4.1. Verbal morphology

We cannot rely on the morphology of English active participles to tell us much about their category.Footnote ¹¹ As already mentioned, some active participles have overt verbalizers, suggesting that they contain at least vP (see Harley Reference Harley, Rathert and Giannankidou2009, a.o.). In BCS, both finite verbs and infinitives, as well as participles (but not members of other categories), contain a so-called verbal theme vowel, which immediately follows the root. This suffix has been argued to be the exponent of v (Svenonius Reference Svenonius2004, Caha and Ziková Reference Caha and Ziková2016, Biskup Reference Biskup2019, Bešlin Reference Bešlin2023). The suffix is different for different classes of verbal stems; I illustrate the three main classes in (15). The verbal theme vowel may vary across the verbal paradigm (present tense being notoriously irregular), but the theme of the active participle is always identical to that of the infinitive.

In Hebrew, the participle consists of a root in a verbal template and a prefix (which I analyze as an adjectivizer in the following section). Setting aside the prefix (and $ \phi $ -marking) for now, note that the verbal template XaXeX found with participles (16a) is also used to form the future (16b), the infinitive (16c) and sometimes the imperative (16d). This template is not used in the derivation of simple nouns or adjectives.

What we have seen is that participles in the three languages show morphological evidence of containing verbal structure. Verbal morphology is found immediately adjacent to the root, suggesting that it attaches low in the structure. In Hebrew in particular, templates are determined based on the first categorizing morpheme (here, v), while further derivation is done with affixation, as we will see immediately in the following section.

4.2. ‘Participial’ marking

The status of the English participial suffix -ing is controversial, with most recent literature treating it as a verbal aspectual suffix. The picture is muddled even further by the fact that the suffix also appears in gerunds (e.g. John’s marrying Jane surprised me.). This has led some researchers to argue that -ing is special and that the result of its affixation may be of any category (e.g. Milsark Reference Milsark1988). However, the distributional facts in Section 5 will lead us to the conclusion that -ing affixation never produces verbs. In addition to attaching to certain roots (e.g. cunning, grueling, fleeting), -ing may attach to verbal stems (e.g. electrifying, energizing, jumping), producing adjectives in both cases. I will therefore conclude that -ing is an exponent of the adjectivizing morpheme (a).Footnote ¹²

For Hebrew, Meltzer-Asscher (Reference Meltzer-Asscher2010, Reference Meltzer-Asscher2011) argues that active participles ‘appear in a morphological form identical to that of verbs in the present tense, in any one of the five non-passive verbal templates of the language (XoXeX, niXXaX, meXaXeX, maXXiX, and mitXaXeX)’ (Meltzer-Asscher Reference Meltzer-Asscher2010: 2212).Footnote ¹³ However, she also claims that, unlike the present tense form (17a), the active participle in (17b) is actually uninflected for Tense (and it instead receives temporal interpretation from the main verb), as in English.

A likely explanation for the identity of the active participle and the ‘present tense’ form in Hebrew is that the ‘present tense’ form in this language is also a participle – a deverbal adjective on my account.Footnote ¹⁴ On Meltzer-Asscher (Reference Meltzer-Asscher2010)’s account, present tense verbs, ‘verbal’ participles and adjectival participles all belong to distinct categories, yet they all share identical morphology. This identity has to be treated as a complete accident, despite the fact that Hebrew generally has very few instances of zero-derivation. On my account, all three forms are deverbal adjectives, which is why they share the same morphology. As we have seen, the verbal template found on participles is accompanied by a prefix, except in the case of XoXeX; cf. (16a). I analyze this prefix as the exponent of the adjectivizer (a).

Note that in order to obtain present interpretations, uncontroversially nonverbal predicates must also appear in the form subject+predicate with no intervening copula (18). This well-known fact shows that Hebrew matrix clauses need not contain a(n overt) verb, so its absence in (17a) cannot be taken as evidence for the verbal status of kotevet. Furthermore, both the participial and the nominal predicate require the same copula for past interpretations (19). If kotevet in (17a) were a true present tense form, (19a) would involve the addition of a past marker to an overtly marked present tense, a typologically unattested pattern.

Overall, the morphological facts from Hebrew do not support the conclusion that the active participle in this language is verbal. On the contrary, the data suggest that this participle patterns with nonverbal predicates. The distributional facts in Section 5 will allow us to pinpoint the category of this nonverbal element as an adjective.

In BCS, I focus on the active participle known in the Slavic literature as the l-participle (see Bešlin Reference Bešlin2023 for a detailed analysis of BCS passive participles).Footnote ¹⁵ Historically, the l-participle was used to express so-called retrospective or resultative meanings (see Migdalski Reference Migdalski2006: Chapter 1). In contemporary BCS, it is difficult to delineate the meaning of the l-participle itself, since it can be used in a number of different contexts, giving rise to meanings as diverse as (active) simple past, present and past perfect, resultative and future II (used in embedded and conditional contexts); see Migdalski (Reference Migdalski2006). In fact, the distinct meanings can largely be attributed to the (independent) aspectual makeup of the copula (biti ‘be’) and the participle; see Migdalski (Reference Migdalski2006) and Todorović (Reference Todorović2016) for thorough analyses. I detail the distribution of the l-participle in Section 5; for now, it suffices to note that it is used attributively (20a) and as a complement of the copula in predicative position (21a). In addition to the verbal theme vowel, the l-participle contains a suffix that appears on some simple adjectives: -o for masculine singular (20b) and -l for all other gender/number combinations (21b). Some other examples of such simple adjectives include kiseo ‘sour’, vreo ‘hot’ and okrugao ‘circular’.

The marker -o/-l seems to be an exponent of the adjectivizing morpheme (a); note that this is not a $ \phi $ -feature marker – I will turn to $ \phi $ -marking next. Before I do that, let me note that the adjectival suffix -l is no longer productive in BCS, except with verbal bases. Furthermore, some simple adjectives derived with the suffix -l likely date back to Proto-Slavic – for example, *kỳslъ ‘sour’ from *kys- $ + $ *-lъ, *-lъ being the participial/adjectival suffix equivalent to the BCS -l (Derksen Reference Derksen2008). However, decomposition into a root and an adjectivizing suffix is still supported in modern BCS. We can contrast adjectives like bel-(a) ‘white-(f.sg’), which were monomorphemic already in Proto-Slavic, with adjectives like kise-l-(a) ‘sour-a-(f.sg’). While derivation involving bel- in BCS always includes the ‘adjectival suffix’ (now part of the root) as in beliti ‘make white’, the root of decomposable adjectives in -l can appear independently (e.g. kisiti ‘to taste sour’, vreti ‘to boil’, zreti ‘to mature’ (not, for example, *zreliti)).

4.3. Φ-marking

As expected, the morphology of English is not particularly telling when it comes to $ \phi $ -marking. However, let me mention that active participles in closely related German have the same agreement (concord) properties as simple adjectives; namely, the participle inherits the $ \phi $ -features of the noun in the attributive position, and it is uninflected in the predicative position, as seen in (22)–(23), adapted from Haiden (Reference Haiden, Fernández and Albizu2001: 195).

As already shown, BCS l-participles can also appear in the attributive or predicative position (24). The l-participle inflects for case, number and gender the exact same way an adjective does in both of these positions; cf. (25). In both cases, the $ \phi $ -marking of the adjective/participle is entirely dependent on the formal features of the noun it is associated with: the head noun being modified in (24a)–(25a), and the subject noun in (24b)–(25b). Note further that a restriction exists on so-called ‘long forms’, in that they are only available in the attributive position; the ‘short form’ is available in both positions, and the pattern is exactly the same for participles and adjectives. In BCS, this morphological distinction is correlated with a meaning contrast between specific and non-specific NPs (Aljovic Reference Aljovic2002).Footnote ¹⁶

In Hebrew, simple attributive adjectives inflect for gender, number, and definiteness, while predicative adjectives inflect only for gender and number (26). The pattern is exactly the same for active participles (27). In this way, Hebrew participles differ from verbs, which index person, in addition to gender and number (28). Analyzing participles as adjectives immediately explains their pattern of agreement.

We have seen that, while adjectives have language-specific morphological patterns, active participles follow these patterns perfectly. In what follows, I will show that BCS, Hebrew and English active participles have the syntactic distribution of adjectives.

5.1. Copula selection

Without going into too much detail for reasons of space, let me note that both adjectives and active participles appear with a copula in predicative position. They therefore pattern with nouns and PPs, but different from other finite and non-finite verbs in the languages under consideration. ‘Be’ and its equivalents in other languages have often been called auxiliaries when they appear with participles, and copulas when they appear with adjectives; however, Becker (Reference Becker2000) shows that they are morphosyntactically identical in English. The same is true of BCS. As shown in Section 6.7.3, certain participles cannot appear with the future copula in Hebrew; comparing the future-copula construction and the synthetic future tense, I will argue that the future copula imposes a semantic restriction on its complement. The inability of certain participles to appear as complements of the future copula will be shown to be a consequence of this semantic restriction.

5.2. Depictives

Let us now look at depictive constructions: constructions that predicate a property of a DP (external or internal argument) that holds throughout the event denoted by the matrix predicate. English depictives can be encoded as root adjectives, participles (active or passive), PPs or DPs (29a); crucially, they cannot be verbal elements, be it infinitives or tensed forms (29b). We observe the same pattern in BCS (30). Hebrew does not have depictives; see Schultze-Berndt and Himmelmann (Reference Schultze-Berndt and Himmelmann2004) for a cross-linguistic overview.Footnote ¹⁷

Finite verbs include Tense, so they are categorially different from participles even on a verbal analysis of the latter. Influential analyses of depictives treat them as small clauses (e.g. Rothstein Reference Rothstein1983), and small clauses never include Tense. What is more difficult is explaining why participles are acceptable in depictive constructions, but (bare) infinitives are not. Related to this point, it is not clear on a verbal analysis of participles why participles do not combine with modals, but infinitives do. While it is possible to describe these contrasts in technical terms (e.g. by stipulating that the participle contains some additional feature), this seems to be unnecessary and uninformative. Considering all the facts presented in this paper, the simplest account is one where the outward-most layer of participles is adjectival, allowing them to appear in positions available to adjectives, but unavailable to verbs.

5.3. Reduced temporal clauses

English adjectives and participles (both active and passive) may occur in what I will call a reduced temporal clause, illustrated in (31a–c). Crucially, the infinitive cannot appear in this construction (31d). At least for English, we can use this test to further show that the distribution of participles mirrors that of adjectives, and not verbs.

Before moving on, I should mention that this test is inconclusive when applied to BCS and Hebrew because none of the equivalents of sentences in (31) are possible, for reasons that are poorly understood.

5.4. Attributive position

Participles also appear in positions that are otherwise only occupied by adjectives. As shown in (20a) and (24a), BCS active participles can act as prenominal modifiers. In addition to passive participles (which are adjectives; see Bešlin Reference Bešlin2023) and simple adjectives, active participles are the only element that can appear in this position in BCS. Meltzer-Asscher (Reference Meltzer-Asscher2010) acknowledges that even the active participles that fail her other diagnostics for ‘adjectivehood’ appear in the attributive position in both English and Hebrew (32).

Meltzer-Asscher (Reference Meltzer-Asscher2010) also incorrectly claims that participles derived from eventive transitive verbs cannot appear in the prenominal position. In fact, it is simply the case that participles derived from transitive verbs need to overtly express their internal argument (see Borer Reference Borer1990). The internal argument cannot follow the attributive participle because English obeys the Head-Final Filter, a generalization that attributive modifiers need to appear adjacent to the noun they modify (Williams Reference Williams1982). This is true both of complements of simple adjectives and of participles (33). English is able to work around the Head-Final Filter by incorporating the object into the participle (34a–b), while some other languages can work around it by expressing fully case-marked internal arguments to the left of the attributive participle, as in Dutch (35).

The claim that the prenominal position in languages like English is occupied only by adjectives has been challenged, but the arguments do not stand up to scrutiny. Sleeman (Reference Sleeman2011) argues that participial modifiers contain verbal structure, but gives no evidence that they are not (ultimately) adjectives. As already noted, the fact that prenominal participles contain verbal structure is problematic for the adjectival hypothesis only if one has lexicalist assumptions. Both Sleeman (Reference Sleeman2011) and Laskova (Reference Laskova2007) also assume that being eventive equals being a verb, and conclude from the possibility of eventive interpretations in cases like (32a) that the prenominal position can be occupied by verbs. However, we saw at the beginning of this article that it is untenable to equate eventivity with verbhood and stativity with adjectivehood. Moreover, authors who accept this position must explain the absence of infinitives in the prenominal position. Once we accept that interpretation cannot determine category, the hypothesis that the prenominal position in languages like English is occupied only by adjectives is conceptually sound again.

Contrary to common belief, the prenominal position in English can accommodate some PPs in addition to adjectives (36). In this position, there is an interesting distributional contrast between adjectives and participles on the one hand, and PPs on the other. As already mentioned, both prenominal participles (36a) and simple adjectives (36b) in English have to obey the Head-Final Filter. On the other hand, PPs are not subject to the same restriction (36c). This is another instance where participles show the same distribution as simple adjectives, suggesting that syntax does not discriminate between the two based on their category.

We have seen that attributive participles pattern with adjectives on two counts: (i) adjectives, but not verbs, can be attributive modifiers, and (ii) adjectives and participles, but not PPs, obey the Head-Final Filter in English. This further strengthens the conclusion that all active participles in the languages under consideration, including those derived from eventive verbs, are adjectives.

5.5. It-clefts

English adjectives and -ing participles are both incompatible with the cleft focus position (37a–b); see Emonds (Reference Emonds, Georgopoulos and Ishihara1991: 97). This is in contrast to infinitives, which appear in this position quite freely (37c).Footnote ¹⁸ The data in (37) provide clear evidence that the distribution of participles mirrors that of adjectives, and not of verbs.

Moreover, Emonds (Reference Emonds, Georgopoulos and Ishihara1991) observes that those dialects of English that allow adjectives in the focus position of a cleft also present participle phrases in that position. In some varieties of Irish English, sentences like (38a) are grammatical. In these dialects, (38b) is also grammatical.

The data we have just seen shows that the distribution of active participles follows that of simple adjectives; where there are dialectal differences in distributional possibilities, the participle still patterns with the uncontroversial adjective. Since distribution is largely determined by the category of an item, I conclude from this that the external syntax of these two elements is identical; namely, they are both adjectives.

5.6. C-selection below the word level

Finally, I discuss the selectional restrictions of the BCS nominal suffix -ic-, broadly ‘one who is X $ {}_{\underline{a}} $ ’ (Babić Reference Babić2002: 565). Even though the present discussion is concerned with elements below the ‘word’ level, I include it in the section on distribution because it pertains to a prime example of c-selection. Namely, the BCS suffix -ic- can select for adjectival input, including participles, but it cannot select for verbs. We can observe examples where -ic- attaches to simple adjectives (39) and active participles (40) in -l, and to simple adjectives (41) and passive participles (42) in -n. In (43), I provide a couple of examples to illustrate the general pattern – namely, that infinitives cannot serve as input to -ic-affixation.

All things equal, if active participles are (deverbal) adjectives, we expect them to be able to serve as input to affixation anywhere that a simple BCS l-adjective can. While this issue requires further investigation, (39)–(43) shows that participles behave the same way as adjectives (but not verbs) in this domain, thus supporting the hypothesis that they are adjectival.

6.1. DP-complements

Bennis and Wehrmann (Reference Bennis, Wehrmann, Bok-Bennema and Coopmans1990) argues that English active participles are verbs because they can have accusative-marked DP complements (44a), while prototypical adjectives cannot (44b). Meltzer-Asscher (Reference Meltzer-Asscher2010) shows that the same contrast obtains in Hebrew (45). A similar pattern obtains in BCS. Namely, participles can have accusative-marked complements just like verbs; simple adjectives can have genitive- but not accusative-marked nominal complements (46).

The conclusion that this makes the participles verbs is warranted only on a lexicalist approach, where ‘being an adjective’ entails having no verbal syntactic structure. On a syntactic approach to word formation, the participle has an accusative-marked complement if it contains the portion of verbal structure that is responsible for licensing it (VoiceP). This does not preclude the claim that the participle is externally adjectival; recall (13). In a similar vein, Bešlin (Reference Bešlin2023) shows that English passive participles derived from ditransitive verbs can have DP complements in unambiguously adjectival positions (47). This should be impossible on a lexicalist account where the adjectival participle is essentially a simple adjective for the purposes of the syntax. On a syntactic account, the pattern can be easily accommodated: the participle in (47) is an adjective which embeds the portion of the verbal structure that licenses the oblique argument.

More generally, the argument that DP complements diagnose verbhood does not stand up to scrutiny given the well-known case of gerunds (48a). The (ultimately) nominal status of gerunds has seldom been questioned, and yet they appear with accusative-marked DP complements, while simple nouns cannot (48b). On syntactic approaches to word formation, this is accounted for by positing a full-fledged VoiceP below the nominal structure.

Taken together, these facts show that having an accusative-marked DP complement – while suggestive of the presence of VoiceP – does not entail that the element in question has the clausal distribution of a verb.

6.2. Word order with modifiers

Laskova (Reference Laskova2007) notes that English eventive passive participles pattern with verbs in that they allow post-modification by adverbs (49a–b). She contrasts this with the behavior of participles that denote a state resulting from an event: resultative participles (Nedjalkov and Jaxontov Reference Nedjalkov, Jaxontov, Nedjakov and Comrie1988, Embick Reference Embick2004). Unlike eventive passive participles and verbs, resultative participles do not allow post-modification by adverbs (49c). Based on this, Laskova concludes that English eventive passive participles are verbs. Building on this work, Meltzer-Asscher (Reference Meltzer-Asscher2010) argues that English -ing participles must necessarily be verbs because they are readily postmodified by adverbs (50).

Bešlin (Reference Bešlin2023) argues that adverbial post-modification in (49c) is ungrammatical because the English resultative participle lacks VoiceP; thus, there is not enough verbal structure for the verbal stem to move past the adverb to Voice. She shows that the movement generally happens by pointing to examples like (51), where the selectional relation between rely and on is disrupted on the surface because the verb has moved.

Therefore, all that needs to be said for (50) is that the verbal structure of active participles is not impoverished in a relevant way when compared to the finite verb or the eventive passive participle. In other words, active participles in (50) include VoiceP, which allows the verbal stem to move past the adverb. This seems correct given that VoiceP hosts thematic agents, and John in (50) is the thematic agent of the event of jumping, denoted by the -ing participle. Therefore, the argument for the category contrast between resultative participles and other participles dissipates. Note that I have provided an analysis, rather than a mere observation of the relevant patterns, in contrast to Laskova (Reference Laskova2007) and Meltzer-Asscher (Reference Meltzer-Asscher2010). My analysis of this data makes no claims about the categorial status of the relevant elements (i.e. their external syntax) and is compatible with the idea that all participles have the external syntax of adjectives.

If active participles project both v and Voice, we predict that they should also license verbal projections which are located above v but below Voice. This includes high applicatives, if they are otherwise available in the language in question (see, for example, Harley Reference Harley2013, Reference Harley, D’Alessandro, Franco and Gallego2017). Accordingly, high applicatives are indeed possible with BCS active participles (52); the structure is illustrated in (53).Footnote ¹⁹ While a closer investigation of this prediction is necessary, I am not aware of any counterexamples.

What happens with modifiers in the other languages? BCS has rampant scrambling; both pre- and post-modification is possible regardless of the category of the modified element. For Hebrew, I am unaware of previous attempts to use modifier placement to diagnose a category contrast between verbs and adjectives. However, my Hebrew consultants report judgments that support my analysis and argue against the analysis that some Hebrew active participles are verbs. In particular, PP modifiers with active participles show the same placement possibilities as adjectives, and not as finite or non-finite verbs. PP modifiers that are available with simple adjectives in Hebrew are always post-modificational (54). For finite and non-finite verbs, both pre- and post-modification are possible (55).Footnote ²⁰ Crucially, the order PP-verb is available, while the order PP-adjective is not. For active participles, only post-modification is available, like with simple adjectives (56).

We have shown that the English modification pattern is independent of the question of category and compatible with the claim that all participles are adjectival. In Hebrew, participles are modified in the same positions as adjectives and not verbs, as expected on an adjectival analysis.

6.3. Phasal verbs

In Meltzer-Asscher (Reference Meltzer-Asscher2010), Emonds (Reference Emonds, Georgopoulos and Ishihara1991) is cited for the claim that phasal verbs (keep, resume, cease) take only verbal, but not adjectival complements. In fact, this is not what is stated in the original paper; Emonds (Reference Emonds, Georgopoulos and Ishihara1991)’s claim is that these verbs select elements with a [+V] feature, regardless of their external syntactic structure (Emonds Reference Emonds, Georgopoulos and Ishihara1991: 99–100). Nevertheless, let us evaluate Meltzer-Asscher (Reference Meltzer-Asscher2010)’s argument at face value. The idea is that (57) demonstrates that phasal verbs only take verbal, but not adjectival complements, and thus that participles (of intransitive verbs) in (58) must also be verbs.

A general point about this diagnostic is that it is not precise to say that the complement of these verbs ‘must be a verb phrase’; in fact, these verbs specifically require -ing complements, and no other verb form can take their place (cf. *keep runs/ran/(to) run).Footnote ²¹ Since the category of the participle is what is at issue, we cannot use this as a diagnostic for categorial status. Furthermore, it is not quite true that these verbs never combine with adjectives; for example, keep can have adjectival complements, as in keep calm, keep busy, keep close, etc.

However, it is true that resume and cease cannot take any (root-derived) adjectives as complements. If participles are (deverbal) adjectives, we still have to explain why -ing adjectives are allowed as complements of these verbs in (57a)–(58), whereas simple adjectives are not (57b). I would like to suggest that the -ing forms with cease and resume are, in fact, not participles at all, but rather nominal phrases (gerunds). The first reason to believe this is that these verbs do actually take simple nominals as their complements, as in (59a). Furthermore, (58) can be expanded to include the nominal possessor his with no change in meaning (59b), suggesting that the -ing form in (58) may be nominal as well. An additional argument for the nominal status of the -ing form in the complement of cease/resume comes from the fact that it can be coordinated with uncontroversial DPs, as seen in (60).Footnote ²²

Of course, complements of cease/resume can be modified by adverbs, as in (61a), which may be taken as evidence for their verbal (or adjectival) status. However, I take the bracketed constituent in (61a) to have essentially the same structure as (61b), which is a nominalized VoiceP (see Kratzer Reference Kratzer, Rooryk and Zaring1996) that can appear in unambiguously nominal positions, as in (61c).

Even more compellingly, we can provide positive evidence that the -ing complements of cease and resume are not adjectival. It has long been noted that very modifies adjectives (though not all adjectives, see below) but not members of other categories (e.g. Brekke Reference Brekke1988, Emonds Reference Emonds, Georgopoulos and Ishihara1991, Meltzer-Asscher Reference Meltzer-Asscher2010). Moreover, some active participles can be modified by very, showing that they are uncontroversially adjectival (62). Now compare (62) to (63), where the AP very thriving is the complement of the phasal verb–the result is ungrammatical. The ungrammaticality of (63) strongly indicates that the -ing complement of cease and resume is not adjectival, thus explaining why simple adjectives cannot appear in this position.

Summing up, the complement-of-keep/cease/resume diagnostic cannot be used to determine verbhood (in English) because (i) no verb form other than the -ing form, whose category is in question, can appear in this position, (ii) some adjectives can appear as complements of keep, and (iii) cease and resume take gerundive, not participial, -ing complements. I have shown that participles and simple adjectives are equally impossible as complements of the verb cease. This is a distributional pattern that sets both apart from infinitives, lending further support to the idea that participles are adjectives.

In BCS, we again see participles patterning with adjectives and not with verbs. Namely, BCS phasal verbs never take participial or adjectival complements (64a). Instead, they can take finite and infinitival verbal complements (64b), in addition to PPs (64c), and bare nominal complements (64d).

We have seen that complements of phasal verbs do not provide a suitable diagnostic for verbhood in English. However, phasal verbs in BCS can take both finite or infinitival verbal complements, but not participial or adjectival complements. Once again, participles can be shown to have the distribution of adjectives and not verbs. In Hebrew, phasal verbs always take infinitival complements, making the test inapplicable.

6.4. Adverbial affixation

In English, the suffix -ly attaches to adjectives to produce adverbs (65a). A number of authors have observed that only some active participles serve as input to -ly suffixation, taking this to indicate that only certain active participles can be adjectival in addition to being verbal (Fabb Reference Fabb1984, Brekke Reference Brekke1988, Meltzer-Asscher Reference Meltzer-Asscher2010, a.o.). Meltzer-Asscher (Reference Meltzer-Asscher2010: 2215) gives the lists in (65b–c) to illustrate the contrast. However, we should first of all recognize that not all simple adjectives serve as input to -ly suffixation either, (65d); therefore, an element that fails to combine with -ly may still be adjectival.

Perhaps even more damaging to the view that the different behavior of the participles in (65b–c) stems from a category contrast is the following: participles that allow -ly suffixation are not necessarily the same participles that appear in other ‘adjectival’ contexts. For example, glowingly, cryingly and jumpingly are well-formed adverbs according to the Merriam-Webster dictionary (contra Meltzer-Asscher Reference Meltzer-Asscher2010), but the underlying participles cannot appear, for example, as complements of seem (e.g. *The girl seemed jumping / crying / glowing). If the category of the participle is supposed to account for both of these facts, we encounter a paradox. On the account developed here, all participles are (deverbal) adjectives. The reason that some participles cannot appear as bare complements of seem has to do with their meaning, not their category, as discussed in Section 6.7.2. While I am not able to provide a definitive explanation for the contrast between cryingly and *walkingly, some such contrasts may also be explained by appealing to meaning. The paraphrase in a walking manner sounds very odd, while in a crying manner is acceptable, possibly because one does not quite know what doing something ‘in a walking manner’ would mean. However, participles describing ways of walking are quite productive in this construction (e.g. in a limping/stumbling/strutting manner), and the difference between walking and limping is unlikely to be one of category. Regardless, the contrast between cryingly and limpingly on the one hand, and *walkingly on the other, shows that the relevant restriction on -ly-affixation is distinct from the restriction on the complement of seem. The restriction on -ly affixation is better understood, at least partly, as a constraint on which participles/adjectives can describe ways of doing things ‘in an X manner’. Furthermore, if we concede that -ly attaches only to adjectives, as is standard in the literature, this means that crying and limping are (eventive) adjectives, contra Fabb (Reference Fabb1984), Brekke (Reference Brekke1988), Meltzer-Asscher (Reference Meltzer-Asscher2010). The existence of adjectives that denote events goes against the idea that adjectival participles must be stative and that they are formed only from stative verbs (contra Meltzer-Asscher Reference Meltzer-Asscher2010).

According to Meltzer-Asscher (Reference Meltzer-Asscher2010), Hebrew has a pattern similar to English -ly affixation, where ‘adverbs can be formed periphrastically using be-ofen Adj (‘in a Adj manner’)’ (Meltzer-Asscher Reference Meltzer-Asscher2010: 2215). Again, some participles can serve as input to be-ofen, while others cannot (66a–b), leading Meltzer-Asscher (Reference Meltzer-Asscher2010) to conclude that only the participles in (66a) are adjectival. The first thing to notice is that, again, not all adjectives can appear with be-ofen (66c), so the unacceptability of (66b) does not provide convincing evidence for their non-adjectival status.

Additionally, be-ofen xasar-xaim ‘in-manner missing-life’ in (66c) is fine if xasar-xaim is interpreted metaphorically to mean ‘lifeless’, but not if it is interpreted literally as ‘dead’. This further supports the idea that a problem may arise not because of an item’s category, but because of its lexical meaning. Simply put, it is difficult to discern what exactly ‘in a dead manner’ is supposed to convey. I therefore take that the expression of manner in Hebrew is constrained by lexical meaning, in addition to the restriction on syntactic category; it is the lexical meaning, not category membership, that drives the contrast in (66).

6.5. Negative un-

Negative un- attaches to adjectives, but not to verbs. It also attaches to some, but not all, -ing participles, as seen in (67) from Meltzer-Asscher (Reference Meltzer-Asscher2010: 2216). From this contrast, Meltzer-Asscher (Reference Meltzer-Asscher2010) concludes that only the participles in (67a) are adjectives. She does acknowledge that un- cannot attach to all adjectives; for example, adjectives like unsmart and ungood are ill-formed, and the reasons for this ill-formedness are unclear.Footnote ²³ This means that the failure of an element to combine with un- does not rule out its classification as an adjective.

We can use the prefix non-, which also attaches to adjectives (and nouns), but not verbs, to show that the contrast between (67a) and (67b) is not one between adjectives and verbs. For example, non-suspecting is possible alongside unsuspecting, and non-jumping (exercises) and non-chewing (diet) are also good, despite these participles’ incompatibility with un-. This provides positive evidence that (at least some of) the purportedly verbal participles in (67b) are also adjectives. Moreover, not all adjectives in (67b) are bad; for example, an uncrying baby or the nucleus of ungrowing cells are attested and acceptable. This is relevant because Meltzer-Asscher (Reference Meltzer-Asscher2010)’s account depends on the idea that all of her diagnostics show a split between the same two groups of participles, which we see is clearly not the case (cf. ungrowing and non-growing, but *growingly and *very growing).

6.6. Coordination

Meltzer-Asscher (Reference Meltzer-Asscher2010) argues that it is impossible to coordinate some active participles with simple adjectives; the judgments in (68) are reported as they appear in Meltzer-Asscher (Reference Meltzer-Asscher2010: 2217). From the purported unacceptability of these coordinated phrases, combined with the view that identity of category is a sufficient (though not necessary) condition for coordination, she concludes that the -ing participles in (68) cannot be adjectives.

The first thing to note is that, while the above examples may be somewhat odd, they are not unacceptable, especially when compared to, for example, *a rude and jump(s) boy, which is judged as emphatically bad. Note that we would have no explanation for this contrast in acceptability on the view that both jumping and jump(s) are verbs. In fact, if jumping were a verb, it is not clear how one would account for the contrast between a jumping boy and *a jump(s) boy, even in cases that do not involve coordination.

Furthermore, we can identify several factors that conspire to make (68) sound odd, none of which have to do with category. First off, the two attributes in (68a) stand in opposition, so using and is a strange way to connect them. Likewise, (69a) is strange compared to (69b), although both examples include coordinated simple adjectives; (69c) sounds much better compared to the original example in (68a).

The degraded nature of (68) may also be due to a common issue in English: coordinating adjectives from different semantic categories often sounds strange, as shown by (70). There are distinct lexical-semantic classes of adjectives which appear in a hierarchy that determines their order in a complex structure (Dixon Reference Dixon1977). In English, the default strategy for attributive adjectives from different classes is to order them according to class, without an overt coordinator.Footnote ²⁴ Examples like (68)/(70) are perfectly acceptable without an overt coordinator (71). The same pattern carries over to the coordination of an -ing participle and a simple adjective (72).

This is not to say two adjectives from different classes can never be coordinated in English; see (73). The same is true for active participles and simple adjectives (74), indicating that the generalization in Meltzer-Asscher (Reference Meltzer-Asscher2010) is incorrect.Footnote ²⁵

In sum, not only does coordination fail to provide evidence for the different categorial status of adjectives and active participles, it in fact shows that they pattern exactly alike (and unlike verbs). Coordination data should therefore be taken to provide positive evidence for the adjectival status of active participles.

6.7. Denoting eventualities vs. (scalar) properties

Next, we turn to tests that have been used to argue that only certain participles are adjectives (while others are verbs), but which are in fact better suited for singling out those participles that denote (scalar) properties, rather than a difference in category. To that end, I discuss modification by very, complements of seem, and the compatibility with the future copula in Hebrew. Eventuality-denoting active participles will be shown to be banned from these positions not because they are verbal, but because the position in question requires that the element occupying it be a predicate of (scalar) properties. In other words, some constructions involving active participles turn out to be syntactically well formed but unacceptable due to a semantic clash. Since this semantic difference is relevant on any account, I will argue that the categorial distinction can be dispensed with completely.

6.7.1. Modification by very

According to a common observation, very generally modifies adjectives; in fact, Brekke (Reference Brekke1988: 169) takes modification by very as ‘the conventional test for true adjective status’ (see also Chomsky Reference Chomsky1957). From here, it has been argued that participles which are not modifiable by very are not adjectives; cf. (75)–(76) from Meltzer-Asscher (Reference Meltzer-Asscher2010: 2216). It is worth noting from the outset that not all simple adjectives can be modified by very (77), so the fact that some participles are incompatible with very cannot by itself be taken as evidence against their adjectival status.

Additionally, Borer (Reference Borer1990) shows that the compatibility of a participle with very and other degree modifiers depends on semantic factors, those that determine whether the related verb related is compatible with the degree modifier very much. In (78)–(79), from Borer (Reference Borer1990: 97–98), we see that very is compatible with a participle only if the verb it is derived from is compatible with (the degree reading of) very much. Footnote ²⁶

Even more strikingly, Borer (Reference Borer1990) observes that Hebrew me’od ‘very’ can modify both verbs and adjectives, and yet only those verbs that can be modified by me’od give rise to participles that allow me’od-modification (80)–(81). Despite me’od’s ability to modify both verbs and adjectives, the split is the same as in English, suggesting that it is not the category of the modified element that is the problem. From here, Borer (Reference Borer1990) concludes that the contrasts we observe have nothing to do with the participles’ categorial status, but rather with a meaning component that distinguishes the two types of verbs, and, by extension, the participles they give rise to.

What is this meaning component? Brekke (Reference Brekke1988) states that the relevant component cannot be gradability because even participles related to gradable verbs such as grow are incompatible with very – for example, *a very growing child. However, it is unclear what criteria Brekke (Reference Brekke1988) uses to determine that grow (or growing) is gradable. In assuming that the participle growing ought to be gradable, Brekke (Reference Brekke1988) seems be thinking that, when a thing is growing, it may change along some dimension that comes in degrees. For example, a growing child may grow in weight or height. But from this, it does not follow that the growing, an event of change, also comes in degrees.

According to Bolinger (Reference Bolinger1972), degree (or scalar) verbs are those that allow modifiers like very much to have an intensifier (degree) reading rather than a quantity (amount) reading. Only verbs like (78a) are considered to be degree verbs, and only they give rise to participles modifiable by very. Since the two classes of verbs already differ in their ability to be modified for degree, the simplest explanation for the contrast in (75)–(76) and (80b)–(81b) is that the distinction in scalarity is inherited by the participles they give rise to. The simplest explanation, then, does not motivate a category difference between the two types of participles any more than it motivates a category difference between eventive and stative verbs.Footnote ²⁷

6.7.2. Complements of seem

As argued at length in Matushansky (Reference Matushansky2002), (perceptual) seem must combine with a complement that denotes a scalar predicate of type $ \Big\langle $ d, $ \Big\langle $ e,t $ \left\rangle \right\rangle $ : a function that maps degrees to functions from individuals to truth values. However, participles that embed full-fledged VoicePs are of type $ \Big\langle $ e, $ \Big\langle $ v,t $ \left\rangle \right\rangle $ : a function that maps individuals to functions from eventualities to truth-values. Simply put, the denotations of these participles involve eventualities, and the verb seem requires a complement with a different denotation, causing a semantic clash.

The fact that seem requires a scalar predicate as its complement also explains why the participles compatible with very are also able to appear as bare complements of seem. This is illustrated in (82); cf. (83) where the participial phrase necessarily denotes an eventuality.

This correlation between degree modification and the ability to appear in the complement of seem position is not restricted to participles. As Matushansky (Reference Matushansky2002) notes, only those nouns that can be modified by degree adjectives like complete or utter can be complements of seem, cf. (84a–b).

Since no one is tempted to posit a syntactic category difference between the nouns fool and postman, the pattern in (82)–(83) should not tempt us to do so for participles either. Certain positions require that the elements occupying them be scalar, and the lexical meaning of some, but not all, nouns and adjectives allows them to be understood as (predicates of) scalar properties.

6.7.3. The future copula in Hebrew

The (in)compatibility of some Hebrew participles is another pattern that may be better explained by appealing to the distinction between property- and eventuality-denoting predicates rather than to a category difference. Meltzer-Asscher (Reference Meltzer-Asscher2010) shows that present participles behave non-uniformly in this context: some are able to follow the future copula (85a–b), and others not (85c). Additionally, Hebrew adjectives but not verbs can follow the future copula (86); see Doron (Reference Doron2003).

From this, Meltzer-Asscher (Reference Meltzer-Asscher2010) concludes that the participles in (85a–b) are adjectives, while those in (85c) are verbs. This conclusion is premature. The same participles that allow me’od-modification can also appear with the future copula; see Meltzer-Asscher (Reference Meltzer-Asscher2010: 2215) for details. We may therefore suspect that, like me’od-modification, compatibility with the future copula depends on the participle’s meaning (namely, whether it denotes a property or an eventuality). In fact, it seems that the future copula in Hebrew cannot combine with eventuality-denoting predicates.

To see this, consider the following. Hebrew has a synthetic future tense, and the roots in (85a–b) can also appear in that construction – for example, (87).

Is there a difference in the interpretation of (85b) and (87)? The answer seems to be ‘yes’. Consider the co-occurrence of the predicate with be-atsmo ‘by itself’ in the two constructions. Co-occurrence with be-atsmo has been argued to diagnose the syntactic presence of a Cause argument (Levin and Rappaport Hovav Reference Levin and Hovav1995, Alexiadou and Anagnostopoulou Reference Alexiadou, Anagnostopoulou, Alexiadou, Anagnostopoulou and Everaert2004, Koontz-Garboden Reference Koontz-Garboden2009, Alexiadou and Doron Reference Alexiadou and Doron2002, Kastner Reference Kastner2017, a.o.). Since the Cause argument is associated with a causing subevent, we expect it to be unavailable with predicates that do not denote eventualities. As expected on my hypothesis, be-atsmo is available with the Hebrew equivalent of ‘flourish’ in the synthetic future tense, but not when the participle combines with the future copula (88a–b). This suggests that only the participle in (88b) denotes an eventuality. Example (88a) is unacceptable even though the participle mesagseget can otherwise combine with be-atsmo (89a). It is specifically the presence of the future copula that precludes be-atsmo. Recall that participles like mesagseget are ambiguous between eventuality-denoting and property-denoting predicates; the future copula can only combine with the property-denoting participle.

The evidence clearly suggests that the predicate following the future copula cannot denote an eventuality. Eventive verbs always give rise to eventuality-denoting active participles, and they are incompatible with the future copula, as seen in (85c). Since this restriction seems to hold in addition to any c-selectional restrictions of the future copula, co-occurrence with the copula does not tell us anything about the participles’ category.