Predicative Adverbs: Evidence from Polish

It is commonly—even if usually tacitly—assumed that adverbs cannot act as primary predicates. Although it is not fully clear whether it is meant as a crosslinguistic generalization, an explicit statement to this effect may be found in a prominent monograph on predication (Rothstein 2001:129): “I assume that the absence of a predication relation is because adverbs are just not syntactic predicates. They never appear in a position in which they can be predicated of a syntactic argument; this is not because they must be predicated of events, since even if the argument denotes an event, it cannot have an adverb predicated of it. The examples in [(1)–(3)] are all unacceptable with adverbial predicates, though the corresponding adjectives are all OK.”

More often, discussions of nonverbal predication concentrate on nominal, adjectival, and prepositional predicates, implicitly assuming that adverbs cannot be productive predicates. For example, the only adverbial predicates mentioned in Roy 2013, a monograph on nonverbal predication, are lexical idiosyncrasies such as the Irish and Spanish equivalents of well in Jane is well.

The aim of this squib is to show that it is not universally true that “adverbs are just not syntactic predicates” or that they may act as primary predicates only exceptionally. We demonstrate on the basis of attested data¹ that, in Polish, a certain—relatively large and varied—class of adverbs may be productively used as predicates in copular constructions, namely, in constructions in which the subject is a verbal constituent—a complementizer phrase (CP) or a bare infinitival phrase (henceforth, InfP; not to be confused with InflP or IP, i.e., Inflectional Phrase)—denoting a fact or an event kind.²

It is well-known that, crosslinguistically, CPs and InfPs may act as grammatical subjects, also in predicative constructions in which they are predicated of, that is, in which they act as subjects of predication. (4)–(5) are examples from Uhrig 2018:chap. 6, a recent corpus study of English subjects.³

Similarly for Polish, as in the attested (6)–(11) from the National Corpus of Polish.⁴

Polish is a relatively free word order language, and the typical (but not exclusive) word order in such constructions is “predicate – copula – verbal subject,” perhaps due to the relative heaviness of such verbal (CP/InfP) subjects. In (6)–(7), the nominal predicate is in the instrumental case, just as in the usual predicative constructions with nominal subjects (see, e.g., Pisarkowa 1965 and Przepiórkowski 1999). Similarly, in (8)–(9), the adjectival predicate is in the nominative, again mirroring the usual predicative constructions with nominal subjects. As shown in (10)–(11), the predicate may be a prepositional phrase, just as in the case of nominal subjects. Examples (6)–(11) show that all six possibilities—nominal/adjectival/prepositional predicate ×CP/InfP subject—are fully acceptable and attested. Note that, in all six examples, the copula occurs in the “default” third person singular neuter⁵ form because the grammatical subject is not nominative (it is not cased at all); this is an instance of so-called default agreement (Dziwirek 1990).

Such verbal subjects of predication may also occur in “small clause” constructions involving verbs such as uważać ‘consider’, where—unlike in (6)–(11)—they are grammatical objects rather than grammatical subjects.

As expected, such subjects of predication in the object position become grammatical subjects in the passive voice (a by-phrase may also be used, as shown in (14)).

In brief, it is uncontroversial that verbal (CP/InfP) constituents may act as subjects of predication—and often as grammatical subjects—in Polish predicative constructions.

It turns out that in such cases the predicate may often be expressed by an adverb.⁶ For example:

Forms marked in (16)–(17) as ADV, while glossed here with English adjectives, are uncontroversial adverbs in Polish. Although they are morphologically related to corresponding adjectives—for example, dobrze.ADV :: dobry.ADJ.NOM.SG.M—they cannot be analyzed as adjectives: (a) they cannot be used adnominally (see (18)); (b) they can be used adverbally (see (19)); and (c) they are not syncretic with any forms in the inflectional paradigms of corresponding adjectives.⁷

Constructions with adverbial predicates are analogous to those with typical—nominal, adjectival, and prepositional—predicates: in both, the copula may be dropped in present tense, as in (20)–(21); in both, other verbs that can combine with nonverbal predicates may be used (e.g., wydawać się ‘seem’, stać się ‘become’), as in (22)–(23); in both, the verbal CP/InfP subject triggers “default agreement” on the verb.

When predicative adverbs are replaced with corresponding adjectives, the meaning normally remains the same;⁸ compare the predicative constructions in (16)–(17) with those in (24)–(25).⁹

Furthermore, it is possible to coordinate an uncontroversial prepositional predicate with an adverb, as in the attested (26).

While it is a matter of debate which cases of unlike category coordination should be analyzed via ellipsis of sentential coordination (so-called conjunction reduction), it is generally assumed that the unlike category predicates are conjoined directly, within a single predicative position, by virtue of bearing the same predicative features (Sag et al. 1985) or predicative “supercategory” (Bruening and Al Khalaf 2020). Indeed, (26) has the same predicative meaning as the prototypically predicative (27), which involves a nominative gerundial subject instead of an InfP and an adjective in place of an adverb.

As noted above, in the case of InfP and CP subjects, the same predicative meaning may be expressed by adverbial and adjectival predicates—compare again (16)–(17) with (24)–(25). While there is no systematic difference in meaning between the two constructions, there are some idiosyncratic lexical differences. For example, there is no adverb with the meaning corresponding to the main meaning of the adjective ważny ‘important’,¹⁰ so only the adjective may be used (see (28)), and—conversely—it seems that only the adverb głupio ‘stupid, awkward’ has the meaning ‘awkward’, so only the adverb can be used in the intended sense in (29).

However, there are systematic syntactic differences between the two constructions. First, only in the case of adjectival predicates may the nominalizer to ‘it’ be added to the subject CP; compare (30) with (16) and (24).¹¹

Second, extraction out of InfP is only possible with adverbial predicates. ¹² The following example, not involving extraction, is attested with the adjectival predicate najtrudniejsze ‘most difficult’, but it is at least equally acceptable with the adverb najtrudniej:

However, when the direct object of the infinitive is topicalized, only the version with an adverb is acceptable.

Third, as noted in ^{footnote 8}, when an InfP subject may occur either with an adverbial predicate or with an adjectival predicate, the adverbial version is much more frequent and often preferred. Thus, the adverbial version of (31) seems more natural than the attested adjectival version, and similarly for (17) vs. (25). No such acceptability contrast is observed with CP subjects, in which case adjectival predicates are often more frequent.

These facts can be explained by postulating that adverbs subcategorize for strictly verbal subjects of predication, while adjectives take nominal(ized) subjects of predication. That is, whenever adjectives seem to combine with verbal subjects of predication, these apparently verbal subjects have an outer nominal layer. This explains the extraction facts in (32): on the assumption that the additional nominal layer results in an island constraint, nominalized InfPs occurring with adjectival predicates are barriers to extraction, while purely verbal InfPs occurring with adverbial predicates are not. Moreover, on the assumption that syntactically more complex structures are dispreferred to synonymous simpler structures, frequent preference for predicative adverbs is also explained (though it is not clear why this preference is only observed in the case of InfP—and not CP—subjects). This categorial difference in subcategorization properties of predicative adverbs and adjectives also explains the fact that—while unlike category coordination of predicates is robust in Polish (see, e.g., (26)–(27))—adjectival and adverbial predicates cannot be coordinated; compare the ungrammatical (33) with the grammatical (17) and (25).¹³

This inability to coordinate adjectival and adverbial predicates is explained by conflicting expectations as to the actual categorial status of the shared InfP subject: predicative adverbs require an InfP, while predicative adjectives require a nominalized InfP. Finally, this analysis also explains the fact that when CPs are nominalized overtly, as in (30), they may combine with adjectival predicates, but not with adverbial predicates.

A potential problem for the proposed analysis is that InfPs cannot be nominalized overtly, whether the predicate is adverbial or adjectival (or in any other context); compare the grammatical (31) with the ungrammatical (34).

While there is a tension between overt nominalization facts and the hypothesis about different subcategorization patterns of predicative adverbs and adjectives, the impossibility of overt nominalization of InfPs does not rule out covert nominalization—a hypothesis supported by extraction facts in (32) and the coordination puzzle in (33), which cannot be explained otherwise. The relevant facts can be modeled by stating that the overt nominalizer to ‘it’ subcategorizes for CPs, while the empty N head acting as the covert nominalizer subcategorizes for both CPs and InfPs.

In summary, adverbs may be used as primary predicates in Polish; they occur in typical predicative constructions, including copular constructions; and they express predicative meanings analogous to those expressed with adjectives. However, the subjects of predication of such adverbial predicates must be strictly verbal—InfP or CP—and, given the limited range of abstract objects that such verbal constituents may denote (events, facts, etc.), the range of predicative adverbs is much smaller than the range of predicative adjectives (which may also predicate of nominal subjects, which may refer to a wide range of abstract and physical entities). This is the issue to which we turn next.

What kinds of adverbs may act as primary predicates? They do not seem to correspond directly to any of the classes postulated within classifications we are aware of (including those proposed in Cinque 1999, Ernst 2001, and Maienborn and Schäfer 2011). They might be preliminarily characterized as “evaluative” in a sense:¹⁴ they describe perception of eventualities (events, states, processes; Bach 1986) or facts as positive or negative, either in general or in some specific respect. The following are among the most common predicative adverbs in the National Corpus of Polish:

• •

  Good/Bad in general: dobrze ‘good’, fajnie ‘cool’, pięknie ‘beautiful’ (used metaphorically), wspaniale ‘wonderful’; źle ‘bad’, niedobrze ‘not-good’

• •

  Evoking good/bad emotions: miło ‘nice’, przyjemnie ‘pleasant’, zabawnie ‘funny’, ciekawie ‘interesting’; przykro ‘sorry’, głupio, niezręcznie ‘awkward’, smutno ‘sad’, straszno ‘scary’

• •

  Reasonable/Unreasonable: rozsądnie ‘reasonable’, mądrze ‘wise’, sensownie ‘sensible’; głupio ‘stupid’

• •

  Easy/Difficult: łatwo ‘easy’, nietrudno ‘not-difficult’, prosto ‘straightforward’, wygodnie ‘convenient’; trudno ‘difficult’, ciężko ‘hard’, niewygodnie ‘inconvenient’, niełatwo ‘not-easy’

• •

  Other: (nie)bezpiecznie ‘(un)safe’; korzystnie ‘profitable’, tanio ‘cheap’; zdrowo ‘healthy’, praktycznie ‘practical’, and others

The complete list of such adverbs is much longer, and many of them occur in predicative constructions very frequently. For example, within the 300-million-token balanced subcorpus of the National Corpus of Polish, there are a few thousand occurrences of predicative trudno ‘difficult’, łatwo ‘easy’, and dobrze ‘good’, and hundreds of occurrences of ciężko ‘hard’, miło ‘nice’, przyjemnie ‘pleasant’, and others. Both the number of different adverbs that can be used predicatively and the frequency of such predicative uses demonstrate the productivity of the construction.

The observation that predicative adverbs often refer to subjective perception of facts or events (or event kinds; see below), rather than to their objective characteristics, is supported by minimal pairs such as the following:

While (35) means that Jan was a skillful swimmer (even though perhaps he did not like swimming), (36) may only mean that swimming felt good to Jan (even though perhaps he was a terrible swimmer).

Since predicative adverbs often—but perhaps not always (see, e.g., tanio ‘cheap’)—express subjective attitudes, it is not surprising that they often take an experiencer argument—a dative nominal phrase, as in the attested (20) and (37)–(38).¹⁵

While in (37) the experiencer is an argument of (bardzo) przyjemnie ‘(very) pleasant.ADV’ only, in (38) the dative dependent of najtrudniej ‘most difficult.ADV’ is at the same time the subject of the infinitival phrase headed by uzyskać ‘gain’; that is, the dative experiencer controls the subject of the infinitival subject of predication.

According to the criteria in Landau 2013:29, this is an instance of obligatory control: the dative experiencer and InfP are codependents (they are dependents of the predicative adverb)¹⁶ and the controlled “PRO” (the subject of InfP) is interpreted as a bound variable, as evidenced for example by the exclusively sloppy readings under ellipsis in (39).

Here, the elided clause may only be understood as ‘It was most difficult for Marysia to ask for help’ (and similarly for the negated version); it cannot be understood as ‘Jan asking for help was most difficult for Marysia’.

While the issue of obligatory control into subject is controversial, it has been reported for languages as diverse as Balinese (Arka and Simpson 1998) and German (Stiebels 2007). Constructions with predicative adverbs, dative experiencers, and infinitival subjects provide another argument against a ban on control into subject.

Let us consider the following constructed examples, with InfP and CP subjects of predication, the latter with a pro-dropped subject, here assumed to refer to Jan:

Both examples seem to be saying that Jan experienced a certain event as nice. However, their meanings differ: the InfP version may only mean that he enjoyed swimming, while the CP version means that he considered it nice that he swam—perhaps because swimming was nice, or perhaps because he thus overcame his fear of water. Hence, only the CP version may be followed by something like ale samo pływanie nie było takie miłe ‘but the swimming itself was not so nice’ without creating a direct contradiction.

Such considerations lead to the conclusion that predicative adverbs may predicate of two sorts of abstract objects: events (or their kinds; see below) expressed by InfPs or facts expressed by CPs. This conclusion is supported by the fact that while all of the adverbs listed in section 5 occur with InfP subjects, only some of them—mostly those expressing general or emotional attitude—may occur with CP subjects. Thus, both events and facts may be good, bad, nice, reasonable, and so on, but only events—not facts—may be easy, safe, healthy, and so on. Also, the textual frequency of predicative adverbs with CP subjects is a couple of orders of magnitude lower than with InfP subjects.

The preliminary conclusion that predicative adverbs occurring with InfP subjects refer to specific event tokens is incompatible with the following example:

The verb wstać ‘get up’ has perfective aspect, so it seems to refer to telic events. But (41) does not assert or presuppose that there was any particular getting-up-by-Jan event—(41) may describe a situation in which Jan did not even seriously start getting up. Rather, what was perceived by Jan as difficult was (instantiating) the getting-up-by-Jan telic event kind. Where InfP is interpreted as referring to particular events, it is because of the veridical entailments of the predicate: unlike trudno ‘difficult’, miło ‘nice’ is veridical in this sense.¹⁷

On the basis of these observations, and representing facts as true propositions, we propose the following partial meaning representations (leaving out tense, etc.) for the InfP and CP versions of (40):¹⁸

These representations follow the neo-Davidsonian approach to event semantics (Davidson 1967, Castañeda 1967, Parsons 1990), in which particular thematic roles such as experiencer or agent are represented as separate functional predicates—for example, exp(s) = j or ag(e) = j. For concreteness, the predicative schema proposed in Rothstein 2001 is assumed here, on which the semantic representation of, for example, Jane is nice would be λs. nice(s) ∧ arg(s) = j (leaving out tense and aspect contributed by the copula); that is, the subject of predication is represented via the arg function from the predicate state s to the subject j. Such states s may host various semantic roles (Rothstein 2001:295), here the experiencer role. What is special about the constructions considered here is that the subjects of predication are not ordinary entities, but kinds of events (see, e.g., Gehrke 2019 and references therein) and facts. Again for concreteness, in (42) we adopt Chierchia’s (1998:348–349) down operator ^∩, which shifts properties into corresponding kinds. In this case, it shifts the property (function from entities to truth values) of being a swimming-by-Jan event to the respective event kind (i.e., an entity). After existential closure over s, this representation leads to the proposition that this event kind is nice for Jan.¹⁹

Similarly, understanding facts as true propositions and employing the standard intensional representation of propositions, (43) is saying that Jan considers the fact that he swam as nice. This encoding of facts results in a formal difference between (42) and (43). In the case of (42), the semantic type of the subject of predication (i.e., the value of arg(s)) is of type e (an event kind understood as an entity), and in the case of (43), it is of type 〈s, t〉 (an intensional proposition, i.e., a function from worlds to truth values). This suggests that perhaps two different—even if clearly related—predicates nice are involved in (42)–(43); let us call them nice_e and nice_st. If so, all adverbs combining with InfP subjects denote predicates such as nice_e, and those additionally combining with CP subjects are ambiguous, as they also denote predicates such as nice_st.²⁰

However, there is another common understanding of facts, on which they are not propositions, but particulars (see, e.g., Kratzer 2002, Mulligan and Correia 2017, and references therein), that is, entities of type e. If so, predicative adverbs are unambiguous. A possible argument for this view could be the acceptability of sentences such as (44), in which an InfP and a CP are coordinated within the subject position.

Assuming that only constituents of the same semantic type may be coordinated, (44) suggests that denotations of InfP and CP subjects are of the same semantic type e.

Unfortunately, neither of the two arguments for the two positions above is very strong. On the “formal ambiguity” view, (44) could be explained via some coercion or type-shifting operation turning an event kind into a fact (that an instance of this event kind took place). Conversely, on the “no ambiguity” view, the cooccurrence of different predicative adverbs with InfP and/or CP subjects could be a matter of syntactic selectional restrictions of particular adverbs. We leave this issue for future research.

The main aim of this squib was to refute the (often tacit) assumption that adverbs cannot act as primary predicates. In Polish, a relatively large class of adverbs may be used in predicative constructions when subjects of predication are abstract objects—event kinds and facts—expressed by InfPs and CPs. This requirement of purely verbal rather than nominal(ized) subjects of predication is the main difference between predicative adverbs and predicative adjectives, explaining differences in their syntactic behavior with respect to extraction and coordination. Predicative adverbs usually express an attitude toward event kinds and facts and may often combine with dative experiencers; in the case of InfP subjects, the dative experiencer acts as an obligatory controller into the subject. Constructions with predicative adverbs are textually frequent and fully acceptable—they clearly belong to the “core” rather than the “periphery,” if such a distinction is made.

We are very grateful to Mary Dalrymple, who provided detailed comments on previous versions of this squib, as well as to anonymous reviewers and the LI Squibs and Discussion editors, whose remarks also led to many significant improvements. We also benefited from comments by Mirosław Bańko, Emily Bender, Anna Bondaruk, Barbara Citko, Jan Fellerer, Steve Franks, Tracy Holloway King, John Lowe, Joan Maling, Geoff Pullum, Bożena Rozwadowska, Ewa Willim, Jacek Witkoś, and the audiences of LFG 2018, SE-LFG26, LAGB 2019, CSSP 2019, and SE-LFG28. The usual disclaimers apply.

Agnieszka Patejuk gratefully acknowledges the Mobilność Plus mobility grant awarded by the Polish Ministry of Science and Higher Education.

This squib is dedicated to the memory of Susan Rothstein (1958–2019).