ity-ary ‘related to writing’ c. caricat-ur-esco vs. *herman-dad-esco caricat-ur-ish brother-ity-ish ‘similar to a caricature’ d. prem-ur-osovs. *herman-dad-oso11 urgent-nom-ous ‘with urgency’ 11 In the dictionaries, there is only one word where apparently,(i)dad is followed by -oso (i). (i) bon-dad-oso good-ness-ous ’kind-hearted’ However, speakers systematically reject new coinages where -idad is followed by this affix, which suggests that the sequence is not allowed, at least in contemporary Spanish, and that bondad- should be analysed as a single root which is further derived by -oso (ii). It is unclear that bondadoso ‘nice, kind-hearted’ has to be related in meaning to bueno ‘good,’ given its high semantic specialisation—it can only refer to humans, unlike bueno. See Embick & Marantz (2008) for other cases where a sequence that could be claimed to be historically derived has been reinterpreted as one single exponent. (ii)bondad-oso kind.heart-ed Antonio Fábregas 30 Similar observations can be made about -ism(o), but in this case its association to qualia structure is more straightforward: -ism(o) is used to form names of political, social and artistic movements which follow some particular set of principles, and it is typical to derive them from proper names, as in (59). (59) Marx>marx-ism(o) MarxMarx-ism The interpretation is that there is a movement that was set in motion by Marx or Marx’ ideals, which means that -ism(o) must be able to identify the base as its agentive quale. The restriction of the affix is, thus, that the base must be interpretable conceptually as something associated with a set of principles that can originate some movement. In a second reading, the affix identifies the telic quale, and the base describes the activity that members of a group perform. This is typical with names of sports and other hobbies. (60) piragü-ism(o) ‘canoeing,’ cicl-ism(o) ‘cycling,’ ajedrec-ism(o) ‘chess playing’ As expected, if -ism(o) also includes a FP layer, it cannot be followed easily by adjectivisers. Sequences like *-ism-al, *-ism-ario, *-ism-esco and *-ism-oso are unattested and rejected by native speakers when they correspond to affixes. 4.3. The structure of -ist(a) Let us now move to the structure of -ist(a). We propose that this exponent lexicalises the complex set of heads in (61); that is, the same structure as -idad or -ism(o), but with an A layer intervening between F and N. (61)FP F AP A NP N … Phrasal Spell Out: An Argument from Haplology 31 This is the same thing as saying that -ist(a) is the spell out of the structure where a noun is adjectivised and the whole structure is dominated by FP. Consider first evidence that the main use of -ist(a) is as an adjectiviser. This aspect is potentially controversial, as many of the words derived with -ist(a) can be used both as adjectives and as nouns—and some are actually more frequent as nouns. For instance, ajedrec-ist(a) can be used as a noun ‘chess player’ (62a) or as an adjective ‘related to chess playing’ (62b). (62) a. un ajedrec-ist(a) a chess-ist b. el mundo ajedrec-ist(a) the world chess-ist ‘the world related to chess’ The reason to consider words derived in -ist(a) as adjectives is that some of the forms can only be used as adjectives, while we are unaware of cases where the word can only be used as a noun. A relevant example is precios-ist(a) ‘affected,’ which rejects nominal uses (63a); another one, for many speakers, is arma-ment-ist(a) ‘related to weapons’ (63b).12 (63) a. ??Ha venido un preciosista. has come an affected Intended: ‘Someone affected has come.’ b. ??Ha venido un armamentista. has come a related-to-weapons Intended: ‘Someone related to weapons has come.’ Evidence that -ist(a) contains the layer FP is straightforward. As FP is the locus of qualia, evidence comes from the semantic interpretation of the words derived by -ist(a). As can be seen from the previous examples, -ist(a) tends to impose on its base a telic quale 12 An independent question which we will not develop here is how adjectives in -ist(a) get converted to nouns. Note that -ist(a) is by no means the only affix that does this; this is particularly frequent with relational adjective affixes (-al, -ario ‘-ary’, -ico ‘-ic’...) and other adjectives. A suggestive solution is the one proposed in Borer & Roy (2010), where nominalisation is a surface effect of an empty pronominal introduced as the subject of the adjective. Antonio Fábregas 32 interpretation: the base does not just denote an object, but it has to be an activity performed by someone. Consider now how this proposal about the structure of -ist(a) explains the haplology of -idad and -ism(o). When building a noun with a high FP projection, the computational system merges N with F, and FP projects (64). Here, -idad or -ism(o) can be introduced. (64) FP F NP N √ -idadpublic-ismociclIn contrast, the adjectival version of (64) involves projecting AP between F and NP, a movement forced by the functional hierarchy: AP is higher than NP, and FP is higher than both. Thus, (65) is built. Here, -idad and -ism(o) are impossible because FP and NP do not form a syntactic constituent to the exclusion of AP. It is in such cases that apparent haplology takes place; the morphosyntactic features associated to -idad or -ism(o) are present—which explains that the structure is not interpreted just as the adjectivisation of a root, but the exponent is unavailable, given the presence of AP, and instead -ist(a) has to be used. (65)FP F AP ANP N √ -ist(a) publicConsider now why -ist(a) does not produce haplology of suffixes like -ura. Unlike -idad, -ur(a) is the spell out of N, not an FP that contains NP. The projection of AP over NP will not make the exponent -ur(a) unavailable, because this projection does not break the syntactic constituent which -ur(a) spells out, as it is trivially one Phrasal Spell Out: An Argument from Haplology 33 single head. In (66), -ur(a) spells out N. (66) caricat-ur-ist(a) caricat-ur-ist ‘person who draws caricatures’ In order to derive (66), above the NP that is spelled out by -ur(a), the structure of -ist(a) is introduced. The lowest head of -ist(a), N, subordinates the N spelled out by -ur(a). (67) FP F AP A NP -ist(a) NNP N √ -ur- caricatMore about this subordination will be said in §4.4. Let us, finally, address the affix -í(a). Remember that speakers fell into two classes: some of them systematically haplologised it in the presence of -ist(a), and some of them systematically kept it before the same affix. These two alternative grammars can be dealt with if each group of speakers associates this item with different constituents. The group that does not haplologise it treats it basically as -ur(a); in their grammar, the affix is associated with N (68a). In contrast, speakers who haplologise it treat it like -idad, associating the affix with FP and NP (68b). (68) a.NPb.FP N ... F NP -í(a) -í(a)N ... What this means, if the analysis is right, is that speakers who haplologise -í(a) should at the same time reject that it is followed Antonio Fábregas 34 by affixes like -esc(o), -ari(o), -os(o) or -al, while speakers who do not haplologise it, should accept those combinations. There are some data that suggest this is on the right track. When confronted with new formations like (69), the speakers who did not haplologise the affix found them more acceptable that those that did not. (69) a. filo-log-í(a) > filo-log-i-esco philo-log-yphilo-log-y-esque ‘typical of philology’ b. agon-í(a)>agon-i-oso agon-y agon-y-ous ‘that causes agony’ c. cardenal-í(a)>cardenal-i-ario cardinal-ate cardinal-ate-ary ‘cardinalate’‘related to a cardinalate’ 4.4. An assumption that needs to be made: an asymmetry between adjectivalisation and nominalisation The attentive reader will have noticed by now that in order to explain the data we had to make an implicit assumption that we will now make explicit: i.e., that nominalisation can be obtained through subordination, but adjectivalisation is always a result of projecting A over N, as the functional sequence dictates. If adjectivisation could subordinate structures, then affix ordering data like (58) would be unexplained, as they could in principle correspond to a structure like (70). Now, if A is unable to subordinate, they follow from the proposal. (70) *AP A FP F NP... In contrast, nominalisation must be able to subordinate other structures, as an example like (18)—lud-ist-ismo ‘Luddistism’— shows. These words must correspond to a structure where over the Phrasal Spell Out: An Argument from Haplology 35 derivation corresponding to an -ist(a) adjective, a noun is defined and the functional sequence is restarted. (71)FP F NP Subordination N FP -ism FAP A NP -ist N ... Note that as -ist(a) contains at its lowest element an N projection, it is also expected to be in principle at least able to subordinate other structures. This is what we suggest happens in caricat-ur-ist(a) ‘person who draws caricatures’ (72), and this is the reason why the lowest N is still spelled out by -ur(a) in this context, as opposed to instances, such as the one discussed in footnote 6, where the base’s N layer is also spelled out by -ist(a). (72)FP F AP Subordination ANP -ist(a) N NP N √ -ur(a) caricatNote, in any case, that (72) is not the adjectivisation of the base noun caricat-ur(a) ‘caricature,’ unlike what happens when there is haplology of -idad or -ism(o). (72) is the adjectivisation of a noun related to the noun caricat-ur(a). Hence the semantics of the word, ‘related to the entity that does something with caricatures.’ Antonio Fábregas 36 We do not have a fully-fledged proposal about the deep reasons why A should be unable to subordinate structures, but we can offer some speculations. Baker (2003: 190-264) has argued that adjectives, as grammatical categories, are not defined by positive properties, but rather by the absence of the properties that nouns and verbs have. Unlike nouns, they do not carry an index of identity, and unlike verbs, they are unable to introduce their own subject (see also Baker 2008). Perhaps this inability to subordinate other structures is related to the absence of positive properties; if subordination implies resetting the functional hierarchy, a head without any substantive features will be unable to reset it because it will be unable to redefine the properties of its complement. Note also that, in contrast to traditional analyses, as the field evolves there are fewer and fewer constructions where an A head is used to subordinate other constituents. In syntax, of course, relative clauses are analysed essentially as nominal subordination with an operator-variable structure which allows it to act as a noun modifier. Embick’s (2004) analysis of participles—the primary example of alleged adjectival derivation of a verb—analyses the adjectivisation process as absence of verbal functional heads, not presence of adjectival heads. Oltra-Massuet (2010), in her analysis of Spanish -ble, another traditional example of subordination of V by A, convincingly argues that when the base is verbal an aspectual head and a modal head can do the work of what seems to be adjectivisation. Obviously, a complete argumentation that A cannot subordinate structures higher than NP must address the remaining cases where there is apparent adjectivisation from verbs, which goes beyond the goals of this paper. Here, therefore, this claim has the status of an assumption, but one that is consistent with the data and one that we believe to have some plausible independent motivations. 5. Conclusions In this paper we have argued that the haplology of -idad, -ism(o) and -í(a), when the word seems to be derived by -ist(a), is the kind of phenomenon that illustrates one of the predictions of a Phrasal Spell Out theory of Late Insertion. We have argued that given the functional hierarchy, the adjectival equivalent of a nominal structure involves a head A which intervenes between N and a higher projection F. If Phrasal Spell Out: An Argument from Haplology 37 the affix lexicalises a phrasal constituent including F and N, but not A, we predict that that affix will not be used when the structure appears in the adjectival version, but instead another exponent, which lexicalises F, A and N, will be introduced. An approach with Fusion does not make this prediction. Fusion, as an operation, would have to be redefined either to allow the simultaneous mapping of three or more heads, or to allow somehow that it makes reference to both morphological and syntactic terminals in order to make the same kind of prediction. Additionally, we have argued that the haplology of -is in front of -ist(a) is, against what has been assumed, not a phonological phenomenon, given that -is can be plausibly segmented as a noun marker. Globally, and beyond the specific problems discussed here, this article provides indirect evidence in favour of a neo-constructionist theory. First, the analyses that we have argued for in this article support a theory where two things are unnecessary: (a) an operation of haplology that would be applied at a morphological level; in our account, haplology is either phonological or motivated by independent syntactic principles having to do with the size of the complement of a lexical head or with the functional hierarchy; (b) idiosyncratic and arbitrary statements in specific lexical entries stating which other exponents undergo haplology in adjacency with that exponent. In this latter account, the fact that -ist(a) triggers haplology of precisely -ism(o), -idad and -í(a), but not -ur(a) or others, is purely accidental. It could have been these affixes as much as any others. In our account, the reason why some affixes undergo haplology, but not others, has to do with their internal complexity, and is matched by their semantics and affix-ordering phenomena. The generalisation that affixes that undergo haplology form a natural class can also be used to argue against an alternative set of operations that theories with Fusion, like Distributed Morphology, might propose to explain haplology. As an anonymous reviewer correctly points out, the same surface result could be obtained in Distributed Morphology through a combination of head movement and a statement that forces a(n intermediate) head to be spelled out as a default ø. Given (76), if for some reason the lowest head Z has to head-move to the highest head X, strict locality dictates that it must displace first to the intermediate head Y. 38 Antonio Fábregas (76)XP X YP Y ZP Z... Assume now that the morphology of the language has an idiosyncratic morphological rule—e.g., with the format that Bonet (1991) proposed for feature impoverishment in clitic sequences—that dictates that when Z and Y bundle together, Y has to be spelled out as ø. This would indeed capture haplology, but note that it requires positing an additional idiosyncratic operation which does not follow from head movement or any other property of the structure. Perhaps more importantly, as morphological rules in DM are idiosyncratic in nature, the rule could make reference to any set of affixes. It remains a mystery, under this account, why the affixes that haplologise are those, and only those, that contain an FP layer and a nominal head. Finally, another potential problem of this theory is that the reason why Z has to move to X is not clear; if it is a formal property that has to be checked by X, the question is what satisfies that property when X is not present (that is, when there is no haplology). Let us be clear at this point, however, that we do not mean to imply that standard DM cannot capture the effect of this kind of haplology: our critique is that it would miss the generalisations having to do with the feature composition of the affixes involved. Secondly, to the extent that we provide evidence for Phrasal Spell Out in lexical entries, we provide evidence for a system where exponents are introduced after the syntactic structure is complete, in line with the claims of the neo-constructionist approach. Insertion in terminal nodes is compatible both with a Late Insertion approach and with a lexicalist view where elements from the lexicon are present from the beginning of the syntactic derivation. However, if at least some exponents are associated with non terminal nodes, insertion of lexical items must necessarily happen after the structure is built, otherwise the context of insertion of exponents is not yet defined. Therefore, a way to interpret this article is also as an argument in favour of not allowing lexical information to be accessed until the computational system has built some structure. Phrasal Spell Out: An Argument from Haplology 39 Works Cited 1. Alber, Birgit & Sabine Arndt-Lappe. 2012. Templatic and substractive truncation. In Jochen Trommer (ed.), The morphology and phonology of exponence. Oxford: Oxford University Press, pp. 289-325. 2. Alemany Bolúfer, José. 1920. Tratado de la formación de palabras en la lengua castellana. Madrid: Librería General de Victoriano Suárez. 3. Arad, Maya. 2003. Locality constraints in the interpretation of roots: the case of Hebrew denominal verbs. Natural language and linguistic theory 21, pp. 737-778. 4. Aronoff, Mark. 1976. Word Formation in Generative Grammar, Cambridge (Mass.): MIT Press. 5. Arregi, Karlos & Andrew Nevins. 2012. Morphotactics: Basque auxiliaries and the structure of spell out. Dordrecht: Springer. 6. Baker, Mark. 2003. Lexical categories. Cambridge: Cambridge University Press. 7. ––. 2008. The Syntax of Agreement and Concord, Cambridge: Cambridge University Press. 8. Bermúdez-Otero, Ricardo. 2007. Morphological structure and phonological domains in Spanish denominal derivation. In F. Martínez-Gil & S. Colina (eds), Optimality-Theoretic Studies in Spanish Phonology, Amsterdam: John Benjamins, pp. 278-311. 9. Bonet, Eulàlia. 1991. Morphology after Syntax, Doctoral dissertation, Cambridge (Mass.): MIT. 10. Borer, Hagit. 2005a. In Name Only, Volume 1 from the ‘Exoskeletal Trilogy.’ Oxford: Oxford University Press. 11. ––. 2005b. The Natural Course of Events, Volume 2 from the ‘Exoskeletal Trilogy.’ Oxford: Oxford University Press. 12. ––. 2012. In the event of a nominal. In Martin Everaert, Marijana Marelj & Tal Siloni (eds.), The Theta System. Oxford: Oxford University Press, pp. 103-150. 13. ––. 2013. Structuring sense: taking form. Volume 3 from the ‘Exoskeletal Trilogy.’ Oxford, Oxford University Press. 14. Caha, Pavel. 2009. The nanosyntax of case. PhD dissertation, Tromsø: University of Tromsø. 15. Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36, pp. 315-332. 16. ––. 2010. The syntax of adjectives. Cambridge (Mass.): MIT Press. 40 Antonio Fábregas 17. Clark, Eve V. & Herbert H. Clark. 1979. When nouns surface as verbs. Language, 55, 767-811. 18. Corbin, Danielle & Marc Plénat. 1992. Note sur l’haplologie des mots construits. Langue Française 96, pp. 101-112. 19. De Lacy, Paul. 1999. Morphological haplology and correspondence. In Paul de Lacy & Anita Nowak (eds.), Papers from the 25th Anniversary. University of Massachusetts Occasional Papers. Amherst: GSLA, pp. 51-88. 20. Dressler, Wolfgang. 1977. Phono-morphological dissimilation. In Wolfgang Dressler & Oskar Pfeiffer (eds.), Phonologica 1976. Innsbruck: Universität Innsbruck, pp. 41-48. 21. Eguren, Luis. 2001. Evaluative suffixation in Spanish and the syntax of derivational processes. In Julia Herschensohn, Enrique Mallén & Karen Zagona (eds.), Features and Interfaces in Romance. 71-85. Amsterdam: John Benjamins. 22. Embick, David. 2004. On the structure of resultative participles in English. Linguistic Inquiry 35, pp. 355-392. 23. ––. 2010. Localism versus Globalism in Morphology and Phonology, Cambridge (Mass.): MIT Press. 24. Embick, David & Alec Marantz. 2008. Architecture and blocking. Linguistic Inquiry 39, 1-53. 25. Fábregas, Antonio. 2013. Diminutives as heads or specifiers: the mapping between syntax and phonology. Iberia, in press. 26.Fradin, Bernard, Fabio Montermini & Marc Plénat. 2007. Morphologie grammaticale et morphologie extragrammaticale. In Bernard Fradin, Françoise Kerleroux & Marc Plénat (eds.), Aperçus de morphologie du français. Saint-Denis: Puv, pp. 2145. 27. Grohmann, Kleanthes. 2011. Anti-locality: too-close relations in grammar. In Cedric Boeckx (ed.), The Oxford Handbook of Linguistic Minimalism. Oxford: Oxford University Press, pp. 260-290. 28. Halle, Morris & Alec Marantz. 1993. Distributed Morphology and the pieces of inflection. In K. Hale and S. J. Keyser (eds), The View from Building 20, Cambridge (Mass.): MIT Press, pp. 111-176. 29. Harris, James W. 1991. The Exponence of Gender in Spanish. Linguistic Inquiry, 22, 27-62. 30. Johnston, Michael & Federica Busa. 1999. Qualia structure and the compositional interpretation of compounds. In E. Viegas (ed.), Breadth and depth of semantic lexicons, 167-188. Dordrecht: Kluwer. Phrasal Spell Out: An Argument from Haplology 41 31.Julien, Marit. 2002. Syntactic Heads and Word Formation, Oxford: Oxford University Press. 32. Lin, T.-H. Jonah & C.-Y. Cecilia Liu. 2005. Coercion, event structure and syntax. Nanzan Linguistics 2, pp. 9-31. 33. Marantz, Alec. 1997. No escape from syntax: don’t try morphological analysis in the privacy of your own lexicon. In A. Dimitriadis, L. Siegel et al. (eds), UPenn Working Papers in Linguistics 4, Pennsylvania: University of Pennsylvania, pp. 201-225. 34. ––. 2001. Words, Unpublished ms., Cambridge (Mass.), MIT. 35. ––. 2013. Verbal argument structure: events and participants. Lingua 130, pp. 152-168. 36.McBride, Alexander I. 2004. A constraint-based approach to morphology. PhD dissertation, Los Angeles: University of California. 37. Neeleman, Ad & Kriszta Szendrői. 2007. Radical pro drop and the morphology of pronouns. Linguistic Inquiry 38, pp. 671-714. 38. Neeleman, Ad & Hans van de Koot. 2006. Syntactic haplology. In Martin Everaert, Rob Goedemans, Bart Hollebrandse & Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax. Oxford: Blackwell, pp. 685-710. 39. Nevins, Andrew. 2012. Haplological dissimilation at distinct stages of exponence. In Jochen Trommer (ed.), The morphology and phonology of exponence. Oxford: Oxford University Press, pp. 84-117. 40. Noyer, Rolf. 1997. Features, Positions and Affixes in Autonomous Morphological Structure. New York: Garland Publishing. 41. Oltra-Massuet, Isabel. 2010. On the morphology of complex adjectives. PhD dissertation, Barcelona: Universitat Autònoma de Barcelona. 42. Ortmann, Albert & Alexandra Popescu. 2001. Haplology involving morphologically bound and free elements: evidence from Romanian. In Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology 2000. Dordrecht: Kluwer, pp. 43-70. 43. Panagiotidis, E. Phoevos. 2011. Categorial features and categorizers. The Linguistic Review 28, pp. 365-386. 44. Piñeros, Carlos Eduardo. 2000. Prosodic and segmental unmarkedness in Spanish truncation. Linguistics 38, p. 63-98. 45. Plénat, Marc. 1991. Le javanais: concurrence et haplologie. Languages 101, pp. 95-117. 46. Pustejovsky, James. 1995. The Generative Lexicon. Cambridge (Mass.): MIT Press. 42 Antonio Fábregas 47.RAE [Real Academia Española] & ASALE [Asociación de Academias de la Lengua Española]. 2009. Nueva gramática de la lengua española. Madrid: Espasa. 48.Ramchand, Gillian. 2008. First Phase Syntax. Cambridge: Cambridge University Press. 49. Richards, Norvin. 2010. Uttering trees. Cambridge (Mass.): MIT Press. 50. Roca, Iggy & Elena Felíu. 2003. Morphology in truncation: the role of the Spanish desinence. In Geert Booij & Jaap van Marle (eds), Yearbok of Morphology 2002, Dordrecht: Kluwer, pp. 187-243. 51. Scheer, Tobias. 2011. A guide to Morphosyntax-Phonology Interface theories. How extramorphological information is treated in Phonology since Trubetzkoy’s Grenzsignale. Berlin: Mouton de Gruyter. 52. Sproat, Richard & Chilin Shih. 1991. The cross-linguistic distribution of adjective ordering restrictions. In Carol Georgopoulos & Roberta Ishihara (eds.), Interdisciplinary approaches to language. Dordrecht: Kluwer, pp. 565-593. 53. Stemberger, Joseph P. 1981. Morphological haplology. Language 57, pp. 791-817. 54. Stump, Gregory. 1998. Inflection. In Andrew Spencer and Arnold M. Zwicky (eds), The Handbook of Morphology, Oxford: Blackwell, pp. 13-43. 55. Svenonius, Peter, Gillian Ramchand, Michal Starke & Tarald Taraldsen (eds.). 2009. Nordlyd 36.1. Special issue on Nanosyntax. Tromsø: Septentrio Press. 56. Van Riemsdijk, Henk. 2008. Identity avoidance. In Robert Freidin, Carlos P. Otero & María Luisa Zubizarreta (eds.), Foundational issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. Cambridge (Mass.): MIT Press, pp. 227-250. 57. Volpe, Marc J. 2005. Japanese morphology and its theoretical consequences: derivational morphology in Distributed Morphology. PhD dissertation: Stony Brook University. 58. Weerman, Fred & Jacqueline Evers-Vermeul. 2002. Pronouns and case. Lingua 112, pp. 301-338. 59. Wurzel, Wolfgang Ulrich. 1976. Zur Haplologie. Linguistische Berichte 41, pp. 50-57. 60. Xu, Zheng. 2007. Inflectional morphology in Optimality Theory. PhD dissertation: Stony Brook University.
© Copyright 2026 Paperzz