Abstract

Taking the Distributed Morphology model as a starting point, this article presents and develops the hypothesis that parallel computations drive some word formation processes. Along the way, some Distributed Morphology assumptions, particularly those concerning contextual allomorphy, are revised. It is argued that event structure is a syntactic head independent of the presence of a vP. Nominalizations in Spanish, which often exhibit verbal thematic vowels between the root and the nominalizing affix, turn out to be an ideal testing ground for theoretical hypotheses.

1 Introduction

For most of its history, generative grammar has adopted the Lexicalist Hypothesis, according to which word formation takes place in a dedicated module separate from syntax, sometimes called the Lexicon, while syntax takes care exclusively of building phrases out of words. Within Lexicalism, morphology and syntax interface only at the X0 node, the endpoint of morphology and the starting point of syntax (see Scalise and Guevara 2005 for an overview of Lexicalism). Alternatively, over the last 20 years various authors have proposed that all combinatorial operations take place in one module (see Lieber 1992, Halle and Marantz 1993, Marantz 1997, Alexiadou 2001, among many others). In this article, I will explore in detail some consequences of the hypothesis that there is only one computational system. This exploration will lead me to revise some mainstream analyses and assumptions and to propose alternatives. The end result will be, or so I hope, a more coherent theory of the computational system, with some empirical advantages. For the purposes of the analyses, I adopt (and eventually revisit) the assumptions that constitute the research program referred to as Distributed Morphology (DM), initiated by Halle and Marantz (1993) and developed in (among many other works) Embick and Noyer 2001 and Embick 2010.

As a starting point, let us assume that the computational system of human language (CHL) is derivational. The empirical fact that specifiers (Specs) and adjuncts exist leads to the conclusion that CHL can engage in parallel computations (henceforth PCs) (Chomsky 1995, Uriagereka 1999). For instance, (1) exemplifies the merger of an external argument with a verb (under simple structural assumptions to be revisited shortly). The derivation of The man of means look(ed) at the painting involves two parallel derivations, one that yields look at the painting and another that yields the man of means. After both derivations are finished, they are merged.

(1)

graphic

Let us further assume that head movement exists—in other words, that two heads create a complex head in the syntax by means of Internal Merge (IM).1 Therefore, two types of projections may exist: a phrasal projection (2a) and a head projection (2b).

(2)

graphic

Let us take it for granted, then, that CHL includes PCs and complex heads. If we additionally adopt the DM postulate that there is no special computational system to build words out of morphemes, we are led to the following two conclusions:

(3)

  • a.

    One should expect to find PC in word formation.

  • b.

    One should expect to find complex heads formed by External Merge (EM).

It seems to me the following sort of example is evidence for (3a) and (3b):

(4)

  • a.

    God-is-dead theology

  • b.

    over-the-fence gossip

  • c.

    don’t-tell-me-what-to-do look

    (Sato 2010:380)

In the Germanic languages and Chinese, the first component of a compound noun can be an entire phrase or even an entire sentence. These compounds are not formulaic, as sometimes claimed; rather, they result from a perfectly productive word formation mechanism (for discussion, see Bauer 1983, Lieber 1992, Harley 2009b, Sato 2010). I take these compounds to show (a) that there is PC in word formation because their first component is built in parallel with the main derivation, and (b) that there are complex heads built from EM.

In DM, complex heads are created as the result of IM, not EM. DM generally adopts only a head-complement structure for word formation, expressed as in (5a) (Marantz 1997). Nominalizations are a type of word formation that follows the pattern in (5a), as shown in (5b). This sort of analysis of nominalization has gained popularity in recent years (see, e.g., Alexiadou 2001, 2009, Roeper 2005, Harley 2009b, Borer 2013).

(5)

  • a.

    [xP x z]

    [yP y [xP x z]]

  • b.

    √Root] v] n

Spanish nominalizations, which include a verbal morpheme usually referred to as the thematic vowel, would seem to support this approach, as shown in (6a) and (6b).

(6)

  • a.

    nominalización

  • b.

    nominaliz – a – ción

    nominaliz] TV] n

However, Spanish nominalizations do not exhibit any of the morphosyntactic properties of verb phrases, as I will show in section 3.2, strongly suggesting that (6a) should be derived in a manner different from (5b).

In this article, I explore a DM model that incorporates PC and EM-derived complex heads. The hypothesis is simple but with wide-ranging consequences. In addition to head-complement structures like those exemplified in (5a), complex words can be built by merging a stem with the output of a parallel derivation that involves affixes. This is shown schematically in (7).

(7)

graphic

(7a) and (7b) are parallel derivations. In (7b), a head w takes the phrase uP as a complement, as usual. In (7a), the affixes x, y, and z undergo EM successively, forming a complex head. In (7c), this complex head takes wP as a complement. I will argue that this sort of approach provides the key to the structure of Spanish nominalizations as well as to some other morphological phenomena.

Within DM, the closest proposal to this that I am aware of is Embick’s (2004) formal approach to conflation (see also, e.g., Mateu 2002, 2008, Acedo-Matellán 2013). Embick (2004) proposes an analysis of resultative predicates like (8a) in which the little v is merged with a root and the resulting compound takes the resultative predicate as a complement, as in (8b).

(8)

  • a.

    John hammered the metal flat.

  • b.

    John [vP [v v_√hammer] [aP the metal flat]]

I would suggest that conflation is a particular case of what is in fact a widespread phenomenon.

The article is organized as follows. Section 2 introduces some key features of DM. Sections 3–7 constitute the formal presentation of the proposal. Section 3 presents a description of nominalizations with a special focus on Spanish; the role that thematic vowels play in derivational morphology in this language and the empirical problems that they raise for DM are discussed in detail. Section 4 presents the analysis of Spanish nominalizations. Section 5 discusses the argument structure of nominals in light of the previous discussion. Section 6 briefly discusses compounding. Section 7 discusses what I call ‘‘Ackema and Neeleman’s objection,’’ a false prediction made by DM that is obviated by the present proposal. The article concludes in section 8 with a summary.

2 Some Features of Distributed Morphology

In this section, I briefly summarize the features of DM that will become relevant in the analyses (for more details, see Marantz 1997, 2001, Embick 2010). I take it for granted that there is no special module for word syntax, separate from phrasal syntax. In place of the traditional lexicon we have a list of roots and functional categories without phonological matrices (so-called List 1). Phonological matrices are stored in a separate list (List 2) and are assigned to terminals in a postsyntactic module by means of an operation of Vocabulary Insertion (VI).

For present purposes, I will adopt the assumption that roots are slices of conceptual structure devoid of grammatical properties. The nature of roots is currently the subject of intense debate (see, e.g., Acquaviva 2009, Harley 2011, Borer 2013), but the arguments made here do not depend on the outcome of this debate. I will also assume (until section 5, when I will need to be more precise) that any internal argument visible in the clause is introduced into the derivation via EM as a complement of a root.

Functional categories consist of feature bundles that set up grammatical dependencies and drive the computation. A subset of the functional categories—identifiable as derivational morphemes— categorize the structure being built as adjectival, nominal, verbal, or prepositional.

(9)

graphic

Categorizing heads also constitute phase heads. Phase heads trigger Transfer of their complements to the interpretive modules. Under Embick’s (2010) condition that only a phase can be transferred (and not a root or a root phrase), the computational cycle works as follows:

(10)

graphic

Consider (10a1) and (10a2). Since the root is transferred at the point where y is introduced, there cannot be any connection between y and the root. In particular, the root cannot trigger any allomorphy or allosemy on y. The root can trigger these phenomena only on x (Marantz 1997, 2001). Take example (11), socialization. The exponent of the final n must be [ation] because that is what [ize] demands. The root √soci has no say in the spell-out of n because it has been transferred before n is merged. As a consequence, although an n merged with the root √soci spells out as [iety], the same n merged with [socialize] cannot spell out as [iety] because at this point the root has no say in the spell-out of n. Thus, society is grammatical, *socializeity is not.

(11)

graphic

Now consider (10b). Z and W transfer together with the root. Therefore, they are subject to allomorphy and allosemy triggered by the root. English T exemplifies this situation.

(12) √] v] T

English T is transferred with v and the root. Therefore, the root can determine which allomorph of T ends up being chosen. If the root is √pit, the past tense is pitted; if the root is √hit, the past tense is hit; and so on.

Embick (2010), and DM in general, add one more condition to trigger allomorphy: the head that triggers allomorphy and the head that is the target of allomorphy must be adjacent. For the purpose of calculating adjacency, silent heads are not counted. Thus, since v in English is always silent, T and the root count as adjacent.

A structure transferred to PF may include complex heads, as in (13).

(13)

graphic

z, x1, and y1 are subwords of the complex word y. z, x1, and y1 are also terminals, which are subject to the rules of VI. The rules of VI consist of feature matrices and contexts of insertion. The structure of a VI rule is as follows:

(14) VI: [y] ↔ /a/ // ___ C

That is, the phonological matrix /a/ is inserted in the position of the syntactic terminal [y] in the context C. (For convenience, I routinely write orthographical representations in / . . . / instead of IPA representations, and I maintain square brackets [ . . . ] for morphemic representation before Spell-Out.) As mentioned, the context for VI rules has been defined in terms of linear adjacency in recent developments of DM. I hope to show that defining VI rules in terms of structure provides some advantages.

Within DM, I have found no discussion of how the creation of complex heads in the syntax affects cyclic Transfer. But such discussion is necessary. Take a simple structure like (15).

(15)

graphic

If the root incorporates into x, and x incorporates into y, we obtain a structure like (16).

(16)

graphic

Here, both x and the root are adjoined to y, and the Transfer triggered by y affects xP but not x or the root. In fact, y, x, and the root will transfer together at the next cycle and therefore there should be no obstacle to allomorphy and allosemy on y triggered by the root. This situation is not hypothetical: numerous languages have verb-second effects in the syntax that exemplify (16). It is unclear how cyclic Transfer is supposed to work out when complex heads are formed in the syntax.

3 Nominalizations

This section presents a description of nominalizations in English and Spanish and the empirical questions raised by the latter. Nominalizations in Spanish are of particular interest because intermediate derivational stages are more transparent in this language than in English, thanks to the ubiquity of thematic vowels. Thematic vowels are obligatory components in the morphological structure of Spanish verbs and allow us to classify verbs into three conjugation classes. But thematic vowels also show up in deverbal nouns and adjectives, with interesting consequences for morphological theory.

As I will show, the presence of thematic vowels in Spanish nominalizations, combined with the absence of any other properties that would suggest the presence of a full-fledged verbal phrase, argue for introducing some heads by means of PC.

3.1 Nominalizations in English

Following the path traced by Abney (1987) and many others after him, I take it that the gerundive nominal is an n head that takes a vP as a complement.2 (In tree representations, I write the spelled-out Vocabulary items associated with the syntactic terminals for the reader’s convenience. But recall that I subscribe to the DM hypothesis that the units of morphosyntactic computation are feature bundles that receive a phonological matrix after Spell-Out.)

(17) John’s watching the vase

(18)

graphic

(19)

  • shows the VI rule for the gerundive nominal.

    • (19) VI: [n] ↔ /ing/ // ___ vP

In mainstream DM, (19) reads, ‘‘The phonetic matrix /ing/ is inserted under a node n when n is adjacent to a vP.’’

This analysis, which includes a full vP with a v as head of a phase, derives the two most salient features of gerundive nominals: (a) accusative case on the complement of the root and (b) the possibility of low adverbs. Both features are exemplified in (20).

(20) John’s intently watching the vase (made the owner suspicious).

Since the morphology of modifiers and the assignment of structural case to internal arguments play important roles in this article, let me state my assumptions explicitly. I start with modifiers:

(21)

  • Given a phase p with a head H(p), H(p) determines the morphology of any modifiers merged in p.3

    • If n(p), modifiers will take on adjectival morphology.

    • If v(p), modifiers will take on adverbial morphology.

Thus, modifiers merged in a vP phase will adopt adverbial morphology, while modifiers merged in an nP phase will spell out as adjectives. There are many types of modifiers not covered by (21), but (21) suffices for our purposes.

The case morphology on an internal argument is also defined in terms of the phase of which it is a constituent.

(22)

  • Given a phase p with a head H(p), H(p) determines the morphological case of an argument DP merged in p.

    • If n(p), case morphology is genitive (of in English, de in Spanish, etc.).

    • If v(p), case morphology is accusative.

Gerundive nominals include a vP, as shown in (18). According to (21), they will accept adverbs; and according to (22), they will assign accusative case.

The assumptions in (22) gloss over all the interesting complications of lexical case, the possible presence of applicatives, and so on. Again, none of this will be relevant in this article. It is worth pointing out that (21) and (22) are simply subcases of the theory of phases and are not to be considered new assumptions.

Let us now move beyond gerundive nominals to consider other nominalizers ([-tion], [-ance], etc.). For the present discussion, we can assume that these English nominalizers attach directly to the root (as in Embick 2010:47), yielding structures like (23).

(23)

graphic

The actual spell-out of n is defined by means of the following VI rules:

(24)

graphic

Since structure (23) contains no verbal phrase, it is expected that there is no accusative case to be assigned and that modifiers will be adjectival. Both expectations hold.

(25)

  • a.

    John’s careless breaking of the vase was annoying.

  • b.

    *John’s carelessly breaking of the vase was annoying.

  • c.

    John’s breaking the vase was annoying is grammatical as a gerundive nominal.

(26)

  • a.

    Mary’s brilliant resolution of the problem was amazing.

  • b.

    *Mary’s brilliantly resolution of the problem was amazing.

  • c.

    *Mary’s resolution the problem was amazing.

(27)

  • a.

    Pat’s stubborn resistance to authority worries his parents.

  • b.

    *Pat’s stubbornly resistance to authority worries his parents.

  • c.

    *Pat’s resistance authority worries his parents.

However, some nominalizations in English do include the overt verbal affix [ize]. Embick (2010) argues that the structure of these words should include a vP.

(28)

  • a.

    itemization, legalization, vaporization, . . .

  • b.

    item] ize vP ] ation nP ]

If itemization does include the verbal phrase itemize, one should expect to find accusative case and adverbs. However, this expectation is not fulfilled: in every respect, the nominalizations in (28) behave like the nominalizations in (25)–(27).

(29)

  • a.

    *his carefully itemization

  • b.

    his careful itemization

(30)

  • a.

    *his itemization newly arrived packages

  • b.

    his itemization of newly arrived packages

Some linguists have argued that the structure of all nominalizations in English includes a vP and, fully cognizant of the data I have just presented, have tried to make sense of them. This issue is discussed in section 4.4.

3.2 Nominalizations in Spanish

I start this section by introducing my assumptions concerning thematic vowels. I then develop a description of nominalizations in Spanish, highlighting the empirical problems they present for mainstream DM.

Consider (31a). The thematic vowel (TV) a is added to the root √cant, which yields a verb of the first conjugation class. The other conjugation classes are defined by the thematic vowels e and i.

(31)

graphic

Thematic vowels have been argued by Oltra-Massuet (1999) and Oltra-Massuet and Arregi (2005) to be associated with little v, and I will follow them on this.4 Since the presence of a thematic vowel on a root entails the presence of a little v, I adopt the simplifying assumption that the thematic vowel itself is a spell-out of little v (nothing hinges on this simplification).

A complete list of VI rules that spell out the thematic vowels would be a complex affair, because we would have to take into consideration the subjunctive mood as well as several cases of syncretism and other complications. For present purposes, we can simply assume the following general rules, which account for most forms of the indicative conjugations:

(32)

graphic

These entries read, ‘‘Spell out v as /e/ when v is attached to a root from the list provided,’’ and so on. The spell-out of /a/ does not have a list of roots, which correctly predicts that /a/ is the only productive thematic vowel in contemporary Spanish. Oltra-Massuet (1999) and Oltra-Massuet and Arregi (2005), as well as Pomino (2008) and Embick (2010), assume that the conjugation class that defines the thematic vowel is a feature of the root. Following these authors, we would have √cant[I], √beb[II], √viv[III], and so on. However, Acquaviva (2009) argues against this solution. Having conjugation features attached to categoriless roots seems paradoxical, since belonging to a conjugation class entails being a verb. Moreover, one of the fundamental tenets of DM is that roots have no grammatical properties, which should lead to the conclusion that roots have no conjugation class features. My assumptions seem conceptually superior: roots remain innocent; the thematic vowel is simply an expression of v; and the spell-out of v is determined by the VI rules, as is standard in DM.5

The following is a short list of nominalizations in Spanish:

(33)

graphic

In many nominalizations (and adjectivalizations, which I do not discuss here), the thematic vowel remains in the final output, as the following examples show:

(34)

graphic

(35)

graphic

(36)

graphic

Here is a list of nominalizations without a thematic vowel:

(37)

graphic

The examples in (34) are nominalizations in [dura], which preserve the thematic vowel. For instance, the verb tachar with thematic vowel [a] yields the nominalization tachadura, while the verb morder with thematic vowel [e] yields mordedura. In the examples in (35) and (36), headed by the affixes /m(i)ento/ and /cion/, respectively, the second and third conjugation thematic vowels conflate into /i/ (which also happens in several tenses of the verbal paradigm). The examples in (37) have no thematic vowel; the forms *insertación and *movición do not exist (but movimiento ‘movement’ does).

The presence of the thematic vowel leads to the conclusion that the structure of the nominalizations in (34)–(36) includes a verbal morpheme, which, within my theoretical assumptions, entails that it also includes a little v. Further evidence that these nominalizations involve some verbal structure comes from parasynthetic verbs. In Spanish, some verbs derived from adjectives or nouns include a prefix, usually [a] or [en].

(38)

graphic

Nominalizations with [dura], [m(i)ento], and [cion] based on verbs of this type always include a prefix.

(39)

graphic

Thus, there is no doubt that the nominalizations in (34)–(36) include a verbal derivation. Within standard DM assumptions, we would have to assume that their structure is [[√] v] n. But this analysis gives rise to two empirical problems.

The first empirical problem is a familiar one: the nominalizations in (34)–(36) should license accusative case and adverbs (see (21) and (22)), but possibly not the ones in (37), where the absence of a thematic vowel suggests that there is no vP (see (23)). However, all of the nominalizations in (34)–(37) behave as if there were no verbal phrase in the structure: they do not admit adverbs, and there is no accusative case. Interestingly, nominalizations whose structure includes a thematic vowel behave like nominalizations that take the root as stem.

(40)

  • a.

    Juan definió el problema rápidamente.

    Juan defined the problem quickly

  • b.

    *la definición el problema rápidamente

    the definition the problem quickly

  • c.

    la definición rápida del problema

    the definition quick of.the problem

    ‘the quick definition of the problem’

The second empirical problem involves the locality of allomorphy and allosemy. Recall that within Embick’s (2010) framework, a morpheme x can trigger allomorphy on a morpheme y only if x and y are transferred within the same cycle and are linearly adjacent. Both conditions are falsified by Spanish nominalizations. The spell-out of n as /dura/, /m(i)ento/, or /cion/ is decided by the root: tachadura (not *tachación), separación (not *separadura). But within DM assumptions, the root should be transferred in the same cycle that introduces v, leaving n to be transferred in the next cycle, and therefore we should not expect the root to be able to choose the n allomorph. Moreover, v is not silent and therefore the root and n are not adjacent. Within the standard DM assumptions sketched in section 2, one should conclude that the root has no influence on the spell-out of n in Spanish nominalizations.

The connection between the root and the nominalizing morpheme involves allosemy too. The root √salv can be related to ‘rescuing’ (what lifeguards do at the beach) but also to ‘saving’ in a spiritual way (what a church does for one’s soul). Both meanings are maintained in the verb salvar. But each alloseme of the root √salv has its own nominalization: salvamento means ‘rescue’, salvación means ‘salvation’. Since the source of the allosemy is the root, the choice of [mento] or [cion] must refer back to it.

It seems fair to conclude that Spanish nominalizations present empirical obstacles for the mainstream DM approach.

4 Nominalizations and Parallel Computation

In section 4.1, I present an analysis of Spanish nominalizations that utilizes PC as introduced in section 1. In section 4.2, I discuss allomorphy, and in section 4.3, I briefly discuss the role of the notion ‘‘word’’ in DM. In section 4.4, I revisit the issue of case and adjunct morphology in nominalizations. The analysis I present also has consequences for the theory of argument structure in nominals and the role of little v in the grammar, as I will discuss in section 5.

4.1 Analysis

The absence of adverbs and accusative case in Spanish nominalizations does not present a problem within a framework that acknowledges the possibility of PC in word formation. Let us assume that the category v does not merge with the root or project a phrase; rather, let us merge it with the n in a PC, forming a complex n head. The complex [n n [v]] merges with the root, projecting an nP. For example:

(41)

  • definición (d)el problema

  • definition (of ).the problem

(42)

graphic

With the nP thus formed, the root √defin can incorporate into n, yielding the complex word in (43)

(43)

graphic

The VI rules listed in (32) suffice to give the form [defini] for the stem √defin+v. But how will the spell-out for n be selected? Recall that it depends on the root. An adjacency-based VI rule cannot operate here. Instead, I suggest a VI rule that takes advantage of the fact that the root and n are constituents of the same complex word.

(44) VI: [n] ↔ /cion/ // [n √ __ ] List of √: { . . . }

This VI rule says that [cion] spells out an [n] terminal if there is a root from the relevant list within the complex word n. The result is that the choice of phonological matrix for [n] depends on the root, as desired.

For the sake of consistency, in (45) I revisit the thematic vowel VI rule, so that it becomes sensitive to structure and not to adjacency.

(45)

graphic

That is, /e/ spells out v when v is in the same complex head as a root from the list; and so on.

Let us return to the empirical problem posed by the morphology of modifiers and case. Recall that the problem is that, if nominalizations include a vP, as often argued, then they should accept adverbs and accusative case. In the present proposal, the head of the phase is an n, v is embedded within n, and v does not project a phrase and is not the head of a phase. Since the head of the phase is n, the morphological case assigned to el problema is genitive, spelled out as de. Any modifiers within the domain of n spell out as adjectives.

Let’s return to English. Without any amendments, the analysis just presented can be used to analyze itemization: specifically, [ization] builds up in a PC and the output is merged to the root √item. The head of the resulting phase is n, and therefore we expect itemization to occur with adjectives and genitive case.

4.2 Allomorphy

I have eschewed both the adjacency requirement and the same-phase requirement for allomorphy. Instead, I propose VI rules in which allomorphy is determined by coconstituents of a complex word (for an independent argument in favor of structural VI rules, see Chung 2007 and the discussion of Chung’s findings in Bonet and Harbour 2012). These changes are required to account for Spanish nominalizations. It is necessary to rebuild the locality constraint of the theory of VI and allomorphy in a DM framework.

Building on rules (44) and (45), I would like to suggest the following principle of allomorphy:

(46) Given x and y, constituents of a complex word W, x and y are free to trigger allomorphy on each other.

(46) seems too open and suggests that anything goes. In fact, not everything goes. Take a structure like the one in (43). We want the choice of root to influence the choice of n, but, under the assumption of VI cyclicity (Bobaljik 2000, Embick 2010), we do not want n to choose the root (n could, on the other hand, trigger readjustment rules on the root; see, e.g., Embick 2010). Still assuming a structure like (43), I suggest the following:

(47) Within a word W, VI applies to the terminals from the top down.

That is, in a structure like (48), the choice of Vocabulary item for terminal y will be resolved after terminal x is resolved, and terminal z will be resolved after terminal y. However, (47) is still too unrestricted. In (48), it could be the case that both x and y can trigger allomorphy on z (not a hypothetical case, as I will show). In this configuration, y wins over x, but this is not reflected in (46) or (47).

(48)

graphic

The history of generative grammar suggests a path inspired by relativized minimality: the phonological matrix that z takes depends on the hierarchically closest allomorphy trigger within W. Thus, if y is not an allomorphy trigger for z, x is free to act. But if y is an allomorphy trigger for z, y applies and x is stymied. This can be implemented in the following manner. Let us assume that there are two VI rules that spell out z. Take z to be a probe in search of a context that will match a VI rule within the complex head. Following the general computational principle of minimal search, z will look for the closest possible context. The closest context it will find will be y, its sister. If there is a VI rule for z that makes reference to y, then it applies. If there is no VI rule for z that makes reference to y, then z probes x. Take a structure such as (43), with a root, a thematic vowel (v), and a nominalizing suffix (n). The n looks for a VI that makes reference to its sister, v. Since it finds none, it probes up the tree and finds the root √defin. n then looks for a VI rule that makes reference to √defin and finds one. The phonological matrix /cion/ is inserted in n. If there is no contextual allomorphy, z will seek a VI rule without context (an elsewhere rule). If there is no VI rule at all, the morpheme is not spelled out. This algorithm yields the possibility that a root may trigger allomorphy on n across a thematic vowel or any other morpheme—the only condition is that the intermediate morpheme does not trigger allomorphy on n.

Let me flesh this out with a more complex example. In Spanish, the causative affix [iz] only admits one nominalizing morpheme: [cion]. ([iz] is a cognate of English [ize]; see (11).)

(49)

graphic

Notice that [iz] has the thematic vowel [a], so again the morpheme that triggers allomorphy is not adjacent to the allomorphy target. Therefore, we will need a structural condition on allomorphy so that [iz] triggers the spell-out of n as [cion].6

(50) nominal a] iz] a v] ción n]

[iz] allows us to perform a simple experiment: we embed below [iz] a root that triggers a different instantiation of n. We find that [iz] always wins. In (51a), the root √mov triggers a nominalization in [miento]. However, the complex word movilizar triggers a nominalization in [cion] because this is what the suffix [iz(a)] requires and [iz(a)] is hierarchically closer to n than the root is. In (51b), the complex word centralizar also triggers a nominalization in [cion], although the root √centr can be nominalized as centro.

(51)

graphic

I surmise that [iz] wins over the embedded root precisely because of its hierarchical position. There are two VI rules that affect the spell-out of n, but /iz/ is probed before the root and so the /iz/ VI rule applies instead of the root one.

(52)

graphic

This shows that the ‘‘relativized minimality’’ approach to the locality of allomorphy accounts for the same range of phenomena that the phase+adjacency system is meant to account for but without the latter’s empirical difficulties.

The most elaborate analysis of allomorphy in terms of adjacency within DM that I am aware of is Embick’s (2010) proposal for the agreement morphemes in the Latin present perfect. The Latin present perfect breaks up the regularity on display in all the other tenses, as exemplified in (53) with the second person singular.

(53)

graphic

As these examples show, the second person singular form of the verb is always /s/, except in the present perfect, where it is /istis/. Embick argues that the allomorph /istis/ is triggered by the perfect morpheme /v/, but only when /v/ and [2sg] are adjacent. In the present perfect, the tense morpheme is silent and so /v/ and [2sg] are adjacent. In the other perfect tenses, there are intervening tense and/or mood morphemes that break up the required adjacency.

However, the data are amenable to an alternative analysis. The absence of a spell-out for present tense in the perfect could be due to fusion of the T and Asp nodes, as in (54).

(54)

graphic

If T does not fuse with Asp (in the past perfect, perfect subjunctive, etc.), the structure in (55) results.

(55)

graphic

If this analysis is correct, the environment for [2sg] differs from the environment for [2sg] in the rest of the conjugation because the present perfect contributes a label to the complex word that includes [2sg] whereas in any other context the label of the complex word is T.

I propose two VI rules for [2sg]. The first is the general one: /s/ is the spell-out for [2sg] whenever [2sg] is in a complex word with label T.

(56) VI: [2sg] ↔ /s/ // [T ― ]

The second, more specific VI rule says that [2sg] spells out as /istis/ when it is a constituent of a word with the labels Perfect and Present.

(57) VI: [2sg] ↔ /istis/ // [Perf/Pres ― ]

Thus, in context (54), the more specific rule (57) wins out over the more general rule (56). In other contexts, (56) applies.

4.3 Morpheme and Word

In a critique of Embick and Noyer 2001, Williams (2007) argues that DM claims to have abandoned the word as a linguistic primitive only to reintroduce it through the back door. In particular, Williams’s (2007) criticism focuses on Embick and Noyer’s (2001, 2007) assumption of a complex word that is the unit for some PF processes such as lowering and local dislocation. My analysis in this section would superficially seem to be subject to the same criticism, since I am taking the word to be the unit within which the VI rules operate (Victor Acedo-Matellán, pers. comm.). Thus, it is important that I make clear what it is that I am claiming.

I claim, following DM tenets, that the word is not a unit of computation for phrasal syntax as distinct from word syntax: there is only one CHL, and the input to CHL is the morpheme. There is no Merge operation that builds words distinct from a Merge operation that builds phrases: there is only one Merge. The output of applying Merge is words and phrases. Both words and phrases can be used as input for PF processes: for instance, they are both relevant units for calculating stress and drawing prosodic boundaries.7 I further argue in this article that the word is also the unit of interest for VI, since it is within the word that allomorphy is defined. Thus, this article is not meant to be read as a rejection of the word as a meaningful concept in linguistic analysis; rather, the word is positioned in the area of the theory where it is most useful—that is, PF. Note also that the assumption that the unit of computation is not the same unit that is used in PF processes does not complicate the theory of grammar: this is the situation that we had before DM, because the ‘‘word’’ of syntax never denoted the same thing as the ‘‘word’’ of phonology and, as Marantz (1997) points out, the ‘‘word’’ of phonology is not coextensive with the ‘‘lexical item’’ of Lexicalist theories.

Morphemes need to be spelled out, and this spell-out is determined by a local context, as the empirical evidence makes clear. My discussion of Spanish nominalizations shows that looking for this local context within phase theory and adjacency, as has been attempted within DM, does not work. Instead, the empirical evidence suggests that the complex word is the unit for VI. Since a new unit of analysis is not being introduced (the word is still needed for stress and prosody, as I mentioned), the result is, in my view, close to optimal. Finally, I would like to point out that in the DM model developed by Embick (2010) and Embick and Noyer (2001), among others, the word is the correct unit for computing some PF processes, such as local dislocation and stress assignment, while the phase is the unit for VI, another PF process. Within my model, the word is uniformly the unit for everything that goes on at PF.

4.4 Adverbs and Case in Nominalizations

Linguists who argue that English nominalizations include a vP structure are aware that nominalizations (except gerundive nominals) do not allow accusative case or adverbial morphology. They have tried two approaches to the problem. The first is to argue that the verbal structure in nominalizations is special in some way that prevents it from displaying the full panoply of verbal properties. The second is to deny that a problem exists and to argue that nominalizations do indeed license adverbs (no mention of case is made within this approach).

As a prominent example of the first approach, Alexiadou (2001) argues that the little v found in nominalizations is always of the unaccusative type. Since unaccusative v does not license accusative case, the case problem is resolved. However, notice that this does not resolve the adverb problem, because adverbs—including low manner adverbs—are licensed in unaccusative structures. Absolute small clauses are examples of unaccusative clauses without T or any functional categories above T, since they consist only of a participle and an optional internal argument (see Pérez-Jiménez 2006 for a detailed analysis of this construction). The following are some examples; they show that absolute small clauses license adverbs.

(58)

  • a.

    Desaparecido misteriosamente, el diamante alimentó la imaginación

    disappear.PTC mysteriously the diamond fed the imagination

    de generaciones de aficionados entusiastas.

    of generations of amateurs enthusiastic

    ‘Having mysteriously disappeared, the diamond fed the imagination of generations of enthusiastic amateurs.’

  • b.

    Disuelto totalmente, el insecticida dejó de surtir effecto.

    dissolve.PTC completely the insecticide stopped of supply effect

    ‘Since it dissolved completely, the insecticide stopped having an effect.’

  • c.

    Mecida la cunita suavemente, el niño al final se durmió.

    rock.PTC the cradle softly the child at.the end SE sleep

    ‘After softly rocking the cradle, the child finally went to sleep.’

    (Pérez-Jiménez 2006:124)

Alexiadou’s argument makes the wrong prediction because if an unaccusative structure cannot license an adverb in nominalizations, then it should not be able to do so in absolute small clauses either.

Additionally, a closer look at Alexiadou’s analysis yields awkward consequences. Consider the following examples:

(59)

  • a.

    Marijuana is legal in California.

  • b.

    Mary legalized marijuana in California.

  • c.

    Mary’s legalization of marijuana in California (was controversial).

The verbal structure that spells out as [ize] is of the transitive type, as shown in the contrast between (59a) and (59b). There is no doubt that the external argument Mary is introduced by the verbal structure in (59b) and therefore that [ize] spells out a transitive v. In fact, a Google search of words ending in [ize] yielded no examples in which [ize] spelled out an unaccusative verb phrase. However, according to Alexiadou (2001), [ize] necessarily spells out an unaccusative v in (59c). Thus, [ize] is transitive in (59b) and unaccusative in (59c). No doubt this problem could be taken care of by duplicating VI rules: for instance, one VI rule ‘‘ize1’’ spells out v(tr) as [ize] and another VI rule ‘‘ize2’’ spells out v(unacc) as [ize]—and the latter is obligatory if v(unacc) is governed by n. Both ize1 and ize2 would apply productively—for example, in deriving a verbal structure from an [al] adjective. But the need to duplicate rules suggests that we are on the wrong track. Moreover, one cannot say that legalization is actually derived from legalize, since the verbal structure included in both is very different.

In another version of the first approach, Harley (2009b) argues that the licensing of adverbs and accusative case depends not on the n or v head but on a functional category that is merged above v called Voice (see Kratzer 1996; the importance of separating v and Voice is argued for in Pylkka¨nen 2008, among others). Voice would then introduce the external argument, assign accusative case, and, presumably, decide on the morphology of modifiers. The reason why there are no adverbs or accusative case in nominalizations is that nominalizations include a vP but not a VoiceP. Since unaccusatives do not have a Voice head, this analysis is in essence a variant of Alexiadou’s and the same criticisms should apply. Additionally, I would argue that Harley’s (2009b) solution merely moves the question one step back without answering it: why is it that a nominalization cannot include a VoiceP between vP and nP? (See also Alexiadou 2001 for arguments that the structure of nominalizations may be fairly complex.)

Let us now consider the second approach. Kratzer (1996), Fu, Roeper, and Borer (2001), Roeper (2005), and others have claimed that nominalizations can license adverbs. Ackema and Neeleman (2004, 2007) and Newmeyer (2009) argue that they do not. If adverbs could be included in nominalizations, then (21) could be maintained without the need for a PC. So it is necessary to look into this in some detail.

As a general rule, adverbs in English nominalizations are sharply ungrammatical.

(60)

  • a.

    *A quickly derivation will show you how my theory works.

  • b.

    *John’s nationalization unjustly of the banks was sharply criticized.

  • c.

    *Lazarus’ suddenly resurrection surprised everyone.

Likewise, Spanish nominalizations, with or without a thematic vowel, are ungrammatical with adverbs.

(61)

graphic

However, the above-mentioned authors provide some examples that do not sound so ungrammatical (at least according to the authors). Here I include the examples that seem to me most convincing (from Roeper 2005).

(62)

  • a.

    ?While the removal of evidence purposefully is a crime, the removal of evidence unintentionally is not.

  • b.

    ?His explanation of the problem immediately to the tenants did not prevent a riot.

  • c.

    ?Protection of children completely from bad influence is unrealistic.

  • d.

    ?His resignation so suddenly gave rise to wild speculation.

(62a) is indeed puzzling. Note that the acceptability of this example depends to some extent on the contrast between the two adverbs in the two sentences. The simple sentence *The removal of evidence purposefully is a crime sounds considerably more degraded, as do similar examples such as *The judge was outraged by the removal of evidence purposefully. The adjective purposeful is always grammatical: while the purposeful removal of evidence . . . .

Example (62b) becomes degraded if to the tenants is omitted, which suggests that immediately might modify the PP and not the nominalized DP. I suspect the same thing could be going on in (62c), since *Protection of children completely is unrealistic is substantially degraded. Finally, in (62d) the acceptability of the adverb depends on the presence of so; compare (62d) with the ungrammatical *His resignation suddenly gave rise to wild speculation. Moreover, so can take either an adjective or an adverb as a complement (so sudden/so suddenly).

Thus, it seems that examples of nominalizations with adverbs must be constructed very carefully. Even a small change in the structure that should be irrelevant actually affects their acceptability substantially. The fragility of the data does not diminish their empirical value, but it should lead us to understand that the factors that affect the acceptability of (62a–d) are poorly understood and to realize that no general conclusions can be drawn from them. Apparently, the general case is that nominalizations do not accept adverbs. If we were to conclude on the basis of (62a–d) that nominalizations do accept adverbs, then we would have to account for the obvious ungrammaticality of the overwhelming majority of examples that can be found or constructed. It seems more economical—at least, from a methodological point of view—to take (62a–d) as outliers and not as the general case.

DM’s construction of complex words by means of cyclic phases not only predicts that nominalizations should include adverbs—it also predicts that denominal verbalizations should have adjectives as modifiers. The results tend to be strongly ungrammatical. Consider the following:

(63)

  • a.

    *John quick revolutionized entomology.

  • b.

    *John sad institutionalized Peter.

In fact, since the first phase is an nP, one should wonder whether genitive case with of might be possible; but it is not.

(64)

  • a.

    *John revolutionized of entomology.

  • b.

    *John institutionalized of Peter.

As discussed above, none of this is a problem if we adopt PC, because (63) and (64) never include an nP; they only include an n head. I take it that the first phase is the vP, which forces modifiers to take an adverbial form, as illustrated in (65).

(65)

graphic

5 Argument Structure

5.1 Argument Structure in Nominalization

Since Grimshaw 1990, it has become standard to distinguish between argument structure (AS) nominalizations and result (R) nominalizations. The following examples show the contrast:

(66)

  • a.

    The examination of the students took three hours.

  • b.

    The examination is in this room.

  • c.

    The exam is on the table.

AS-nominalizations denote an event, while R-nominalizations denote a result. The nominalization in (66a) is an AS-nominalization, involving the complement of the students. At the same time, (66a) necessarily denotes an event. These two properties—event denotation and argument structure—seem to occur together, and so sometimes I refer to the resulting structure as argument structure/event structure (AS/ES). The DP in (66c) necessarily denotes a result (in fact, an entity). The DP in (66b) is ambiguous, since it could denote either an event or a result. Many nominalizations are ambiguous in this way.

I think it is fair to say that the consensus analysis for AS-nominalizations is that their AS and ES are inherited from a vP included in the structure (see Alexiadou 2001, 2009 for a particularly clear exposition).8 In this view, R-nominals are distinguished from AS-nominals precisely in the absence of a vP.

(67)

graphic

Grimshaw (1990) argues that zero-derived nominals do not have AS, hence the fact that exam in (66c) has no AS.

My approach cannot adopt the assumption that AS-nominals always include a vP. Although I acknowledge the existence of nPs that do include a vP—gerundive nominals in English, for example—I also argue that other instances of nominalization in English and Spanish simply include a v embedded within an n (see (42)). These nominalizations may be AS-nominalizations, as shown in (68).

(68)

graphic

I take it to be implausible that a v embedded within another head could project an AS/ES. Thus, my approach commits me to the claim that the presence of an AS/ES is independent of a vP. In this section, I argue in fact that AS/ES can be built directly on the root. The structure that I will end up proposing is shown in (69). It includes an Event Phrase (EventP) selecting for the root and an internal argument (IA; see Borer 2012 on why the IA must be separated from the root, as well as Harley 1995 on the original proposal for a syntactic representation of events). The v is a component of n and is not involved in ES.

(69)

graphic

If the denotational semantics of the root is compatible with an AS/ES, the EventP is licensed and the derivation can proceed. In the case of an AS-nominalization, the next step is to merge an n with the EventP, forming an nP. It may well be possible that the n includes a v as a constituent, as in (69). I am not claiming that (69) is the only structure that can introduce an AS/ES; for present purposes, it is enough for me to argue that (69) is possible.

The structure in (69) divorces the presence of an AS/ES from vP. In sections 5.2 and 5.3, I present empirical evidence in favor of this separation. First, I show that there exist several types of AS-nominals whose structure does not include a verbal morpheme. Second, I show that result nominals whose structure contains a v are possible. The result is that the correlation between the presence of a verbal morpheme and the presence of an AS-nominalization is undermined in both directions. In section 5.4, I reintroduce the structure in (69) and discuss the role of v in grammar. In section 5.5, I explore an additional empirical consequence of my proposal.

5.2 AS-Nominalization without vP

The Spanish morpheme [(t)azo] is a nominal suffix that is inherently eventive. Consider the following examples:

(70)

graphic

(70a) shows that the noun paraguazo ‘blow with an umbrella’ is constructed with the noun paraguas ‘umbrella’ and the suffix [azo].9 The resulting word denotes an event (of semelfactive Aktionsart). However, no verbal morpheme is discernible in the structure, in particular, no thematic vowel. The AS consists of an agent and a beneficiary (or rather, maleficiary). The latter apears in dative case while the former appears in genitive case. This can be seen in (70b) (a Iñaki) and (70c) (de Cristina).

Here are some more examples:

(71)

graphic

Occasionally, the resulting noun may acquire an idiomatic meaning.

(72)

graphic

The [(t)azo] nominals pass the (appropriate) tests for AS-nominals (see Grimshaw 1990). They can be modified by an agent-oriented adjective.

(73)

  • graphic

  • the deliberate ‘blow with an umbrella’ of Cristina DAT Iñaki

  • ‘Cristina’s deliberate ‘‘blow with an umbrella’’ to Iñaki’

The agent of the [(t)azo] nominal can control PRO in an infinitival clause, even when this agent is implicit.

(74)

graphic

The event that the noun denotes can also control PRO in an infinitival.

(75)

graphic

Other tests do not work because of the characteristics of the event itself. Since [(t)azo] denotes a semelfactive, it sounds awkward with classic event modifiers such as en tres horas ‘in three hours’ or durante dos días ‘for two days’. It is not compatible with a modifier like frecuente ‘frequent’ in the singular, although it is in the plural.

(76)

  • a.

    la frecuente destrucción del medio ambiente (por grupos humanos

    the frequent destruction of.the environment by groups human

    de gran densidad demográfica)

    of great density demographic

    ‘the frequent destruction of the environment (by human groups of great demographic density)’

  • b.

    *el frecuente paraguazo

    the frequent ‘blow with an umbrella’

  • c.

    los frecuentes paraguazos

    the frequent ‘blows with an umbrella’

The contrasts in (76), are, as I suggested, a consequence of the type of event that the nominalization refers to. If John whacked Peter for three hours, this can only mean that John whacked Peter repeatedly for three hours, not that there was a single, three-hour-long event of whacking Peter. Since paraguazo refers to a single act of hitting someone with an umbrella, it follows that the modifier frecuente ‘frequent’ will not collocate with it.

Another feature of the nominalization paraguazo ‘blow with an umbrella’ is that the agent does not appear in a por ‘by’-phrase; rather, it appears in the genitive case. This follows the regular pattern in Spanish nominalizations: with the transitive voice, the IA is genitive while the EA (external argument) is in a ‘by’-phrase; but if the IA is absent or receives a lexical case (such as dative), the EA appears in the genitive case, as shown in (77). Since the IA in paraguazo receives dative case, it follows that the EA should receive genitive case, in agreement with the general pattern of case distribution in nominalizations.10

(77)

  • a.

    la reparación del coche por Juan

    the fixing of.the car by Juan

    ‘Juan’s fixing the car’

  • b.

    la reparación de Juan

    the repair of Juan

  • c.

    *la reparación por Juan

    the repair by Juan

It seems clear that nominalizations in [(t)azo] exemplify the possibility of an AS-nominal without a verb phrase as a constituent. I would suggest that the structure of a [(t)azo] noun is as in (78). The affix [(t)azo] is a spell-out of Event incorporated into n.

(78) [nP EA n [EventP IA Event [nP paraguas]]]

The suffix [ada] also yields AS-nominals that do not include any verbal morphology. The meanings of [ada] nominals are varied. The following are some examples:

(79)

graphic

Let’s pause on (79e) for a moment. The noun puñal has a derived verb, apuñalar ‘to stab’. But since the noun puñalada does not include the prefix [a], we can safely conclude that it does not derive from the verb apuñalar (on the other hand, the noun apuñalamiento ‘stabbing’ is derived from the verb). The other nouns listed in (79) do not have verbal equivalents.

Some of the [ada] nouns have IAs and EAs (puñalada, dentellada).

(80)

  • La puñalada/dentellada de Juan a Pedro asustó a todos los presentes.

  • the stabbing/biting of Juan DAT Pedro frightened ACC all the present

  • ‘Juan’s stabbing/biting Pedro frightened everybody present.’

The others have an intransitive meaning and do not have IAs. But they all pass other tests of AS/ES. They accept subject-oriented modifiers, implicit argument control, and control by the event. Example (81) includes an agent-oriented modifier.

(81)

  • La deliberada españolada de la alcaldesa de Madrid irritó a muchas

  • the deliberate españolada of the mayor of Madrid irritated ACC many

  • personas.

  • people

  • ‘Madrid’s mayor’s deliberate españolada irritated many people.’

The following example is ambiguous because PRO could be controlled by the event puñalada ‘stabbing’ or by the implicit agent of the event:

(82)

  • [La puñaladaiproj a Iñaki] para PROi/j asustar al público fue muy

  • the stabbing DAT Iñaki for frighten.INFACC.the public was very

  • efectiva.

  • effective

  • ‘The stabbing of Iñaki to frighten the public was very effective.’

Finally, some of these nominalizations also allow for (a)telic temporal modifiers.

(83)

  • La novillada de tres horas agotó las fuerzas de los mozos.

  • the novillada of three hours exhausted the strength of the men

  • ‘The novillada for three hours exhausted the men’s strength.’

Other nominalizations, such as puñalada ‘stabbing’ and dentellada ‘biting’, fail the temporal modifiers test. As was the case with paraguazo ‘blow with an umbrella’, this seems to me to be a natural consequence of the type of event they denote: for instance, if Brutus stabbed Caesar for three hours, this can only mean that Brutus stabbed Caesar repeatedly for three hours, not that there was a single, three-hour-long event of stabbing Caesar. Since puñalada refers to a single act of stabbing, it follows that the modifier de tres horas ‘for three hours’ will not collocate with it.

A third type of AS-nominal that does not include a verbal head consists of noun/verb pairs that are derived directly from the same root. The noun is inflected with masculine or feminine gender by means of a simple word marker.

(84)

graphic

As one can guess from the glosses, many of these nouns are ambiguous between an AS-nominal and an R-nominal. The following examples show their behavior as AS-nominals:

(85)

  • a.

    La compra de las acciones X por el banco Y es muy sospechosa.

    the purchase of the stocks X by the bank Y is very suspicious

    ‘Bank Y’s purchase of X stocks is very suspicious.’

  • b.

    La cosecha de trigo por Juan empezó en octubre.

    the harvest of wheat by Juan started in October

    ‘Juan’s wheat harvest started in October.’

  • c.

    El rechazo de María a las acciones innobles del banco la puso

    the rejection of Maria DAT the actions ignoble of.the bank CL.ACC put.PAST

    de patitas en la calle.

    of legs in the street

    (Also: El rechazo de las acciones innobles del banco por María . . . )

    ‘Maria’s rejection of the bank’s ignoble actions got her fired.’

  • d.

    El retraso del banco Y en los pagos desequilibró los mercados.

    the delay of.the bank Y in the payments unbalanced the markets

    ‘Bank Y’s delay in the payments unbalanced the markets.’

All of the examples in (84) pass Grimshaw’s (1990) tests for AS-nominals.12

(86)

  • a.

    la compra deliberada de las acciones X por el banco Y

    the purchase deliberate of the stocks X by the bank Y

    ‘bank Y’s deliberate purchase of X stocks’

  • b.

    la compra de las acciones X por el banco Y en tres días

    the purchase of the stocks X by the bank Y in three days

    ‘bank Y’s purchase of the X stocks in three days’

  • c.

    la constante compra de acciones X por el banco Y

    the constant purchasing of stocks X by the bank Y

    ‘bank Y’s constant purchasing of X stocks’

  • d.

    la compra de las acciones para animar a los inversores

    the purchasing of the stocks in.order.to encourage ACC the investors

    ‘the purchasing of stocks to encourage investors’

The question now is how to analyze this kind of noun. If we wanted to maintain the assumption that AS-nominals include verb phrases, as in mainstream DM, we would need to adopt a complicated analysis for these examples. For instance, we would have to derive a structure like the one in (87a) with both v and n terminals, and then delete the v (87b) before VI takes place (87c).

(87)

  • a.

    Syntax

    √retras] v] n

  • b.

    Delete v

    √retras] n

  • c.

    VI

    √retras+o

This deletion operation would have to operate with lookahead. It is not the case that v-deletion takes place generally in nominalizations in Spanish (as shown by the nominalizations in [m(i)ento], [cion], [dura], and so on that I discussed above). Only if n is to be spelled out as a word marker would v be deleted.

Alternatively, we could allow VI of v and n and subsequently delete the spelled-out thematic vowel by means of a readjustment rule. This readjustment rule would apply only when followed by a word marker.

(88)

  • a.

    Syntax

    √retras] v] n

  • b.

    VI

    √retras+a+o

  • c.

    Delete thematic vowel

    √retras_o

However, deletion of the thematic vowel is not motivated by Spanish phonotactics, or for any other reason that I can think of. It would stand out as an arbitrary operation. In fact, deriving the noun from the verb is itself fairly arbitrary; one could just as well assume that the verb is derived from the noun. An analysis in which the noun is derived from the verb could only be motivated exclusively for theory-internal reasons. The most parsimonious approach to these noun/verb pairs is to view them as being derived directly from the root. It so happens that this solution is also the closest to the spirit of what DM tries to accomplish. But if we adopt this solution, then we have to abandon the notion that AS-nominals necessarily receive their AS/ES from a verbal structure dominated by the nP.

Fábregas (2013) discusses noun/verb pairs in Spanish in detail and shows that there is a subset of them that have no AS/ES.

(89)

graphic

Fábregas uses a battery of tests to show the difference between the examples in (84) and those in (89). For instance, he shows that nouns in Spanish generally do not accept prepositions other than de ‘of ’, con ‘with’, or sin ‘without’. The only exception are AS-nominals, which do accept a wider range of prepositions. Interestingly, the nouns listed in (84) accept prepositions, while the nouns listed in (89) behave like normal nouns in not allowing prepositions.

(90)

  • a.

    la compra de zapatos elegantes por los aristócratas

    the purchase of shoes elegant by the aristocrats

    ‘the purchase of elegant shoes by aristocrats’

  • b.

    *la conserva de atún por Pescanova

    the canning of tuna by Pescanova

Fábregas proposes that the difference between the nouns in (84) and those in (89) can be captured by adopting a structure in which the AS-nominals in (84) are derived from a verb while the non-AS-nominals in (89) are derived directly from the root. As mentioned, deriving nouns from verbs forces us to incorporate some complicated operations deleting terminal nodes or exponents. Fortunately, the framework I am proposing allows for a simple approach to the distinction without complicating things on the PF side of the grammar: the structure of nominals with an AS/ES includes an EventP, while the structure of those without an AS/ES does not. The tests used by Fábregas can be reinterpreted accordingly with no loss of insight. For instance, AS-nominals accept all sorts of PPs because the Event head licenses them. Non-AS-nominals do not include an Event head and consequently do not license prepositions. This way, it is possible to maintain that there is no verb in the structure of AS-nominals and therefore no verb needs to be deleted.

It seems that the [(t)azo] and [ada] nominalizations, together with the noun/verb pairs, suffice to make the case that AS-nominals do not require a verbal structure. There are other nominalizations in Spanish that seem to denote complex events but, since they seem to fail Grimshaw’s (1990) tests systematically, I add them here only tentatively. One such case is the nominalizer [eria]. This suffix attaches to a noun (or maybe a root) that refers to an entity and yields an activity that is practiced regularly. Again, the structure of these nouns contains no verbal morphemes (unlike the structure of the English glosses, which seem very close equivalents). The following are some examples:

(91)

graphic

There are some [eria] nouns that seem to have a verbal equivalent, although the connection in meaning between them tends to be somewhat distant and the noun never includes a thematic vowel.

(92)

graphic

At least some [eria] nominals can take complements easily, which suggests that they can be AS-nominals. However, they seem to resist an EA.13

(93)

  • a.

    La jardinería de rosas (*por Iñaki) es muy rentable.

    the gardening of roses (by Iñaki) is very profitable

    ‘Rose gardening (by Iñaki) is very profitable.’

  • b.

    La minería de carbón (*por Iñaki) requiere mucho trabajo.

    the mining of coal (by Iñaki) requires much work

    ‘Coal mining (by Iñaki) requires a lot of work.’

  • c.

    La familia O. se ha dedicado a la ganadería de reses de lidia.

    the family O. SE has devoted to the ranching of bulls of fight

    ‘The O. family has devoted itself to the ranching of bulls for bullfighting.’

5.3 R-Nominals with a Little v

R-nominals whose structure does exhibit a thematic vowel—hence a v—are also possible. For example:

(94)

  • a.

    ampliación ‘expansion’

    Este barrio se construyó como una ampliación al casco antiguo.

    this neighborhood SE built as an expansion of.the town old

    ‘This neighborhood was built as an expansion of the old town.’

  • b.

    separación ‘separation’

    Entre el punto X y el punto Y hay una separación de 20 metros.

    between the point X and the point Y there.is a separation of 20 meters

    ‘Between point X and point Y is a distance of 20 meters.’

  • c.

    definición ‘definition’

    La definición de inercia se encuentra en la página 20.

    the definition of inertia SE finds in the page 20

    ‘The definition of inertia can be found on page 20.’

These nominalizations also have an AS-nominal version.

(95)

  • a.

    la constante ampliación de la deuda por el gobierno

    ‘the constant expansion of the debt by the government’

  • b.

    la ampliación de la deuda en tres días

    ‘the expansion of the debt in three days’

  • c.

    la ampliación de la deuda para animar a los inversores

    the expansion of the debt in.order.to encourage ACC the investors

    ‘the expansion of the debt to encourage investors’

Thus, the presence of a verbal morpheme in the structure does not guarantee the presence of an AS/ES. Maintaining the assumption that AS/ES structures grow out of a v would force us to claim in addition that this AS/ES can be excised from a v—but only if that v is the complement of a nominalizing affix.

5.4 Argument Structure/Event Structure and the Role of v

In the previous sections, we saw that DP can include an AS/ES while vP does not need to (if it is embedded within a nominalization). The conclusion seems to be that the presence of an AS/ ES does not depend on the category involved; rather, it depends on the denotational properties of the root in interaction with the structure it finds itself in (a root like √table will never be a constituent in an AS/ES without significantly altering its denotation while a root like √cough may be). My proposal, represented in (69), is repeated here for convenience.

(96)

graphic

The root does not select for an IA; Event does. But when the structure is translated into a semantic representation, the resulting representation must be coherent: if the denotation of the root accepts an AS/ES, the structure can be interpreted, but if it does not, the representation is thrown out. The EA and other arguments are introduced by n (or v,a), by Voice, or by Applicative heads (this is not crucial for my purposes).

One question that needs to be addressed in this framework is this: why do all vPs require an AS/ES while many nPs do not? (That is, why are verbs sensitive to the θ-Criterion while nouns are not?) This question would have an immediate answer if AS/ES always depended on the presence of a vP—a position that, I believe, is sufficiently refuted by the evidence presented here. I do not have a fully developed answer at this point, but I would like to suggest that the reason why it seems as if vPs require AS/ES lies in the properties of T, not v. The denotation of T is an instruction for the semantic component to locate a certain event within a time frame. If there is no event in the structure, T has nothing to locate in the time frame. A nominal structure does not include a T, and therefore the presence of AS/ES is optional (here I leave aside notorious troublemakers such as the future president). A clausal structure does include a T, and therefore AS/ES is obligatory.14

A second question is this: what is the role of v? In nominalizations, v can project a full vP (as in gerundive nominals), or it can be embedded within n. When v is embedded within n, the morphosyntactic properties of the structure are altered, but the actual meaning of the sentence remains (almost) unchanged. I believe this reveals how light the light verb is. When v takes a complement and projects a full phrase, it fulfills several grammatical functions: it provides a label for the structure, it establishes morphosyntactic dependencies (and therefore, assigns Case (and case) and values or checks agreement), and can trigger movement. However, it is unclear that v has an inherent semantic denotation or function beyond categorization, particularly if we assume that EA is introduced by a separate head, Voice or Init (Kratzer 1996, Ramchand 1997). Rather, v exists in the syntax because it has an ‘‘effect on output’’ to the extent that the grammatical dependencies that it triggers do entail semantic consequences. For instance, an influential line of research initiated in Chomsky 2001 (but with many precedents) connects the presence of a certain feature in v with Icelandic object shift. Object shift has an interpretive consequence: the shifted object acquires a specific interpretation.

If that is the case, it is not surprising that a nominalization that includes only a v is so similar in meaning to one that introduces a vP. The difference between them is more noticeable in the properties of the morphosyntactic structure that results. When v is introduced as a component of n, nothing semantic is lost; only its grammatical properties are ‘‘swallowed’’ within the bigger head (see López 2007 and Preminger 2011 for arguments that leaving unvalued features on a functional head unresolved does not lead to ungrammaticality).

5.5 Event Spells Out

I would like to finish this section by suggesting that the structure in (96) may provide a solution for an old if small mystery of Spanish grammar. Many Spanish nouns have a ‘‘double,’’ different only to the extent that it includes a suffix [e] between the root and the word marker. However, the difference in meaning between the noun and its double is stark: the noun with [e] denotes a complex event, the noun without it denotes an entity. The nouns with [e] have verbal counterparts. Likewise, some adjectives have [e] counterparts. Here are some examples:

(97)

graphic

Fábregas (2011) points out that there are some [e] nouns and adjectives without a verbal counterpart and vice versa.

(98)

graphic

The [e] nouns pass Grimshaw’s (1990) tests for complex event nominals. In the following example, the noun blanqueo ‘laundering’ is modified by an agent-oriented adjective, it has an IA and an EA, and the event itself can control PRO in the nonfinite clause:

(99)

  • [El deliberado blanqueo de dinero por los bancos]i para PROi proteger

  • the deliberate laundering of money by the banks for protect.INF

  • a los inversores excitó la ira del público.

  • ACC the investors aroused the anger of.the public

  • ‘The deliberate money laundering by the banks to protect investors aroused the public’s anger.’

The [e] morpheme is commonly taken in Spanish descriptive grammars to be a verbal suffix (or, rather, the whole [ear] is taken to be a verbal suffix), and [e] nouns are defined as being deverbal (see Serrano-Dolader 1999). Alternatively, Fábregas (2011) argues that the verb is derived from the noun. I believe that both approaches present some drawbacks.

First, both approaches would have to confront the gaps exemplified in (98). For instance, analyzing famoseo ‘related to celebrities’ as derived from *famosear is somewhat implausible (as Fábregas (2011) points out). Second, analyzing [e] words as being deverbal or denominal/deadjectival raises the issue of arbitrariness that I discussed in relation to (84a–d). Finally, neither approach provides an analysis for [e] as an independent morpheme. Rather, both approaches take [ear] and/or [eo] as unanalyzable blocks—but [e] clearly makes a meaningful and systematic contribution, as the contrasts in (97) demonstrate. How, then, should we analyze [e]? For the sake of the straw-man argument, let us assume that [e] is an exponent of v and therefore the thematic vowel is a morpheme associated with v rather than v itself. So far, so good. The tricky part is that, when v becomes the complement of n, the thematic vowel must disappear (bombeo, not *bombeao). Thematic vowels do not disappear in nominalizing contexts generally, as has been amply demonstrated here; rather, thematic vowels would only disappear when sandwiched between [e] and a nominalizing affix. It is of course possible to propose such a deletion of the thematic vowel, but it is worth considering a more elegant possibility; this is in fact supplied by the structure in (96).

Adopting the structure in (96) dispels the mystery: [e] is a spell-out of the Event head that builds an AS/ES on top of a root (or maybe on top of a noun, at least sometimes). The resulting EventP can be selected by an n or a v, as usual.

(100)

graphic

A final consideration is this. If we assume that having an AS/ES is a property of verbs, then it follows that [e] nouns must be deverbal and the traditional approach is the only one possible. But if AS/ES is not a privilege of verbs and an AS/ES can appear in the absence of any verbal morphology, as I have argued, then there is no reason why [e] nouns should be regarded as derived from verbs. If there is no derivation of nouns from verbs or vice versa, then the gaps in (98) are not surprising and there is no problem with arbitrariness. Since [e] is not a verbal morpheme, we do not have to concern ourselves with the putative disappearance of the thematic vowel in the nominal forms. Additionally, the morpheme [e] receives an analysis that reflects its contribution to sentence meaning. Finally, (100) respects the conclusion reached above, namely, that event structures exist independently of the category that selects them.

As an anonymous reviewer points out, the [e] morpheme is sometimes preceded by an affix that introduces a manner or aspect nuance (see Fábregas 2011 for discussion).

(101)

graphic

(102)

graphic

Interestingly, these affixes seem to appear only in nouns and verbs that also include the morpheme [e]. If that is the case, we could then see these affixes as adjuncts to Event, merged to Event in a parallel derivation, which are spelled out only when the Event head is as well.

(103)

graphic

(104)

graphic

6 Compounding

The database on compounding has grown exponentially in recent years and, correspondingly, so has linguists’ realization of the complexity of the phenomena (see, e.g., Štekauer and Lieber 2005, Scalise and Masini 2012). I do not try to present a general approach to compounding in this article. Instead, I take an approach to compounding that explicitly attempts to cohere with mainstream DM, that of Harley (2009a), and show that integrating the two ingredients of (3) into the analysis provides empirical advantages.

6.1 Compounding in DM

Harley (2009a) argues that compounding in English (she does not discuss other languages) is derived by syntactic head-to-head movement, as in Baker’s (1988 and later work) analysis of noun incorporation (NI) in Mohawk and Inuit. For a compound like truck driver, Harley takes it that truck is originally merged as the complement of drive. The resulting drive truck is then selected by the agentive nominalizer [er]. Successive head-to-head incorporation yields truck driver.

(105)

graphic

The outcome of this way of deriving compounds is that the DM assumption that there is no separation between morphology and syntax is maintained. This is desirable. However, it is worthwhile exploring the possibility that incorporating PC into the theory might give a better result.

Let us start by noting that there are two main differences between compounds like truck driver and NI as described in the literature (see, e.g., Gerdts’s (1998) review article as well as Mithun 1984, 1986, Baker 1988, 1996, Rosen 1989). The first is that compounds may involve any relationship between the head and the second constituent. In truck driver, truck can be seen as the complement of drive, but compounds with adjunct modifiers also exist. For example:

(106)

NI, on the other hand, involves only relations between a verbal head and its IA.15

Harley tackles this apparent difference between NI and compounding by pointing out that if there is a complement in the structure, incorporation of an adjunct is impossible (see also Roeper and Siegel’s (1978) First Sister Principle). Consider the following examples:

(107)

  • a.

    The farmer grows wheat quickly.

  • b.

    a wheat-growing farmer

  • c.

    *a quick-growing farmer

  • d.

    The wheat grows quickly.

  • e.

    quick-growing wheat

The verb (or the root) grow can form a nominal compound with its object wheat (107b) but not with the modifier quick (107c); presumably, the direct object is invisibly present in the structure of (107c). (107e) is grammatical because there is no direct object; wheat is a subject in this example. Harley argues that if we assume a phrase structure without vacuous projections, the head-complement and the head-modifier (sans complement) configurations are identical. If so, compounding and NI could be assimilated as operations that apply to sisters.

This argument entails that the Merge position of direct objects and unaccusative subjects is not the same (as Harley points out in footnote 7); in other words, wheat occupies different positions in (107a) and (107d), only the first instance being a sister to the root (see also the fast-falling snow example, (106)). Thus, Harley commits herself to the theory of IA introduction illustrated in (108).

(108)

graphic

However, the literature on NI shows many examples of subjects of unaccusatives that incorporate—which led to Baker’s (1988) original hypothesis that the argument that incorporates must be governed in D-Structure by the incorporating head. Baker’s analysis—and indeed, a good deal of independent literature on argument projection—should lead to the consequence that wheat in (107a) and wheat in (107d) are configurationally identical. It is therefore unclear what the examples in (107) show. Moreover, it seems to me that a case like (109) provides a counterexample.

(109)

  • a.

    John hit the ball quickly.

  • b.

    a ball-hitting youth

  • c.

    a quick-hitting youth

In (109b) there is a hidden object, just as there is in (107c). But as (109c) shows, this hidden object does not present an obstacle to forming a compound with the adjunct.

The second difference between English compounding and NI can be illustrated with the help of the compound in (110).

(110) small appliance factory

This phrase is ambiguous: it could refer to a small factory that builds appliances or to a factory that builds small appliances. Let us focus on the second meaning. Following Harley’s (2009a) proposal, one should be able to derive it as in (111).

(111)

  • factory [small appliance] →

  • appliance-factory [small t] →

  • *This is an appliance-factory small.

That is, small appliance starts off as complement of factory, and then appliance incorporates into factory, stranding small. This yields an impossible phrase in English. For some reason, appliance must incorporate into small before the whole phrase small appliance incorporates into factory. This does not look at all like what we see in canonical incorporating languages. In incorporating languages, there are two possibilities: either a noun can incorporate, stranding other constituents (as in Mohawk; see Baker 1996), or the noun is base-generated as a bare singular and then incorporated (see Gerdts 1998 and especially Rosen 1989). But the case where a noun incorporates into an adjective and the resulting compound incorporates again is, I believe, unattested.

6.2 Compounding in PC

Examples like small appliance factory lead to considering the possibility of deriving compounds in PC and then adjoining the output to the core derivation. Small appliance factory is derived by merging small and appliance in PC and adjoining the output to factory.

(112)

graphic

Within my assumptions, truck driver is derived as in (113).

(113)

graphic

Thus, the difference between English compounding and NI can easily be accounted for within DM if we adopt the assumptions in (3). NI is based on one single computation path that, by definition, involves head-complement relations. Compounding is based on adjoining the output of a PC to a root or a word. Adjunction does not have a head-complement requirement.

Let us now return to phrasal compounds, like the one in (114).

(114) He gave me his favorite don’t-you-tell-me-what-to-do look.

As pointed out in section 1, this form of compounding is perfectly productive in English, as well as in Germanic generally and in Chinese. Notice that phrasal compounds seem to be ideally suited as evidence for PC in word formation. In the current example, don’t you tell me what to do forms a CP in PC, which is then adjoined to n.

(115)

graphic

In fact, I do not see how phrasal compounding could possibly be accounted for in a DM-inspired model without PC.

Sato (2010) analyzes phrasal compounds in two steps. (a) The phrase is turned into a noun by zero derivation. (b) The noun is reintroduced into the derivation by means of an operation that Sato calls renumeration: the zero-derived new word becomes a member of the numeration (for the notion of numeration, see Chomsky 1995:chap. 3). However, this analysis leaves something unexplained: Why can a renumerated item only be part of a compound? Why can’t it project its own phrase? An item in a numeration should be able to project a whole phrase. However, renumerated items never project a whole phrase; they can only be used as a component of a complex n.16Harley (2009a) implements Sato’s step (a) in a DM vocabulary by having the phrase become the complement of an n whose semantic function is to ‘‘reify’’ the phrase: the phrase becomes a concept and as such is reintroduced into the derivation. It is this n that is reintroduced into the derivation.

Within my system, step (b) is rendered unnecessary by PC. That is, it is not necessary to ‘‘reintroduce’’ an item into the derivation or to ‘‘renumerate’’ a phrase. One constructs a phrase in PC, which is then adjoined to an n, both operations generally available in CHL.

As for step (a), I am not entirely certain that it is necessary either. It seems that what triggers the perceived necessity of step (a) is the need to force a big foot into a small shoe: in essence, step (a) makes the big foot small so it can fit into the shoe. But it might be the case that there is nothing wrong with having an entire phrase as a subword. There is nothing in bare phrase structure that rules it out, as far as I can tell. In effect, I would like to claim that the computational system itself, by creating the structure in (115), does the job of reconceptualization or reification that Harley describes: since the CP is a constituent of n, it necessarily becomes a modifier of n.

7 Ackema and Neeleman’s Objection

Ackema (1999) and Ackema and Neeleman (2004, 2007) argue against the idea that there is only one CHL. In the model they propose, word formation and phrase formation constitute two submodules of one larger computational module. Here I present their major empirical challenge to the one-module theory and show how the PC model overcomes it.

Ackema and Neeleman point out that the DM-style single combinatorial system predicts that examples like (116) and (117) should be grammatical.

(116) *Centralize more to our arguments than we thought.

(117) *Washable carefully by dipping repeatedly.

Consider first example (116) and the three-step derivation in (118).

(118)

  • Derivation of (116)

    1. [DegP more central to our arguments than we thought]

    2. [vP ize [DegP more central to our arguments than we thought]]

    3. centralize more t(central) to our arguments than we thought

In step 1, we build a Degree Phrase (DegP) headed by the adjective central, which includes the comparative correlative more . . . than and the PP to our arguments. This DegP is then selected by the verbalizer [ize], as in step 2. In step 3, the adjective incorporates into the verbalizer, yielding an ungrammatical structure.

(117) presents a similar phenomenon. In step 1, we construct a vP with two modifiers. When we derive an aP out of this vP, the presence of these modifiers makes the result ungrammatical.

(119)

  • Derivation of (117)

    1. [vP wash carefully by dipping repeatedly]

    2. [aP able [vP wash carefully by dipping repeatedly]]

    3. washable t(wash) carefully by dipping repeatedly

Regarding (117), one could argue that washable is built not from the verb to wash but from the root √wash—and that therefore there is no vP in the structure and (117) never surfaces. But (116) remains, since central is composed of the root √centr and the adjectival affix [al]. One could imagine that building parallel examples would not be very difficult.

Ackema and Neeleman (2004, 2007) point out that the same empirical problem arises if compounds are derived in a Harley-style manner. If city center is the output of NI, then the example in (120) and the derivation in (121) should be possible.

(120) *the city center of a prosperous medieval in northern Italy

(121)

  • Derivation of (120)

    1. the center of [a prosperous medieval city in northern Italy]

    2. the city center of [a prosperous medieval t(city) in northern Italy]

Ackema and Neeleman correctly argue that these problems are avoided if centralize, washable, and city center are fully derived in the word formation system (or the ‘‘Lexicon,’’ as it is traditionally called) before they are inserted into the phrase formation system, where they take the modifiers that are appropriate for their category label.

However, adopting PC in word formation allows us to rule out structures like (116)–(117) without resorting to separate modules. Consider (116), for instance. We take little v and little a to attach in PC; that is, v attaches to a before the latter merges with its complement, adjuncts, and so on, as shown in (122).

(122)

graphic

There is no aP projection; that is, the head a is embedded in a v and never takes the root √centr as a complement. This is the reason why there are no aP modifiers: since central is never a constituent, there is no periphrastic comparison, as in more central than. The same approach applies to (123).

(123)

graphic

The derivation in (123) does not yield the verb wash, so it follows that we should not expect any verbal modifiers.

Finally, consider (120). The compound city center is built in a PC. Thus, it can take modifiers to form DPs such as the prosperous medieval city center, but the DP in (120) is simply never constructed.

8 Conclusions

I have argued that we should allow for parallel computations in word formation, as a natural consequence of assuming that there is only one CHL for phrasal and word syntax. This simple assumption has yielded empirical dividends concerning the structure of nominalizations in Spanish, compounds in English, and the possibility of overcoming the Ackema-Neeleman objection. Along the way, I have revisited DM’s assumptions concerning allomorphy and AS-nominals. With respect to allomorphy, I have presented an algorithm based on minimality that derives the fact that in Spanish nominalizations the root can decide on an allomorph across a little v. With respect to AS-nominals, I have argued that AS/ES can be built on a root without including a verbal morpheme. This proposal extends the empirical scope of DM without sacrificing any of its fundamental tenets.

Notes

I would like to thank Víctor Acedo-Matellán and two reviewers for Linguistic Inquiry for their comments on an earlier version of this article, as well as Karlos Arregi, Kay González-Vilbazo, and Xavier Villalba for discussions on matters related to this article. For many insightful questions, I would also like to thank the various audiences that heard presentations of this material, at University of Hamburg, University of Frankfurt, Autonomous University of Barcelona, University of Leipzig, University of Paris-Diderot, and University of Leiden. I would like to thank Natascha Müller and the Romance Languages Department at the University of Wuppertal as well as the Theoretical Linguistics Group at the Autonomous University of Barcelona for their hospitality. I would also like to gratefully acknowledge a Fulbright Commission fellowship that gave me six months of nothing to worry about except research. All remaining errors are mine.

1 This assumption has become controversial as a result of the widespread adoption of Chomsky’s (1995) bare phrase structure. Indeed, if the only computational mechanism is Merge of two units a and b resulting in the set {a,b}, it is unclear how Merge can sometimes yield complex heads and sometimes phrasal structures. There have been some attempts at eliminating head movement from the theory of grammar, most notably by Matushansky (2006). Matushansky proposes that head movement in fact involves movement of a head to a Spec position followed by morphological fusion. However, Matushansky’s proposal runs into implementation problems with successive head movements. Whatever conclusion the field reaches concerning head movement, I take it that the existence of complex heads is a fact and that they are created by means of IM, directly (head-to-head movement) or by means of some intermediate step (see Baker 2009, Roberts 2010 for discussion).

2Malchukov’s (2004) typological study shows that nominalizations can affect the clausal spine at different points: within my assumptions, above vP, above TP, or above CP. Thus, (18) should be taken as representing one particular type of nominalization in one particular language.

3 As Anke Assman ( pers. comm.) points out, (21) seems to run counter to the traditional approach to adjectives and adverbs in Spanish and many other languages, according to which an adverb is an adjective with an extra piece of morphology (e.g., cleverly = [clever]+[ly]). I can think of two approaches to this issue. Within one approach, we maintain the traditional assumption that adverbs include an extra morpheme. This would be consistent with (21) if this extra morpheme is inserted postsyntactically as a consequence of a dependency with an appropriate phase head (compare Embick and Noyer’s (2001) analysis of morphological case morphemes). Alternatively, we can assume that adjectives and adverbs have the same morphosyntactic structure: √] a] x, where x defines the structure as adverbial or adjectival in a dependency with a phase head.

4Oltra-Massuet (1999) makes the novel claim that a thematic vowel is attached to any functional category of the verbal extended projection, which allows for an original account of stress in Catalan (and in Oltra-Massuet and Arregi 2005, in Spanish as well).

5Acquaviva (2009) presents independent arguments against having class features on roots. Bermúdez-Otero (2013) takes a different tack, arguing that List 1 includes roots with their category heads and that the latter include the theme vowels.

6 I leave [iz] unspecified for category in the example. The tradition begun with Marantz 1993 considers it to be a version of little v. In López 2001, causative morphemes are analyzed as roots.

7Marvin (2003) develops a DM approach to stress in English based on phases but she only deals with secondary stresses within the word; the final word stress is relegated to another, unspecified linguistic module.

8 In Grimshaw 1990, nominalizations are built on verbs that attach to nominalizing affixes. The output is a noun that inherits its AS from the verb and its event argument from the affix (see in particular Grimshaw 1990:66). Work inspired by DM tends to regard a nominalization as composed of a nominalizing affix that takes a verbal phrase as a complement, the verbal phrase contributing the event semantics.

9 The derivation of paraguazo ‘blow with an umbrella’ is more complex than is presented in the text. Paraguas ‘umbrella’ is itself a compound noun: para ‘for’ + aguas ‘water’. The noun paraguazo deletes the final /s/ of paraguas. The details of the derivation do not affect the argument.

It seems that [(t)azo] may take nouns as well as roots as complements. The word manotazo ‘slap’ is constructed from mano ‘hand’ + [tazo]. Manotazo displays the word marker [o], which is a sign that a noun has been formed (see Harris 1991 on word markers). In contrast, manaza ‘big hand’ is constructed on the stem [man] by adding the augmentative affix [az]. On the other hand, we also find zapatazo ‘blow with a shoe’, constructed on the root √zapat rather than on the noun zapato ‘shoe’; as far as I can tell, *zapatotazo does not exist.

10 Morphological case in Spanish nominals could be the subject of another article. Apart from the remarks in the text, I think the following generalizations hold:

  • (i)
    • a.

      Psych verbs: The experiencer appears in the genitive case, the theme in a ‘by’-phrase or the dative case.

    • b.

      el interés de María por las matemáticas

      the interest of Maria by the mathematics

      ‘Maria’s interest in math’

  • (ii)
    • a.

      Statives: Both arguments appear in the genitive case.

    • b.

      la dulzura de carácter de Juan

      the sweetness of character of Juan

      ‘Juan’s sweetness of character’

11 The stems chiquill- and novill- themselves include the diminutive morpheme [ll]. This does not affect the analysis.

12 The Spanish noun/verb pairs look like cases of English conversion, also known as Ø-derived nouns. However, the similarity seems to be only superficial if, as has frequently been argued, English Ø-derived nouns do not have AS/ ES (Grimshaw 1990, Borer 2013). An important question, unfortunately beyond the limits of this article, is the reason for this difference.

13 An anonymous reviewer does not accept jardinería de rosas ‘rose gardening’. He or she does not comment on the other examples.

14 As an anonymous reviewer points out, fleshing out this suggestion should include an account of nonfinite T. Nonfinite T seems to involve a spectrum of possible values: it is sometimes anaphoric, sometimes generic, sometimes indistinguishable from a nominalization. For the time being, I limit myself to pointing out that all forms of T establish some relation between a time and an event, leaving the details for future research.

15Spencer (1995) argues that the NI in Chukchi may involve adjuncts. Similarly, it has been argued that Greek allows adverb incorporation (see the critical discussion in Smirniotopoulos and Joseph 1997). Mathieu (2009) discusses examples of NI in Ojibwa in which the noun clearly drags a lot of material along with it. In fact, Mathieu argues that Ojibwa’s NI is actually movement of an entire DP to the Spec of a light verb. At present, I cannot explain why Chukchi’s NI differs from the mainstream cases discussed by Gerdts (1998), for example. For the purposes of this article, I assume

16 An anonymous reviewer points out that there is an apparent counterexample to this assertion. In the ‘‘self-reference’’ function, a renumerated noun can actually project a DP.

  • (i)

    That eternal don’t you tell me what to do of his is getting on my nerves.

References

Abney,
Stephen
.
1987
.
The English noun phrase in its sentential aspect
.
Doctoral dissertation, MIT, Cambridge, MA
.
Acedo-Matellán,
Víctor
.
2013
.
Merging roots in bare phrase structure and the conflation/incorporation distinction
. In
NELS 42
, ed. by
Stefan
Keine
and
Shayne
Sloggett
,
1
14
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Ackema,
Peter
.
1999
.
Issues in morphosyntax
.
Amsterdam
:
John Benjamins
.
Ackema,
Peter
, and
Ad
Neeleman
.
2004
.
Beyond morphology
.
Oxford
:
Oxford University Press
.
Ackema,
Peter
, and
Ad
Neeleman
.
2007
.
Morphology ≠ syntax
. In
The Oxford handbook of linguistic interfaces
, ed. by
Gillian
Ramchand
and
Charles
Weiss
,
325
352
.
Oxford
:
Oxford University Press
.
Acquaviva,
Paolo
.
2009
.
Roots and lexicality in Distributed Morphology
.
York Papers in Linguistics Series 2
10
:
1
21
.
Alexiadou,
Artemis
.
2001
.
Functional structure in nominals: Nominalizations and ergativity
.
Philadelphia
:
John Benjamins
.
Alexiadou,
Artemis
.
2009
.
On the role of syntactic locality in morphological processes: The case of Greek derived nominals
. In
Quantification, definiteness, and nominalizations
, ed. by
Anastasia
Giannakidou
and
Monika
Rathert
,
253
280
.
Oxford
:
Oxford University Press
.
Baker,
Mark
.
1988
.
Incorporation
.
Chicago
:
University of Chicago Press
.
Baker,
Mark
.
1996
.
The polysynthesis parameter
.
Oxford
:
Oxford University Press
.
Baker,
Mark
.
2009
.
Is head movement still needed for noun incorporation? The case of Mapundung
.
Lingua
119
:
148
165
.
Bauer,
Laurie
.
1983
.
English word-formation
.
Cambridge
:
Cambridge University Press
.
Bermúdez-Otero,
Ricardo
.
2013
.
The Spanish lexicon stores stems with theme vowels, not roots with inflectional class features
.
Probus
25
:
3
103
.
Bobaljik,
Jonathan David
.
2000
.
The ins and outs of contextual allomorphy
.
Ms., McGill University, Montréal
.
Bonet,
Eulàlia
, and
Daniel
Harbour
.
2012
.
Contextual allomorphy
. In
The morphology and phonology of exponence
, ed. by
Jochen
Trommer
,
195
235
.
Oxford
:
Oxford University Press
.
Borer,
Hagit
.
2012
.
In the event of a nominal
. In
The theta system: Argument structure at the interface
, ed. by
Martin
Everaert
,
Marijana
Marelj
, and
Tal
Siloni
,
103
149
.
Oxford
:
Oxford University Press
.
Borer,
Hagit
.
2013
.
Derived nominals and the domain of content
.
Lingua
141
:
71
96
.
Chomsky,
Noam
.
1995
.
The Minimalist Program
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2001
.
Derivation by phase
. In
Ken Hale: A life in language
, ed. by
Michael
Kenstowicz
,
1
52
.
Cambridge, MA
:
MIT Press
.
Chung,
Inkie
.
2007
.
Suppletive negation in Korean and Distributed Morphology
.
Lingua
117
:
95
148
.
Embick,
David
.
2004
.
On the structure of resultative participles in English
.
Linguistic Inquiry
35
:
355
392
.
Embick,
David
.
2010
.
Localism versus globalism in morphology and phonology
.
Cambridge, MA
:
MIT Press
.
Embick,
David
, and
Rolf
Noyer
.
2001
.
Movement operations after syntax
.
Linguistic Inquiry
32
:
555
595
.
Embick,
David
, and
Rolf
Noyer
.
2007
.
Distributed Morphology and the syntax-morphology interface
. In
The Oxford handbook of linguistic interfaces
, ed. by
Gillian
Ramchand
and
Charles
Weiss
,
289
324
.
Oxford
:
Oxford University Press
.
Fábregas,
Antonio
.
2011
.
Dos formas de estar callado: Nominalizaciones desinenciales
.
Revista de Investigación Lingüística
14
:
167
189
.
Fábregas,
Antonio
.
2013
.
Argument structure and morphologically underived nouns in Spanish and English
.
Lingua
141
:
97
120
.
Fu,
Jingqi
,
Tom
Roeper
, and
Hagit
Borer
.
2001
.
The VP within process nominals: Evidence from adverbs and the VP-anaphor do so
.
Natural Language and Linguistic Theory
19
:
549
582
.
Gerdts,
Donna B
.
1998
.
Incorporation
. In
The handbook of morphology
, ed. by
Andrew
Spencer
and
Arnold
Zwicky
,
84
100
.
Oxford
:
Blackwell
.
Grimshaw,
Jane
.
1990
.
Argument structure
.
Cambridge, MA
:
MIT Press
.
Halle,
Morris
, and
Alec
Marantz
.
1993
.
Distributed Morphology and the pieces of inflection
. In
The view from Building 20
, ed. by
Kenneth
Hale
and
Samuel Jay
Keyser
,
111
176
.
Cambridge, MA
:
MIT Press
.
Harley,
Heidi
.
1995
.
Events and licensing
.
Doctoral dissertation, MIT, Cambridge, MA
.
Harley,
Heidi
.
2009a
.
Compounding in Distributed Morphology
. In
The Oxford handbook of compounding
, ed. by
Rochelle
Lieber
,
129
144
.
Oxford
:
Oxford University Press
.
Harley,
Heidi
.
2009b
.
The morphology of nominalizations and the syntax of vP
. In
Quantification, definiteness, and nominalizations
, ed. by
Anastasia
Giannakidou
and
Monika
Rathert
,
321
343
.
Oxford
:
Oxford University Press
.
Harley,
Heidi
.
2011
.
On the identity of roots
. .
Harris,
James
.
1991
.
The exponence of gender in Spanish
.
Linguistic Inquiry
22
:
27
62
.
Kratzer,
Angelika
.
1996
.
Severing the external argument from its verb
. In
Phrase structure and the lexicon
, ed. by
Johan
Rooryck
and
Laurie
Zaring
,
109
137
.
Dordrecht
:
Kluwer
.
Lieber,
Rochelle
.
1992
.
Deconstructing morphology
.
Chicago
:
University of Chicago Press
.
López,
Luis
.
2001
.
On the (non)complementarity of θ-theory and checking theory
.
Linguistic Inquiry
32
:
694
716
.
López,
Luis
.
2007
.
Locality and the architecture of syntactic dependencies
.
London
:
Palgrave-Macmillan
.
Malchukov,
Andrej
.
2004
.
Nominalization/verbalization: Constraining a typology of transcategorial operations
.
Munich
:
Lincom
.
Marantz,
Alec
.
1993
.
Implications of asymmetries in double object constructions
. In
Theoretical aspects of Bantu grammar
, ed. by
Sam
Mchombo
,
113
150
.
Stanford, CA
:
CSLI Publications
.
Marantz,
Alec
.
1997
.
No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon
. In
Proceedings of the 21st Annual Penn Linguistics Colloquium
, ed. by
Alexis
Dimitriadis
,
Laura
Siegel
,
Clarissa
Surek-Clark
, and
Alexander
Williams
,
201
225
.
Pennsylvania Working Papers in Linguistics 4.2. Philadelphia
:
University of Pennsylvania, Penn Linguistics Club
.
Marantz,
Alec
.
2001
.
Word
.
Ms., MIT, Cambridge, MA
.
Marvin,
Tatjana
.
2003
.
Topics in the stress and syntax of words
.
Doctoral dissertation, MIT, Cambridge, MA
.
Mateu,
Jaume
.
2002
.
Argument structure: Relational construal at the syntax-semantics interface
.
Doctoral dissertation, Universitat Autònoma de Barcelona
.
Mateu,
Jaume
.
2008
.
On the l-syntax of directionality/resultativity: The case of Germanic preverbs
. In
Syntax and semantics of spatial P
, ed. by
Anna
Asbury
,
Jakub
Dotlačil
,
Berit
Gehrke
, and
Rick
Nouwen
,
221
250
.
Amsterdam
:
John Benjamins
.
Mathieu,
Eric
.
2009
.
Noun incorporation and word formation via phrase movement
.
Ms., University of Ottawa
.
Matushansky,
Ora
.
2006
.
Head movement in linguistic theory
.
Linguistic Inquiry
37
:
69
110
.
Mithun,
Marianne
.
1984
.
The evolution of noun incorporation
.
Language
60
:
847
894
.
Mithun,
Marianne
.
1986
.
On the nature of noun incorporation
.
Language
62
:
32
37
.
Newmeyer,
Frederick
.
2009
.
Current challenges to the Lexicalist Hypothesis
. In
Time and again
, ed. by
William D.
Lewis
,
Simin
Karimi
,
Heidi
Harley
, and
Scott O.
Farrar
,
91
117
.
Amsterdam
:
John Benjamins
.
Oltra-Massuet,
Isabel
.
1999
.
On the notion of theme vowel: A new approach to Catalan morphology
.
MA thesis, MIT, Cambridge, MA
.
Oltra-Massuet,
Isabel
, and
Karlos
Arregi
.
2005
.
Stress by structure in Spanish
.
Linguistic Inquiry
36
:
43
84
.
Pérez-Jiménez,
Isabel
.
2006
.
Las cláusulas absolutas
.
Madrid
:
Visor
.
Pomino,
Natascha
.
2008
.
Spanische Verbalflexion
.
Tübingen
:
Niemeyer
.
Preminger,
Omer
.
2011
.
Agreement as a fallible operation
.
Doctoral dissertation, MIT, Cambridge, MA
.
Pylkkänen,
Liina
.
2008
.
Introducing arguments
.
Cambridge, MA
:
MIT Press
.
Ramchand,
Gillian
.
1997
.
Aspect and predication: The semantics of argument structure
.
Oxford
:
Oxford University Press
.
Roberts,
Ian
.
2010
.
Agreement and head movement
.
Cambridge, MA
:
MIT Press
.
Roeper,
Tom
.
2005
.
Chomsky’s Remarks and the transformationalist hypothesis
. In
Handbook of word-formation
, ed. by
Pavol
Štekauer
and
Rochelle
Lieber
,
125
146
.
Dordrecht
:
Springer
.
Roeper,
Tom
, and
Muffy
Siegel
.
1978
.
A lexical transformation for verbal compounds
.
Linguistic Inquiry
9
:
199
260
.
Rosen,
Sarah T
.
1989
.
Two types of noun incorporation: A lexical analysis
.
Language
65
:
294
317
.
Sato,
Yosuke
.
2010
.
Complex phrase structures within morphology
.
Lingua
120
:
379
407
.
Scalise,
Sergio
, and
Emiliano
Guevara
.
2005
.
The lexicalist approach to word-formation and the notion of the lexicon
. In
Handbook of word-formation
, ed. by
Pavol
Štekauer
and
Rochelle
Lieber
,
147
187
.
Dordrecht
:
Springer
.
Scalise,
Sergio
, and
Francesca
Masini
.
2012
.
Introduction to the volume
.
Probus
24
:
1
3
.
Serrano-Dolader,
David
.
1999
.
La derivación verbal y la parasíntesis
. In
Gramática descriptiva de la lengua española
, ed. by
Ignacio
Bosque
and
Violeta
Demonte
,
4683
4757
.
Madrid
:
Espasa-Calpe
.
Smirniotopoulos,
Jane
, and
Brian
Joseph
.
1997
.
On so-called adverb incorporation in Modern Greek
. In
Greek Linguistics ’95: Proceedings of the 2nd International Conference on Greek Linguistics
, ed. by
Gaberell
Drachman
,
Angeliki
Malikouti-Drachman
,
Ioannis
Fykias
, and
Chrysoula
Klidi
,
1
:
117
128
.
Graz
:
Neugebauer Verlag
.
Spencer,
Andrew
.
1995
.
Incorporation in Chukchi
.
Language
71
:
439
489
.
Štekauer,
Pavol
, and
Rochelle
Lieber
, eds
.
2005
.
Handbook of word-formation
.
Dordrecht
:
Springer
.
Uriagereka,
Juan
.
1999
.
Multiple Spell-Out
. In
Working Minimalism
, ed. by
Samuel David
Epstein
and
Norbert
Hornstein
,
251
282
.
Cambridge, MA
:
MIT Press
.
Williams,
Edwin
.
2007
.
Dumping lexicalism
. In
The Oxford handbook of linguistic interfaces
, ed. by
Gillian
Ramchand
and
Charles
Weiss
,
353
382
.
Oxford
:
Oxford University Press