2. Spanish and Portuguese

In Spanish, ear verbs generally behave as scalar change-of-state (henceforth COS) verbs, paralleling zero or -en English counterparts, as shown in (1). They allow causative formation, as in (2a–b); passive, and middle alternation, as in (2c); and often external and instrumental cause, as in (2d). These properties are central to the definition of verbs derived from gradable adjectives as degree achievements (henceforth DA), as generally expected in colour roots (Levin Reference Levin1993: 245): as DAs, they describe events in which an individual or object undergoes change over time (Dowty Reference Dowty1979, Hay et al. Reference Hay, Kennedy and Levin1999, Kennedy and Levin Reference Kennedy, Levin, McNally and Kennedy2008).

Verbs like blanquear ‘to whiten’ further show the variable (a)telicity associated with DAs; that is, a kind of telicity sensitive to the type of scale associated with the root, with bounded-scale roots giving default telic predications (as in Hay et al. Reference Hay, Kennedy and Levin1999, but also the sense of variable telicity (as in Kearns Reference Kearns2007). As noted in the literature (McNally Reference McNally, Nouwen, van Rooij, Sauerland and Walter Schmitz2011 and Clapp Reference Clapp2012, for analytic and experimental data, respectively; M. Rasia Reference M. Rasia2017, for Romance), colour roots are generally associated with bounded property scales. The DAs yielded by the Spanish suffix -ear on chromatic-denoting roots produce, by default, telic predications allowing resultativity, telicity indicators, endpoint, maximality, completion (3a), and adjectival passives (3b). Thus, they support the idea that the aspectual properties of deadjectival DAs follow from the scalar structure lexicalized in the root (Hay et al. Reference Hay, Kennedy and Levin1999, Winter 2006, Kennedy and Levin Reference Kennedy, Levin, McNally and Kennedy2008).Footnote ²

Brazilian Portuguese derivational grammars (Rio-Torto et al. Reference Rio-Torto, Soares-Rodrigues, Pereira, Pereira and Ribeiro2016) describe amarele(j)ar, negrejar, verde(j)ar as causative-transitive resultative verbs as well. Verbs like branque(j)ar resemble Spanish in this sense (e.g., Está encardido, tem de esfregar muito para branquear ‘It's filthy, you have to scrub hard to whiten [it]’). Yet, dictionaries and native speakers surveyed seem to associate verde(j)ar or amare(j)lar with different (stative, nonresultative) properties resembling Catalan counterparts shown in the next section.

3. Catalan

Studies on Catalan –ejar, such as Oltra and Castroviejo (Reference Oltra and Castroviejo2013), show that, unlike Spanish and Portuguese, -ejar verbs yield monadic occurrences showing stative-like properties. These authors analyze these instances as inherently caused changes of state describing nonresultative stative attributions of a property (Oltra and Castroviejo Reference Oltra and Castroviejo2013: 132). They report a systematic behaviour with -ejar forms on similar (colour) roots.Footnote ⁴

Although the behaviour advanced by Oltra and Castroviejo is evident, there seems to be a nontrivial ambiguity in certain colour-rooted verbs. According to dictionaries, corpus data and the judgment of Catalan native speakers, a telic, transitive change-of-state predication like (5a) is possible, and genuinely distinct from atelic (stative, non-resultative) monadic variant claimed for (5b). Hence, even if Oltra and Castroviejo's stative readings are natural, verbs like blanquejar ‘whiten’ in (7) – as well as rossejar ‘redden’ and others – can also yield the causative alternation and variable telicity seen in Spanish deadjectivals above. Partitive (ne) cliticization supports a common unaccusative frame for monadic guises (8); however, stative uses can be told apart from antipassive ones based on other facts, such as not allowing external causation nor middle constructions, as opposed to the monadic (transitive-alternant) variant, and as a logical consequence of their noncausative structure.

This alternative in event and argument structure realization readily accommodates otherwise puzzling behaviours like variable (in)compatibility with perfective tenses (e.g., (5a) and (24)) correlated with the distribution of antipassive morphology (se-clitic in (9), attested also by Oltra and Castroviejo).

A caveat therefore arises. Unlike what has been previoulsy assumed, several Catalan varieties could allow a nontrivial alternation in event type and meaning which is not seen in Spanish (5a–b). In Catalan, however, the alternative and the consequent structural contrast remain morphologically opaque.

4. Italian

Italian carries distinct, complex commitments. Transitive/unaccusative uses paralleling Spanish, and the Portuguese and Catalan cases, are possible. They yield eventive, change-of-state predications with variable telicity (see M. Rasia Reference M. Rasia2021).

Nonetheless, Italian -eggiare is generally limited to the production of intransitive verbs lacking all the expected properties of colour-rooted DAs. These verbs fail to license internal arguments and lack change-of-state (COS) denotation. More important, they show stat(iv)e behaviour, contrasting with the mixed eventive-stative behaviour seen in Catalan monadic variants analyzed by Oltra and Castroviejo. This has two key consequences. On the one hand, Italian -eggiare yields ‘deadjectival’ verbs that do not convey scalar change, even if they are building on gradable-property-denoting roots, as in (19). Also, they do not enter the causative alternation, given their argument structure (unergative) and aspectual type (eventless predication) (next). The eventive, causative deadjectival – that is, the semantic and syntactic type corresponding to Spanish -ear in (2) and the Catalan alternate in (5) – obtains, instead, via different morphology (zero, -ificar, -esc), hence yielding a nontrivial alternation where derivational options depending upon the choice of verbalizer correlate with maximally distinct semantic and syntactic structures.

Italian is also dinstinct in that it contributes empirically clearer results, with several widely-studied tests for split intransitivity which are partially unavailable in other Romance languages like Spanish and, to a lesser extent, Catalan (which allows partitive-cl).Footnote ⁵ As expected, these tests reveal maximally divergent patterns, with -eggiare forms showing typical unergative behaviour – namely, auxiliary (avere) distribution in (12). Argument structure alternations involving internal-argument-licensing are also disallowed (13), correlating with se-marking (or the absence thereof). This is important as the clitic si is associated with inchoativity in the causative alternation, but also with passive, middle, and reflexive structures in Romance – that is, transitivity-dependent constructions. This contrasts with Spanish (see (14)) where passive, impersonal and middle readings are natural in cliticized ear forms, unlike Italian occurrences.

As a result, Spanish homogenous patterns for different deadjectival verbs (distinct verbalizers), noted above, contrast sharply with the divergent patterns seen in Italian (14b) , where unaccusative vs. unergative behaviour generally correlates with verbalizer choice (as in (12)).

Additional tests support an unexpected – i.e., unergative – structure in Italian counterparts. As anticipated, -eggiare verbs are not often found in middle constructions (15), fail passivization (16), and, unlike in Catalan (see the constructions reported by Oltra and Castroviejo, in (9)), disallow ne-cliticization (17). Verbs formed by combining the same roots with different (zero/-esc) verbalizers are fine in these contexts, yielding the expected COS, transitive, eventive type.

The significant divergence as a result of language-specific grammatical processes – notably, the role of each verbalizer in each subsystem of verb derivation and the derivational choices available – is particularly important here, especially since the lexical base (i.e., the root) remains invariable across very different instantiations. The asymmetry gains theoretical and empirical relevance in the case of Romance, since the availability of such alternatives points to structural (i.e., VP-internal configuration) differences in languages sharing Latin-inherited derivational morphology (Cockburn Reference Cockburn2012, a.o.), and are generally seen as grammatically similar. These types of contrasts are not often considered in the literature on Romance verb formation, especially the trivial vs. crucial status of the derivational choice in Spanish and Italian illustrated in (12) to (17), although they bear direct implications on argument structure in deadjectival verbs and aspect.

Moreover, unergative deadjectival verb formation is a phenomenon rarely discussed in the traditional construction(al)ist literature. The existence of this class is a major problem for influential theories on verb formation (e.g., Hale and Keyser Reference Hale and Keyser2002, Harley Reference Harley, Erteschik-Shir and Rapoport2005, Ramchand Reference Ramchand2013), and a yet underexplored option in lexically-driven approaches (e.g., Kennedy and Levin Reference Kennedy, Levin, McNally and Kennedy2008, Levin and Rappaport Reference Levin and Rappaport2010). One of the problems is that, for example in Hale and Keyser (Reference Hale and Keyser2002) and Harley (Reference Harley, Erteschik-Shir and Rapoport2005), argument structure follows from the category (on a structural definition) of the base. Ideally, transitive-unaccusative COS verbs like redden result from incorporating an ‘adjectival’ root at the base of the VP configuration, while unergative configurations should obtain from noun-incorporation instead. By systematically producing unergative verbs from ‘adjectival’ roots, Italian forms raise a pressing theoretical question in this respect.

There are also serious implications concerning semantic accounts. Namely, Italian could challenge the assumption that in deadjectival verbs, roots associated with gradable bounded scales yield inherently telic verbs (Hay et al. Reference Hay, Kennedy and Levin1999, a.o.), which is the pattern reflected by Spanish variants (see (3)). If, for deadjectival verbs, (a)telicity (and, presumably, resultativity) is expected to follow as a direct result of the type of scale lexicalized in the root (especially with colour-denoting bases (see Levin Reference Levin1993, and McNally Reference McNally, Nouwen, van Rooij, Sauerland and Walter Schmitz2011 on English), one could expect deadjectivals sharing a lexical root to show similar aspectual properties. Yet Italian breaks up this relation by yielding stative, eventless patterns: -eggiare forms are not felicitous with endpoint adverbials.Footnote ⁶ Similarly, the for-x-time modification, if admitted (i.e., if timespan adverbials do occur at all with -eggiare), yields the reading expected in pure statives; that is, a temporal frame on a state homogeneously extending for x time, thus indicating for how long the property holds of the referent, as in (18).

Unlike Catalan, Portuguese, and Spanish occurrences, these Italian verbs do not necessarily describe an event(uality) (a scalar change) progressing over time (DA). Hence, degree adverbials fail to mark scalar progression (see (24b)); but only mark a degree of property statively attributed to the subject, as (19) shows. This follows from the absence of change-of-state denotation proposed.Footnote ⁷ Similarly, in combining with verbs like stop but not with culminative/resultative verbs like finish (see Rothstein Reference Rothstein2004, a.o.), Italian variants contribute further stative patterns (see Dowty Reference Dowty1979).

Thus, Italian variants contrast with the traditional characterization of (colour-based) deadjectival verbs (see Levin and Rappaport Reference Levin and Rappaport2010). The patterns shown by Italian are of further interest to studies on stativity (e.g., Maienborn Reference Maienborn, Comorowsky and von Heusinger2008, Rothmayr Reference Rothmayr2009), as they offer further evidence of a finer-grained distinction between two distinct (stative) types of predication, illustrated in the next section.

5. Italian and Catalan: Distinct stative types

In Catalan, stative -ejar uses license three types of elements: (a) locative/temporal, (b) manner and (c) progression adverbials (22). This suggests not only a certain eventive status but also a sense of incremental scalarity, shown, namely, by the value difference in yellowness returned by the inceptive adverbial in (24a). This value difference is crucial as it replaces the notion of change over the course of an event in order to capture DA properties and different facts more accurately (Deo et al. Reference Deo, Francez, Koontz-Garboden and Snider2013). These, along with internal-argument-licensing, are two properties crucially missing from Italian equivalents.

Italian counterparts are odd in progressive tenses (#La camicia stava gialleggiando (√yellow-ear.ger) ‘The shirt was looking yellow(ish)’) and disallow location in space, manner, or progression adverbials, as in (23) and (19b)). Moreover, they show the expected behaviour in pure states (Rothmayr Reference Rothmayr2009: 29, a.o.), as opposed to statives with mixed eventive-stative properties (known as Davidsonian Statives), which allow such combinations, as in (22) (Maienborn Reference Maienborn, Comorowsky and von Heusinger2008, Rothmayr Reference Rothmayr2009). Mixed eventive-stative (Davidsonian) behaviour in Catalan in turn explains adjuncts like those in (21a), with entailment patterns following the one expected for eventive forms entailing change of state (21b), consistent with the sense of progression and gradability/ scalarity returning value difference in (21c). Crucially, the pattern in (21a) passes the tests used to identify stative-like DA occurrences (Gawron Reference Gawron2009) – that is, somehow expressing a difference in the value of this argument over a contextually given ordered domain (Deo et al. Reference Deo, Francez, Koontz-Garboden and Snider2013).Footnote ⁸

This distinction is not only consonant with Oltra and Castroviejo's description of Catalan verbs as Davidsonian statives, but it also accounts for the asymmetric distribution in (22–23). A refined view of DA and the possibility of certain eventivity in predicates with stative-like properties is key.

This contrast, as well as the alternation suggested above, would also be helpful to explain that the occurrence of Catalan forms in perfective tenses may not be as restricted as initially assumed (see (24)). Note the coherent contrast in morphological and adverbial marking indicating progress or change (25). Conversely, both progressive periphrases and graduality adjuncts are odd in Italian.

The choice of a gloss (and approximate translation) building on a verb like to go or to grow in Catalan stative variants (“go.yellow”, Oltra and Castroviejo Reference Oltra and Castroviejo2013) is not trivial (recall (5), (22) and (24)); it reflects a sense of incremental scalarity supported by the distributional asymmetries in specific aspectual tests on eventivity and stat(iv)es seen above. Specifically, Catalan data seem to suggest that incrementality and scalarity in Davidsonian statives are possible, raising a problem worth bringing into the discussion. The description of the stative verb's basic denotation in terms of ‘a formal approaching relation’ (Oltra and Castroviejo Reference Oltra and Castroviejo2013: 150) is consistent with this idea. This would, however, not apply to Italian alternates, which do not seem to be generally associated with ingressive or incremental scalar values. Thus, Italian verbs systematically fail tests for scalarity and gradability (Levin and Rappaport Reference Levin and Rappaport2010), such as comparative, superlative, and degree modification, as (26) illustrates.Footnote ⁹ This crosscut is also predicted under the distinction between stative (Davidsonian vs. pure eventless) types.

Hence, two important differences arise, insofar as Italian verbs yield an eventuality (state) that may extend over a limited time span (recall (18)) but not one that evolves or progresses over time, thus neutralizing the aspectual relevance of (the scale lexically coded in) the Root as a natural structural result (no COS denotation, no internal-argument-licensing).Footnote ¹⁰ Given these syntactic and semantic conditions, Italian verbs could also challenge other assumptions; namely, that verbs building on these lexical bases (‘colour verbs’) generally involve external causation (Levin Reference Levin1993; Levin and Rappaport 1995: 95) or causative readings (Harley Reference Harley, Giannakidou and Rathert2009: 329); or, further, that they produce clear-cut result verbs (Levin and Rappaport 2014: 337) or DAs. Catalan (see Oltra and Castroviejo Reference Oltra and Castroviejo2013: 136) and Italian stative -ejar/-eggiare variants, however, are similar in that they generally resist causativization, with the necessary proviso that this holds only for stative variants, especially in light of evidence like (7) and the alternative illustrated in (5).

Finally, there is the fact that Italian verbs show unergative patterns, while Catalan counterparts, according to Oltra and Castroviejo, instead behave as unaccusatives. This, if correct, is crucial, as it means that Italian yields a maximal contrast in aspect and argument structure reflected by the choice of derivational morphology, with change-of-state and eventive-like behaviour correlating with split transitivity and clitic (se-ne) distribution. In turn, Catalan retains an(other) important relation in which scalarity and transitional (change) denotation go hand in hand with internal-argument licensing (28), and where a difference in external-argument licensing (transitive vs. unaccusative frames) correlates with full eventivity (proper COS events vs. Davidsonian states), yet somehow, it uniformly involves an affected theme. Eventive patterns like progressive and perfective tenses thus seem to depend on transitivity (dyadicity). On the one hand, Italian dramatically contrasts by giving pure state noncausative, nonresultative (changeless, scalarless) patterns correlated with unergative frames.

On the other hand, Spanish and Brazilian Portuguese -ear/-ejar data, together with the relative triviality of choice of verbalizer, point to two subsequent conclusions. First, that an important alternative in argument and event structure like the one seen in Italian is unavailable. Second, that the notion of transitivity, change-of-state denotation, and eventivity as constant properties of deadjectivals works better as a sweeping generalization for these two Romance languages. Even if Catalan nevertheless allows such properties in some instances, also allowing for an eventive vs. stative option, it still fails to reflect the alternation which is derivationally transparent and nontrivial in Italian (see (11) vs. (2)).

Finally, it may be worth noting that French, and European Portuguese, apparently allow the stative atelic alternative in unergative patterns.Footnote ¹¹ Yet, productivity is limited, in contrast to the quantitative and qualitative regular productivity noted in Italian and, to a lesser extent, in Catalan.

In conclusion, a major contribution of the Catalan vs. Italian contrast proposed in this section touches on scalarity in statives. Apparently, (incremental) scalarity is dependent on certain eventivity. This sets a clear but generally undebated contrast between stative – Davidsonian vs. pure (completely eventless, Kimian-like) – types (see Maienborn Reference Maienborn, Comorowsky and von Heusinger2008, a.o.). By describing a change in the value of a scalar property over time vs. some other (nontemporal) domain, the opposition would empirically connect with recently claimed distinctions between eventive and non-eventive flavors of scalarity and change (see Rappaport Reference Rappaport2021) and more comprehensive approaches on DA denotation.

6. Summary and conclusions

In this article, I have argued the following: First, that the derivational alternative seen in Italian defines a consistent crosscut in event and argument structure, with (non)eventivity correlating with (a)transitivity. Second, that a relatively similar structural option is either collapsed under the same morpheme (as in Catalan), or unavailable in other Romance languages (such as Spanish), in both cases rendering an opaque derivational choice. Third, by systematically producing unergative, change-of-state-less verbs from ‘adjectival’ roots (specifically, those designating properties like colours), Italian poses an empirical problem and a theoretical question not only for mainstream theories of lexical syntax and verb formation, but also for lexically-driven approaches which also fail to predict (or, at least, discuss) this possibility. If correct, Romance – and, especially, Italian – data support the need to revisit and, ultimately, refine, nontrivial syntactic and semantic generalizations on deadjectival verbs and verb formation that have been circulating in the literature for the last few decades.