The argument structure of long and short form adjectivesand participles in Russian

Olga Borik *

CLT Centre de Lingüística Teòrica, Universitat Autònoma de Barcelona, Departament de Filologia Catalana,Edifici B, Campus UAB, 08193 Bellaterra, Spain

Received 23 June 2013; received in revised form 24 April 2014; accepted 30 May 2014Available online 19 July 2014

Abstract

This paper addresses a long-standing problem of argument realization patterns with short and long form adjectives and participles inRussian: long forms can be used in both predicative and attributive positions, but realize their arguments only in adnominal positions,whereas short forms, which are only used in predicative position, do not show any restrictions on argument realization. Despite the factthat adjectives and participles take different types of arguments, they show the same general pattern of argument realization, hence thequestion that I will try to answer in this paper is of a more general character: what kind of mechanism allows for long form adjectives andparticiples to express their arguments (whatever they are), but only in adnominal position? I propose a unified account that covers bothclasses, which is based on the following idea: if a complex modifier headed by a short form is used in adnominal position, the long formaffix is necessarily added to this configuration to enable obligatory agreement in Case between a head noun and the modifier. I furtherargue that in predicative position, there is no reason to allow for a similar derivation due to the absence of Case agreement requirements,which rules out the use of long forms with realized arguments in this position.© 2014 Elsevier B.V. All rights reserved.

Keywords: Adjective; Participle; Short and long forms; Russian; Argument realization

1. Introduction

This paper discusses and analyzes the distribution and argument realization patterns of long and short forms(henceforth LFs and SFs, respectively) of adjectives and past passive participles in Russian.1 The problems raised by thesyntactic and semantic behavior of long and short forms are well-known and widely discussed in the literature (see Siegel,1976; Babby, 1975, 1999, 2009; Bailyn, 1994; Schoorlemmer, 1995; Paslawska and von Stechow, 2003; Geist, 2010), butthe analyses that have been proposed usually address some isolated issues either for adjectives or for participles and,hence, do not always capture all the relevant facts considered in this paper.

One of the conclusions that seems to have emerged from the literature addressing various aspects of short and longforms in Russian is that SFs behave more like verbal elements, whereas LFs are considered purely adjectival (see Babby,1975 and elsewhere, Bailyn, 1994; Pereltsvaig, 2001; Geist, 2010, among others). From a typological perspective, this isa well-grounded conclusion, since the class of adjectives is typologically one of the most diverse classes due to the ability

www.elsevier.com/locate/lingua

Available online at www.sciencedirect.com

ScienceDirect

Lingua 149 (2014) 139--165

* Tel.: +34 93 586 8149; fax: +34 93 581 2782.E-mail address: Olga.Borik@uab.cat.

1 In this paper, I will only consider past passive participles. Active and present participles behave differently in various respects and will not beanalyzed here.

http://dx.doi.org/10.1016/j.lingua.2014.05.0190024-3841/© 2014 Elsevier B.V. All rights reserved.

of adjectives to exhibit both nominal and verbal features to a different extent in different languages (see Dixon, 2004).Thus, in Russian adjectives and participles seem to fall into two categories, one of which, namely, SFs, exhibitsmore verbal features, whereas the other, LFs, has more nominal characteristics. One of the nominal characteristics that Iwill take particularly seriously in this paper is Case marking, which only LFs (of both adjectives and participles) candisplay.

The empirical data that I will try to account for are the following. First of all, while both LFs and SFs are allowed inpredicative position, only LFs can be used in adnominal position. Secondly, LFs can realize their arguments only inadnominal, but not in predicative position, but SFs do not exhibit any restrictions on argument realization in predicativeposition. This pattern is common to both adjectives and past passive participles, hence I will try to develop an analysiswhich is general enough to be extended to both categories. I will focus mainly on the features that unify adjectives andparticiples, whereas many syntactic and semantic differences that can be drawn between these two categories and thedetails of their representations will have to be disregarded for the present purposes.

I will argue that the distributional differences between short and long forms are due to a Case agreement requirementthat is imposed on adjectival and participial modifiers in adnominal position. Crucially, only LFs can agree in Case, thepossibility of Case marking being enabled by a special ‘thematic’ element -oj, which is also a morphological LF marker(see Halle and Matushansky, 2006). This will account for the fact that only Case-marked LFs can be used in adnominalpositions.

To capture the argument realization patterns of LFs and SFs, I will introduce a distinction between ‘simple’ and‘complex’ LFs, suggesting that the latter are derived on the basis of SFs and hence can ‘inherit’ their argument realizationproperties. The main idea is that when a SF is wanted in an adnominal position, it is forced to inflect for Case. In order toenable Case marking, a LF marker has to be added to a SF, creating thus a complex (i.e., potentially argument-realizing)LF. Predicative position does not impose any Case agreement requirements, so a need to derive a complex LF neverarises in this position, and all predicative LFs are therefore simple. Simple LFs will be treated as canonical adjectivesderived by an adjectivizing head which merges with a predicative root. To this structure, the LF suffix is added. As properadjectives, LFs will be able to appear in canonical adjectival positions, i.e., both as nominal modifiers and in predicativeposition.

The paper is organized as follows. Section 2 gives a data overview and discusses the distribution and argumentrealization patterns of both long and short forms of adjectives (section 2.1) and past passive participles (section 2.2). I thenproceed to discuss some of the most relevant previous analyses of the differences between long and short forms insection 3. In particular, in section 3.1, I will argue against an influential hypothesis first advanced in Babby (1975) andSiegel (1976) that all LFs are attributive, i.e., predicative LFs should be analyzed as modifiers of a null N. In section 4, aproposal will be developed which aims at explaining the observed empirical generalization for both SFs and LFs ofadjectives and participles. Section 5 concludes the paper.

2. Data overview

In this section, I will present a set of data that should be taken into consideration by any theory that attempts to account forthe distribution and argument realization patterns of long and short forms of participles and adjectives. These data will formthe empirical basis for the analysis proposed in this paper. For ease of exposition, adjectives and participles will be presentedin different sections, but, as will become apparent by the end of this section, the relevant properties are similar for bothclasses.

2.1. Short and long forms of adjectives

Many qualitative adjectives in Russian come in two forms, as exemplified in (1).2

(1) krasivyj krasivbeautiful.LF.NOM.SG.MASC. beautiful.SF.SG.MASC.dovol’nyj dovolencontent.LF.NOM.SG.MASC. content.SF.SG.MASC.veselyj veselmerry.LF.NOM.SG.MASC. merry.SF.SG.MASC.

O. Borik / Lingua 149 (2014) 139--165140

2 I will use the following abbreviations: NOM -- Nominative, INSTR -- Instrumental, DAT -- dative, ACC -- accusative, SG -- singular, PL -- plural, MASC --masculine, FEM -- feminine, NEUT -- neuter, PRES -- present, PST -- past, INF -- infinitive.

As can be seen from (1), the main morphological difference between the two forms is the presence of a special ‘ending’ --yjin the LFs, italicized in the examples above. Historically, this ending is related to a demonstrative pronoun (see Gorskovaand Haburgaev, 1981). The ending has several morphophonological variants, which are used depending on thephonological properties of the stem. Moreover, the ending changes depending on number and gender marking. This,in all the cases illustrated in (1), the ending --aja would correspond to feminine singular forms and the ending --oje -- toneuter forms. LFs and SFs are almost always derived from the same stem, with the exception of several suppletive formssuch as bol’soj (big.LF) vs. velik (big.SF).

An important morphosyntactic difference between LFs and SFs that should be pointed out right away concerns Casemarking: SFs cannot be inflected for Case, whereas LFs are always Case-marked. Thus, if we look again at the examplesin (1), we can see that LFs are marked as Nominative, whereas SFs are not.3 The ability to be Case-marked seems todirectly correspond to the presence of the LF ending, a relationship that will be central to my analysis of LFs in section 4.LFs used adnominally will always agree in Case with the head noun. Additional examples to illustrate Case agreement inadnominal positions are given in (2):

(2) a. Kloun rassmesil dovol’nuju devo!ku.clown.SG.MASC.NOM make.laugh.PST.SG.MASC. content.SG.FEM.ACC. girl.SG.ACC.‘The clown made a happy girl laugh.’

b. Dovol’nym detjam razdali podarki.content.LF.PL.DAT. children.DAT. give.out.PST.PL. present.PL.ACC.‘Presents were given to the happy children.’

The next important contrast between short and long adjectives that we will be concerned with is distributional: both formscan be used as predicates, but only LFs are allowed in adnominal positions. This pattern is illustrated in (3).

(3) a. Deti byli dovol’ny/ dovol’nye/ dovol’nymi.children.NOM. be.PST.PL. content.SF.PL./ content.LF.PL.NOM./ content.LF.PL.INSTR.‘The children were happy’

b. (*Dovol’ny/) dovol’nye deti prodolzali igrat’.(*content.SF.PL./) content.LF.PL.NOM children.NOM. continue. PST.PL. play.INF.‘Happy children kept playing.’

Once again, we see that the LF in predicate position in (3a) can be marked with either Nominative or Instrumental Case4,whereas the SF does not bear any Case marking. In the adnominal position (see (3b)), the LF agrees with the head noun inCase, in addition to number and (in singular) gender.

When we turn to argument realization patterns, we observe that the behavior of LF and SF forms also differ. As hasalready been shown in (3), in predicative position both LFs and SFs can be used, although only SFs will allow for theirarguments to be realized in this position, as the following examples with the LF adjective gordyj (proud) illustrate onceagain.

(4) a. Otec byl gordyj (*svoim synom).father.SG.NOM was.PST.SG.MASC. proud.LF.SG.MASC.NOM. (his.INSTR. son.INSTR.)

b. Otec byl gord svoim synom.father.SG.NOM was.PST.SG.MASC. proud.SF.SG.MASC.NOM. his.INSTR. son.INSTR.‘The father was proud of his son’.

Curiously, however, the arguments of LF adjectives can be realized in adnominal positions. As shown in (5a) and(5b) below, the LF adjective can realize its internal argument in this type of construction when used both post- and

O. Borik / Lingua 149 (2014) 139--165 141

3 Babby (2009:82) actually assumes that SF adjectives simply bear Nom Case and this is the only Case available for them. This assumption israther implausible, since there is no empirical motivation for it. SF adjectives are never used in Case-marked argument positions, even when aNom Case is required (e.g., a canonical subject position). LF adjectives, on the other hand, can be used in such positions.

4 An option of being marked either by Nom(inative) or Instr(umental) in predicative position is a common feature of nouns and LF adjectives inRussian. The differences between the two Case forms will not concern us here, but there is abundant literature on the issue, e.g., Nichols (1981),Timberlake (1986), Pereltsvaig (2001), Richardson (2001), etc. In all the examples below, the choice between the two Case forms is madearbitrarily, but it is never the case that another form cannot be used.

pre-nominally, respectively.5,6 Not surprisingly, and in accordance with the data in (3b), SF adjectives are banned fromthese constructions altogether, as witnessed once again by (5c,d).

(5) a. Otec, gordyj svoim synom, . . .father.NOM. proud.LF.SG.MASC.NOM. his.INSTR. son.INSTR. . . .

b. Gordyj svoim synom otec. . .proud.LF.SG.MASC.NOM. his.INSTR son.INSTR. father.NOM.

c. *Otec, gord svoim synom. . .father.NOM. proud.SF.SG.MASC.NOM. his.INSTR. son.INSTR. . . .

d. *Gord svoim synom otec. . .proud.SF.SG.MASC.NOM. his.INSTR. son.INSTR. father.NOM.

Thus, the examples in (1)--(5) illustrate the main descriptive generalization already stated in the Introduction. In particular,LF forms can be marked for Case and are used both predicatively and attributively, although they can realize their(internal) arguments only in attributive position. SF adjectives freely appear in predicative position, but cannot be usedattributively. SFs, as opposed to LFs, do not exhibit any restrictions on argument realization. In the next section, I willillustrate that the same generalization holds for past passive participles.

2.2. Long and short forms of participles

Past passive participles in Russian are formed from verbal stems by means of the participial suffix --(e)nn or --t. Not allthe verbs, however, have a corresponding past participle. The general rule is that these participles can be formed on thebasis of transitive verbs and specify a consequent state of the event described by the verb (see Schoorlemmer, 1995;Paslawska and von Stechow, 2003; Borik, 2013). The consequent state requirement is not specific to Russian, though, asevident from the literature on passive participle formation (see Kratzer, 2000; Maienborn, 2009; Gehrke, 2012; McIntyre,2013, among others). Thus, in German, (6a) is fully acceptable with an accomplishment verb which lexically specifies aconsequent state, but (6b) is deviant since it is formed on the basis of an activity verb, which, according to standardassumptions, does specify any consequent state (examples are from Gehrke, 2012):

(6) a. Die Lampe ist repariert.the lamp is repaired

b. #Die Katze ist gestreichelt.the cat is stroked

Note also that Russian allows for participles being formed on the basis of both morphologically perfective and imperfectiveverbs (Borik, 2013), a fact that is not that widely acknowledged in the literature on participle formation. As we will see insection 4.2.1, imperfective participles do have some properties that distinguish them from perfective ones, although thedescription of these differences will remain rather impressionistic in the present paper.

Past passive participles, just like adjectives, can appear in long and short forms. The distribution of the two formsmirrors the distribution of LF and SF adjectives in Russian (see example (3)): LF participles can appear in both predicativeand attributive positions, whereas SFs are banned from appearing adnominally, as illustrated in (7):

O. Borik / Lingua 149 (2014) 139--165142

5 As is well-known, in English the arguments cannot be realized in a pre-nominal position (see Embick, 2004, for instance). Russian is differentfrom English in this respect, i.e., there are no relevant differences between pre- and post-position, apart from punctuation: all complex nominalmodifiers (i.e., all modifiers with arguments) are separated by commas in post-nominal position. I will not address the nature of the differencebetween English and Russian in the present paper.

6 An anonymous reviewer points out that the appositive use of an adjectival phrase in (5a) suggests that the LF in this case is derived from areduced relative clause. This is an important observation and I will come back to this point in more detail in section 4.3, after the analysis for LFshas been proposed. For now, let me point out that if (5a) (and even (5b)) are indeed instances of reduced relative clauses, this still does not castdoubt on the generalization that LFs with arguments cannot be used in predicative position, since in a full relative clause counterpart of (5a) apredicative adjective still has to be used in a SF:

(i) Otec, kotoryj byl gord /*gordyj svoim synom,father.NOM. which be.PST.SG.MASC proud.SF.SG.MASC. /proud.LF.SG.MASC.NOM. his.INSTR. son.INSTR

A LF thus becomes possible only in a reduced relative, i.e. in the attributive use, assuming reduced relative clauses are merged DP internally (seeCinque, 2010).

(7) a. Deti byli udivleny/ udivlennye/ udivlennymi.children.NOM. were.PST.PL. surprised.SF.PL/ surprised.LF.PL.NOM/ surprised.LF.PL.INSTR.‘The children were surprised’.

b. (*Udivleny/) udivlennye deti pritihli.(*surprised.SF.PL/) surprised.LF.PL.NOM. children.NOM. become-quiet.PST.PL.‘Surprised children got quiet.’

Just like adjectives, LF participles in adnominal position agree in Case with the head noun:

(8) a. Kloun rassmesil udivlennyh detej.clown.SG.MASC.NOM make.laugh.PST.SG.MASC. surprised.LF.PL.ACC. child.PL.ACC.‘The clown made surprised children laugh.’

b. Udivlennym detjam razdali podarki.surprised.LF.PL.DAT. children.DAT. give.out.PST.PL. present.PL.ACC.‘Presents were given to the weary children.’

Now consider argument realization possibilities for participles. Once again, just as in the case of adjectives, we will startwith the predicative position, which admits both LF and SF participles, although only SFs can realize their arguments inthis position. Consider the examples in (9):

(9) a. Portret byl narisovannym (*karandasom/ *izvestnym hudoznikom)portrait.NOM. was painted.LF.INSTR. (*pencil.INSTR./ *famous.INSTR. artist.INSTR.)

b. Portret byl narisovan karandasom/ izvestnym hudoznikomportrait.NOM. was painted.SF. pencil.INSTR./ famous.INSTR. artist.INSTR.‘A portrait was drawn with a pencil/by a famous artist.’

In adnominal participial constructions,7 the LFs become free of any restrictions on argument realization. SF participles,like SF adjectives, are not allowed in this position (cf. (10)--(5) from section 2.1).

(10) a. Portret, narisovannyj karandasom/ izvestnym hudoznikom, . . .portrait.NOM. painted.LF.SG.NOM. pencil.INSTR./ famous.INSTR. artist.INSTR.

b. Narisovannyj karandasom/ izvestnym hudoznikom portretpainted.LF.sg.Nom pencil.INSTR./ famous.INSTR. artist.INSTR. portrait.NOM.‘A portrait drawn with a pencil/by a famous artist’

c. *Portret, narisovan karandasom/ izvestnym hudoznikomportrait.NOM. painted.SF. pencil.INSTR./ famous.INSTR. artist.INSTR.

d. *Narisovan karandasom/ izvestnym hudoznikom portretpainted.SF. pencil.INSTR./ famous.INSTR. artist.INSTR. portrait.NOM.

Thus, to sum up the data as far as the distribution and the argument realization patterns are concerned, both adjectivesand past passive participles exhibit exactly the same behavior. The main generalizations are the following:! LFs appear both predicatively and adnominally, SFs only predicatively;! SFs in predicative positions can always realize their arguments;! LF adjectives and participles can realize their internal arguments when they appear as adnominal modifiers, but not inpredicative positions.

A summary of the results is given in Table 1.In addition to the results presented in Table 1, we have also seen that LFs in adnominal positions always agree in Case

with the head noun.

O. Borik / Lingua 149 (2014) 139--165 143

7 The construction headed by a participle is called prichastnyj oborot (participial phrase). As should be clear from the data presented in (5),adjectives also appear in this type of construction. In footnote 6 I already mentioned that these constructions can be seen as reduced non-finiterelative clauses, and participles behave just like adjectives in full relative clauses (i.e., LFs with arguments are excluded from predicate position ofa full relative).

As I already pointed out in section 1, the data presented above are not new and several proposals have been made inthe literature to account for them. In the next section, I will review some of the previous proposals, examine their problemsand evaluate their potential to account for the pattern summarized in Table 1.

3. Short and long forms: some analyses

One of the most detailed syntactic analyses of LF vs. SF adjectives and, to a certain extent, participles is presented inBabby (1975, 1999, 2009). Babby argues that there is a fundamental distinction in the argument structure properties ofLFs and SFs. Simplifying, we could say that the basic idea of his proposal is that SFs select their argument whereas LFsdo not.8 I will use many of the insights of various versions of Babby’s analysis in this paper, but my analysis will differ in onecrucial respect: in section 3.1, I will argue against the view adopted by Babby (and many others after him) that all LFs areattributive.

Apart from Babby’s proposal, the existing literature does not treat both adjectives and participles in a unified analysis.However, the proposals aiming at explaining the distinctions between SFs and LFs, even if only adjectives or onlyparticiples are taken into account, could potentially be extended to cover all similar cases across-the-category. Two suchproposals will be discussed in sections 3.2 and 3.3.

However, before we turn to a discussion of particular proposals, let me mention a wide-spread intuitive characterizationof LFs vs. SFs stated in some traditional Russian grammars, e.g., Svedova et al. (1980). The basic idea is that thedistinction between long and short forms reflects a semantic difference between individual-level vs. stage-levelpredicates. Under this hypothesis, LFs of both adjectives and participles would be analyzed as expressing permanent,i.e., individual-level properties, whereas SFs would denote temporary properties.

Even though this proposal might seem to be intuitively appealing at first sight, a closer examination reveals that intuitionin this case is rather misleading because, from a theoretical perspective, the hypothesis is untenable. In particular, it isdifficult to see how the explanation in terms of permanent/temporary properties can be extended to the followingexamples:

(11) a. Cemodan byl polon.suitcase.NOM. was full.SF

b. Cemodan byl polnyj.suitcase.NOM. was full.LF.NOM.‘The suitcase was full’

(12) Celovek smerten.man.NOM. mortal.SF‘Man is mortal’

In (11), there is no difference in the interpretation of (11a) vs. (11b) that could be described by means of temporary/permanent properties. Thus, (11b) does not have to mean that the suitcase is permanently full. Example (12) (taken fromGeist, 2010:247) illustrates the point even better: the property of being mortal is clearly a permanent one for men, but it isnaturally expressed by means of a SF. Thus, the choice between LFs and SFs cannot be reduced to the temporary/permanent property distinction. Let me now turn to Babby’s proposal for long and short form adjectives and participles inRussian.

O. Borik / Lingua 149 (2014) 139--165144

Table 1Distribution and argument realization options for LF and SF adjectives and participles in Russian.

Predicative position Adnominal position (both pre- and post-nominal)

LF + +LF + arguments " +SF + "SF + arguments + "

8 Babby (2009) presents a fully developed theory of argument selection which includes a number of complex rules for changing ‘diathesis’(which is, roughly, Babby’s term for argument structure). I will not go into the details of his analysis here, since it would require a long technicalexposition of his rather specific machinery of argument selection and realization. I will only mention that I am not convinced that the specificimplementation proposed by Babby is advantageous over the one which is more commonly assumed.

3.1. LFs are not always attributive

One of the best-known analyses of SFs and LFs is based on the idea put forward and advocated by Babby (1975, 1999,2009), Siegel (1976) and Bailyn (1994, 2012), among others. A defining feature of all the proposals in question is that theyanalyze LFs as attributive and SFs as predicative. One of the main implications of this analysis is that when a LF appears in apredicative position, it actually appears within a noun phrase with an empty head, that is, an LF in predicative position isanalyzed as a modifier of a phonologically unrealized noun. From now on, I will refer to this proposal as the attributivehypothesis (for LFs).

One of the strongest arguments provided to support the attributive status of LF has to do with the agreement patternthat different predicative elements show with the 2nd person pronoun ‘vy’ in its polite use (cf. Sie in German, vous inFrench, Usted in Spanish, etc.). The pronoun is morphologically plural, but it can be used as a polite form of addressingone person. In this case, as illustrated in the examples below, in sentences where vy is a subject, nouns (i.e., idiotka ‘idiot’)and LFs (i.e., krasivaja ‘beautiful’) in predicate position are used in singular, whereas SFs (i.e., krasivy ‘beautiful.SF’) andverbs (i.e., byli ‘were’) appear in a plural form:

(13) a. Vy (*byla/byli) idiotka/ *idiotkiyou.PL. (*be.PST.SG.FEM./be.PST.PL.) idiot.SG.NOM./ *idiot.PL.NOM.‘You (addressing a female person) were an idiot.’

b. Vy (*byla/byli) krasivaja/ *krasivyje.you.PL. (*be.PST.SG.FEM./be.PST.PL.) beautiful.LF.SG.FEM.NOM./ *beautiful.LF.PL.NOM.

c. Vy (*byla/byli) *krasiva/ krasivyyou.PL. (*be.PST.SG.FEM./be.PST.PL.) *beautiful.SF.SG.FEM./ beautiful.SF.PL.‘You (addressing a female person) were beautiful.’

Thus, we observe in the above examples that LFs pattern with nouns, whereas SFs pattern with verbs with respect to thisagreement test. These data are taken to indicate that the LF adjective in (13b) is actually used adnominally, thus showingthe same agreement pattern as the noun in the same position in (13a).9 Hence, we expect LFs to always follow thenominal pattern, if they are taken to be nominal modifiers. However, the parallel between nouns and LFs observed in (13)breaks down when, instead of the polite form of a pronoun, a plural noun (including pluralia tantum) is used in the subjectposition. In this case, a noun in predicate position shows up in the singular, as illustrated in (14a) below, whereas LFadjectives (14b) pattern instead with SFs (14c), both taking a plural form.10

(14) a. Eti konfety byli *podarkami/ podarkom.these.NOM.PL. sweet.NOM.PL. be. PST.PL. present.PL.INSTR./ present.SG.INSTR.‘These sweets/chocolates were a present.’

b. Eti konfety byli vkusnymi/ *vkusnymthese.NOM.PL. sweet.NOM. be.PST.PL. tasty.LF.PL.INSTR./ tasty.LF.SG.INSTR.

c. Eti konfety byli vkusny/ *vkusnathese.NOM.PL. sweet.NOM. be.PST.PL. tasty.SF.PL./ tasty.SF.SG.FEM.‘These sweets/chocolates were good.’

In (14), we see that SFs and LFs pattern together, differently from nouns. In principle, this is unexpected under thehypothesis that LFs are always used attributively and hence should show the same agreement as predicative nouns. Apotential explanation could be that the LF adjective in (14b) agrees with a null noun which would be the same as thesubject, i.e., konfety (sweet.PL).11 This is a reasonable expectation, but the proponents of the attributive hypothesis havenever explicitly specified what kind of noun we should expect to appear as the head of a noun phrase in predicate position.This point itself has been brought up as a criticism of the attributive hypothesis. For instance, as noted by Geist (2010),with many adjectives, an attempt to lexicalize the null N does not lead to sensible results, as illustrated in (15), where

O. Borik / Lingua 149 (2014) 139--165 145

9 Anticipating the proposal for SFs to be presented shortly, I should mention here that an alternative explanation for the pattern in (13) is that theSF, being essentially a verbal element, merely follows the verbal agreement pattern: note that the auxiliary be in all the examples is in the pluralform. This alternative explanation is fully compatible with my analysis presented in section 4.

Moreover, as pointed out by an anonymous reviewer, the nominal character of LFs (i.e., the fact that the LF affix is historically related to ademonstrative pronoun, see section 2.1) could also help to explain the contrast between LFs and SFs in (13).10 In (14c) I used a feminine form of the SF adjective simply because konfeta (sweet/chocolate) in the singular is feminine. The gender markingdoes not matter for grammaticality judgments: all the singular SFs are unacceptable in (14c).11 This point was raised by an anonymous reviewer.

repeating Ivan as a predicative noun does not make more sense than using the other listed options. This difficulty,however, is not expected, since it would require an additional implausible stipulation that the noun in predicate position willnecessarily remain empty.

(15) Posle ekzamena Ivan byl zloj ØN

after examen Ivan.NOM. be.PST.SG.MASC. angry.LF.SG.MASC.NOM.(??!elovek, ??student, ??Ivan)(person, student, Ivan)‘After the exam, Ivan was angry.’ (but ??an angry person/student/one)

Coming back to the pattern in (14) and the problem of a phonologically silent head noun that the LF adjective is presumed toagree with, in the absence of any clear rule concerning the nature of the head noun, a number of candidates could beavailable, as in, for instance, (16a), where one potential candidate to replace a zero noun is singular, and another one is plural.In this case we would expect an adjective to show variable singular/plural agreement also when a noun is not realized.However, this is not the case, and only a plural LF form in (16b) (see also (14b)) is grammatical in the absence of any overtnoun.

(16) a. Ovos!i byli poleznym (produktom)/ poleznymi (ovos!ami).vegetable.NOM.PL. be. PST.PL. healthySG.INSTR. (product.SG.INSTR.)/ healthy.PL.INSTR. (vegetable.PL.INSTR.)‘Vegetables were (a) healthy (thing)/healthy(vegetables).’

b. Ovos!i byli *poleznym Ø/ poleznymi Øvegetable.NOM.PL. be. PST.PL. healthy.SG.INSTR. (Ø.SG.INSTR.)/ healthy.PL.INSTR. (Ø.PL.INSTR.)‘Vegetables were healthy.’

What the discussion of the examples in (13) through (16) is meant to show is that LFs and overt nouns in predicate position donot always follow the same agreement pattern, as would be expected under the attributive hypothesis. In addition, in somecases it is rather difficult to come up with a possible overt counterpart of an assumed zero head noun in the case of predicativeLFs. Thus, I conclude that the argument used by Babby (1975 and elsewhere) based on the agreement of LFs with the politeform of 2nd person singular pronoun vy (you) in Russian is insufficiently robust, as a more careful examination reveals.

In addition to agreement, there is another unexpected difference between LFs and nouns with respect to Casemarking. The observation comes from Geist (2010), who shows that if we look at depictive predicates, which can be bothnominal and adjectival, we notice that, while nominal depictives can only be used in Instrumental but not Nominative Case,Nominative LFs are actually grammatical as depictives, as shown in the examples below:

(17) a. On vernulsja domoj *oficer/ oficerom.he.NOM. return.PST.SG.MASC.REFL. home officer.NOM./ officer.INSTR.‘He came back home an officer.’

b. On vernulsja domoj sedoj/ sedym.he.NOM. return.PST.SG.MASC.REFL. home grey-haired.NOM./ grey-haired.INSTR.‘He came back home grey-haired.’

Thus, (17) provides yet another context where the behavior of LF adjectives deviates from what would be expected underthe hypothesis that LFs are always attributive. If this were the case, we would expect LFs to exhibit exactly the samemorphosyntactic properties as nouns in the same syntactic environment, yet this is not what we find empirically.

Babby (1975, 2010) and Siegel (1976) also use some semantic arguments that, at first sight, seem to provide evidencein favor of the attributive status of LFs in Russian. In particular, they argue that LFs, being part of the NP structures,semantically always function as restrictive modifiers. If this is the case, then the following prediction is made: a LF inpredicative position cannot appear with a noun whose reference is unique, as exemplified by (18) below12:

(18) *Prostranstvo beskone!NOE.space.NEUT.NOM. infinite.LF.SG.NEUT.NOM.‘The space is infinite’ (but *The space is an infinite one)

O. Borik / Lingua 149 (2014) 139--165146

12 This example is from Siegel (1976:297). The grammaticality judgments are given as in the original.

An objection that can be raised with respect to this argument is that, if attributive LFs were restrictive modifiers, then theprediction made above should extend beyond predicative uses. In other words, phrases like beskone!NOE prostranstvo(‘infinite space’) should simply be impossible to construe.13 However, this is not the case, as witnessed by (19) (taken fromhttp://otvet.mail.ru/question/90516):

(19) Kak vygljadit beskone!NOE prostranstvo?how look.PRES.3SG. infinite.LF.SG.NEUT.NOM. space.SG.NEUT.NOM.‘How does infinite space look like?’

Another problem for this line of reasoning is posed by examples with LF modifiers which do not necessarily yield a restrictiveinterpretation, like in the case of ‘a beautiful dancer’ (see Larson, 1998; Cinque, 2010).14 In a restrictive interpretation, theconstruction is interpreted with respect to an individual x as ‘x is beautiful as a dancer’, although x can be entirely unattractiveotherwise. In a non-restrictive interpretation, x is both a beautiful individual and a dancer. If attributive LFs were alwaysrestrictive, the latter would not be a possible reading for the Russian counterpart of ‘a beautiful dancer’. The prediction,however, is not borne out: the Russian phrase krasivyj tancor (beautiful dancer) has the same ambiguity as thecorresponding English phrase and can refer to a person who is beautiful and also a dancer, i.e., can have a non-restrictiveinterpretation.

Finally, the argument realization pattern with LFs described in sections 2.1 and 2.2 presents yet another problem forthe attributive hypothesis. In particular, if an adjective or a participle can have its arguments realized in an attributiveposition with lexically overt nouns, we would need an explanation of how this option is blocked when a LF is used in apredicative position as the modifier of a zero noun. It is difficult to imagine what kind of independent motivation could beprovided to explain the ban on argument realization for LFs specifically in a predicative position, so this problem does notseem to have an easy solution within the attributive hypothesis.

To sum up, as I hope to have illustrated in this section, there are substantial reasons to cast serious doubt on thevalidity of the attributive hypothesis adopted in much of the syntactic literature on LFs. Treating LFs in predicativepositions as modifiers of an empty N head seems to create more problems than it solves, so I reject the attributivehypothesis as untenable.15

3.2. State vs. property

Geist (2010) proposes an analysis for SF vs. LF adjectives suggesting that SFs should be treated like stative verbs, whilegiving a property analysis to LF adjectives. She assumes that technically the difference between properties and stative verbsis encoded in their semantic representation: the latter but not the former introduce a state argument s, i.e., belong to a class ofeventualities. She argues that this categorical difference can explain the differences in the argument structure of LFs andSFs. Her representations for LF and SF adjectives are given in (20a) and (20b), respectively (Geist, 2010:254--255):

(20) a. [aP a [!P<i>]]] LF adjectivesb. [vsP DPi [vs [!P<i>]]] SF adjectives

Schematically, the proposal works as follows. The adjectival head a in (20a), which derives LF adjectives, does not project aspecifier and takes a root P which lexicalizes some property or quality as a complement. Due to the absence of a specifierposition, the theta-role <i> carried by P has to be assigned ‘externally’, i.e., outside of the aP projection. As for SFs, they arederived by a stative verbal head vs which does project a specifier, therefore providing a way for a P root to discharge its theta-role within the same projection. Thus, Geist essentially suggests that SF adjectives should be treated like stative verbs,following the analysis of Babby (1999, 2009) and argues that SFs really select for their arguments, i.e., exhibit a typical verbalbehavior when it comes to argument selection, whereas LFs combine with their arguments like modifiers.

An immediate objection to this analysis is that it does not solve the problem of how LF adjectives in adnominal positioncome to realize their arguments if they never project them.16 Moreover, it seems problematic that in her analysis based on

O. Borik / Lingua 149 (2014) 139--165 147

13 In addition, let me point out that (18) does not really sound very bad for many speakers, myself included. A simple Google search also yields anumber of perfectly natural examples with the LF of infinite in predicate position, just like in (18).14 I thank an anonymous reviewer for bringing this point to my attention.15 More arguments in favor of the attributive hypothesis can be found in Babby (1975, 2009, 2010). For obvious reasons, I cannot go through allof them. I hope, however, that the problems raised for those arguments that have been discussed here constitute sufficient counter-evidence.16 Apart from a short remark in fn. 5 of her paper, Geist does not deal with adnominal positions at all. In this footnote, she remarks that LFs inadnominal positions could be derived from SFs, but provides no further details.

(20b), SF adjectives are essentially indistinguishable from stative verbs. Thus, (20b) can, in principle, give rise to both(21a) with a SF adjective and (21b) with a lexical verb to be sick/ill:

(21) a. Rebenok bolen.child.SG.MASC.NOM. sick.SF.SG.MASC.

b. Rebenok boleet.child.SG.MASC.NOM. be-sick.PRES.3SG.‘The child is sick.’

However, there is an important difference between SFs and canonical verbs: the verb in (21b) is inflected for tense,whereas the SF in (21a) is not. This difference is unexpected and problematic if both predicates are in fact associated withthe same syntactic structure. I will get back to this point in section 4.1 and suggest a modification of the analysis proposedby Geist (2010) to handle this problem.

Apart from the internal problems with the analysis just mentioned above, it would be difficult to extend Geist’s accountto participles for two reasons. First of all, Geist argues that SF adjectives always have a stative interpretation, but SFparticiples can be eventive, as witnessed by the fact that passive sentences with SF participles allow for by-phrases andvarious types of event related modifiers:

(22) Dom byl pokrasen nasimi rabo!imi/ za 2 !asa/ bystro.house.NOM. be.PST.SG.MASC. painted.SF.SG.MASC. our.INSTR. workers.INSTR./ in 2 hours/ quickly‘The house was painted by our workers/in two hours/quickly.’

An easy way to accommodate the eventive character of SF participles would be to assume that in structure (20b) theverbal head can be either stative or eventive (Arad, 1999). However, in this case we would have to explain why SFadjectives cannot be eventive17 and it would become even harder to make a distinction between verbs and SF forms.Thus, what appears to be an easy way quickly turns into a problematic one.

The second problem is that there are semantic differences between LF participles and LF adjectives as well, so we donot find a full parallel even among LFs themselves. In particular, LF participles do not seem to merely denote properties,like their adjectival counterparts, but presuppose a reference to a prior event. One way to illustrate this difference is bycomparing the interpretation of a sentence with a LF participle nadutaja (inflated.LF.SG.FEM.NOM.), on the one hand, and asentence with a full-fledged adjective naduvnaja (inflatable.LF.SG.FEM.NOM.), both derived from the same verbal root.18

(23) a. Lodka byla uze nadutoj, hotja nikto eje ne naduval.boat.SG.FEM. was already inflated.LF.SG.FEM.INSTR. though no one her not inflated

#‘The boat was of the type that can be pumped up/was inflatable, although no one has pumped it up yet’OK: ‘The boat has been pumped, although no one has pumped it up’

b. Lodka byla (*uze) naduvnoj, hotja nikto eje ne naduval.boat.SG.FEM. was already inflatable.LF.SG.FEM.INSTR. although no one her not inflated.‘The boat was inflatable, although no one had (yet) inflated/pumped it up.’

As can be seen from the translations, the purely stative interpretation of naduvnoj in (23b), which simply denotesa certain property that can be attributed to a subject lodka (boat), is not available for (23a), where a past participleis used instead of the adjective. That is, the boat has to have been inflated at some point, and although (23a) onlyrefers to a consequent state of the inflating event, it does presuppose a prior event that led to this consequent state,as indicated by an acceptable translation. (23b), however, does not necessarily refer to or presuppose any precedingevent.19 The ungrammaticality of the adverb uze (already) in this example emphasizes that there was no change ofstate (cf. Löbner, 1989, Krifka, 2000), which follows if there is no reference (actual or presupposed) to an underlyingchange of state event.

O. Borik / Lingua 149 (2014) 139--165148

17 And they cannot, as I will illustrate in the next subsection.18 The examples are taken from Borik (2013:121--122), who contrasted the Russian data presented in these examples with the comparableGreek data from Alexiadou and Anagnostopoulou (2008).19 A reviewer points out that -- able adjectives can refer to a (potential) future event: an inflatable boat is one that can potentially be inflated. Thismight well be the case, but my point is that the reference to a prior event is never obligatory with adjectives but seems to be presupposed in thecase of participles. If we take an arbitrary adjective which is not of the -- able type, it might not be eventive at all, whereas any participle will have ahidden reference to a preceding event of a relevant type.

Thus, past participles, even in the LF, seem to always entail the existence of a prior event, whereas adjectives do not,yet the analysis in (20a) would not capture this difference. It is exactly for this reason that I doubt that Geist’s propertyanalysis can be extended to the case of participles and their argument realization properties. That is, even if the distinctionbetween states and properties proposed by Geist could be extended to fully account for the argument patterns withadjectives, it cannot form a solid foundation for an account that aims to explain the argument realization properties of LFand SF of both adjectives and participles.

3.3. Stativity vs. eventivity

In the recent literature on the syntax and semantics of passive participles, it has often been proposed that several typesof passive participles should be distinguished (cf. Kratzer, 2000; Embick, 2004; Alexiadou and Anagnostopoulou, 2008;Sleeman, 2011, etc.). One of the properties that appears relevant in making such distinctions is the property of eventivity.In particular, eventive participles can form verbal passives, whereas other types of participles cannot. Eventive participles,it has also been claimed, cannot occur in a prenominal position (Embick, 2004, but see section 4.2.2). With respect toRussian, it has also been argued that long vs. short forms of participles can encode an eventive vs. stative distinction. Inparticular, Schoorlemmer (1995) and Borik (2013) claim that LF past passive participles can only be stative while SFparticiples can be both stative and eventive20, and these authors develop a semantic analysis to capture this fact. Thefollowing data provide empirical support to this claim.

(24) a. Dom byl pokrasennym (*za 2 !asa/ *bystro/ *special’no).house.NOM. was painted.LF.SG.MASC.INSTR. (*in 2 hours/ *quickly/ *on purpose)

b. Dom byl pokrasen za 2 !asa/ bystro/ special’no.house.NOM. was painted.SF.SG.MASC. in 2 hours/ quickly/ on purpose‘The house was painted in two hours/quickly/on purpose’

This pair of examples illustrates that event-related modifiers such as temporal, manner and purpose adverbs are ruled outwith LFs, as witnessed by (24a), but not with SFs, as shown in (24b), partially repeated from (22). These data indicate thatLF participles in Russian are not really eventive in the sense that they do not make a direct reference to an underlyingevent. SF participles, on the other hand, seem to make both an underlying event and a consequent state available formodification, as (25) illustrates:

(25) Vorota (byli) otkryty storozhem rovno v 6 utra na 2 !asa.Gates (were) opened.SF.PL watchmen.INSTR. exactly in 6 morning for 2 hours‘The gates were opened by the watchman at exactly 6 in the morning for 2 hours.’

On the basis of these and some other observations, Borik (2013) argues that SF participles can form both adjectival andverbal passives, i.e., they are ambiguous between a stative and an eventive interpretation. If the realization of by-phrasesand instruments is related to the eventive character of the predicate and syntactically requires a more complex verbalfunctional structure (see Embick, 2004; Alexiadou and Anagnostopoulou, 2008, etc.), then the stativity of the LFs canexplain the fact that the arguments cannot be realized with LFs, as was illustrated in (9a) above, repeated below as (26):

(26) Portret byl narisovannym (*karandasom/*izvestnym hudoznikom)portrait.NOM. was painted.LF.INSTR. (*pencil.INSTR./*famous.INSTR. artist.INSTR.)‘The portrait was drawn (*with a pencil/*by a famous artist)’

Borik (2013) does not discuss adnominal uses of LF participles, but it seems evident that the fact that LF participles canrealize their arguments in adnominal position is unexpected under the hypothesis that the argument realization ban isrelated to the inherent stativity of LFs. Moreover, the argument realization options with long vs. short forms are notexclusive to participles. Just like participles, adjectives do not realize arguments in predicative positions when a LF is usedyet impose no argument restrictions when it comes to SFs. If we extend the analysis in stative/eventive terms proposed forparticiples also to adjectives, we would have to claim that SF adjectives can be eventive. If this were the case, we would

O. Borik / Lingua 149 (2014) 139--165 149

20 An alternative view, namely, that all SF participles in Russian are also stative (and hence all periphrastic passives are adjectival) was putforward in Paslawska and von Stechow (2003), who argue against Schoorlemmer’s (1995) view. For the arguments against their proposal thereader is referred to Borik (2013).

expect SF adjectives to combine with different types of eventive adverbials, similarly to SF participles (see (24b)). Thisprediction, however, is not borne out, as shown in (27):

(27) Cemodan byl polon (ves!ej) (*za 2 !asa/ *bystro/ *special’no).suitcase.NOM. was full.SF (things.GEN.) (*in 2 hours/ *quickly/ *on purpose)

While an internal argument ves!ej (things) can be optionally realized with the SF polon (full), none of the event-orientedadverbials are actually admissible in the sentence. This points to the conclusion that having an event argument might berelevant for the realization of some particular type of argument and modifier, but cannot be a general defining property forargument realization, as witnessed by SF adjectives. Thus, even if the analysis in terms of stative/eventive distinctionmight be useful for participles, it cannot be extended to cover the case of adjectives, which, as we have seen above, alsorealize their internal arguments in predicate positions with SFs, but not with LFs.

To sum up the discussion so far, having rejected the attributive hypothesis advocated by Babby (1975, etc.) in section3.1, I have argued that various analyses that have been proposed for LFs and SFs of adjectives, on the one hand, andparticiples, on the other, cannot be easily extended to the other category. However, we also saw in section 2 of this paperthat adjectives and participles behave alike with respect to the (non-)realization of their arguments in different syntacticpositions. From this I conclude that since the existing analyses cannot help to explain the argument realization patterns oflong vs. short forms, an alternative must be sought. The alternative that I propose will be discussed in the next section.

4. The analysis

In this section I will suggest an analysis that accounts for the data presented in section 2 and summarized in Table 1. Inparticular, my account will explain both the distribution and the argument realization pattern of long and short adjectivesand participles. Let me start by introducing the highlights of the proposal.

Following the insight of Babby (1975, 1999, 2009) and many others, I adopt the hypothesis that SFs are essentiallyverbal elements. In section 4.1, I will discuss two arguments in favor of the verbal character of SFs: their agreementpattern in predicative position and their argument selection properties. Following a rather common assumption thatsyntactically, arguments are introduced in a layer of functional structure associated with verbs (see Alexiadou, 2001;Borer, 2003; Alexiadou et al., 2006, etc.), I will assume that a syntactic configuration associated with SFs includes thecorresponding structure. The verbal character of SFs ensures that they can always function as predicates. At the sametime, in principle SFs are not tolerated in adnominal position since they lack some necessary morphosyntactic featuresrequired for agreement purposes in this position, namely, they cannot be marked for Case. However, a SF can beendowed with the necessary features, which will lead to the derivation of a complex LF, which inherits all the argumentsand other properties (e.g., possible eventivity, aspect, etc.) of the corresponding SF. The details of my analysis of SFs arespelled out in section 4.1, whereas complex LFs are discussed in section 4.2.2.

Simple LFs, on the other hand, are purely adjectival, i.e., agreeing elements, and are generated in a simple adjectivalconfiguration. As full-fledged adjectives, they can be used freely in both adnominal and predicative positions, which isexactly what we observe, but due to the absence of any verbal functional layers they will not introduce any arguments.Simple LFs are discussed in section 4.2.1.

Thus, in the analysis that I propose, I crucially rely on one clear morphosyntactic difference between LF and SFadjectives and participles that was mentioned in section 2: only LFs can be Case marked. Not all of the ingredients of thisanalysis are new, but the present proposal will make it clear why and how both LF adjectives and participles can realizetheir arguments in adnominal positions. I claim that we find complex LFs in these positions for the reasons of Caseagreement. On the other hand, complex LFs are excluded from predicate position due to economy considerations whichprevent unnecessary derivations: SFs are not required to agree in Case with the subject, so they simply remain SFs.

In the rest of this section, we will look at various parts of the proposal, dealing, in turn, with SF forms, simple LFs andcomplex LFs. Let me start by introducing the analysis for SFs.

4.1. SFs

As was already stated in the Introduction, typologically, adjectives do not form a uniform class. In particular, adjectivesshare some nominal characteristics and some verbal characteristics, and in this paper I will adopt one of the most widelyaccepted claims in the literature (see Babby, 1975 and elsewhere, Pereltsvaig, 2001; Geist, 2010, etc.) that SFs areessentially verbal rather than nominal elements. I will discuss two arguments that prove the verbal nature of SFs: (a) theiragreement pattern with the main subject and (b) their argument selection properties.

With regard to the first argument, it was already emphasized in the classical work by Babby (1975), who in turn citedsome older Russian grammarians as having made similar statements (e.g., Sakhmatov, 1941), that the agreement pattern

O. Borik / Lingua 149 (2014) 139--165150

of SFs is the same as for verbs in the past tense in Russian: both classes agree with the subject in number and, if thenumber is singular, in gender. This is illustrated in (28) for a SF adjective polon (full) with an optional copula verb21 and in(29) for past tense forms of the verb bolet’ (to be sick):

(28) a. Cemodan (byl) polon.suitcase.NOM.SG.MASC. (was.SG.MASC.) full.SF.SG.MASC.

b. Sumka (byla) polna.bag.NOM.SG.FEM. (was.SG.FEM.) full.SF.SG.FEM.

c. Sumki /!emodany (byli) polny.bag.NOM.PL. /suitcase.NOM.PL. (were.PL.) full.SF.PL.‘A suitcase/a bag/bags was/were full.’

(29) a. Mal’!ik bolel.boy.NOM.SG.MASC. be-sick.PST.SG.MASC.

b. Devo!ka bolela.girl.NOM.SG.FEM. be-sick.PST.SG.FEM.

c. Deti boleli.children.NOM. be-sick.PST.PL.‘A boy/a girl/children was/were sick.’

Note that verbs in the present tense agree differently. The main verb agrees with a subject in person, i.e., an additionalperson agreement feature is realized on the verb in the present tense, whereas gender agreement is absent altogether:

(30) a. Mal’!ik/ devo!ka boleet.boy.NOM.SG.MASC./ girl.NOM.SG.FEM. be-sick.PRES.3SG.‘A boy/a girl is sick’.

b. My boleem.we.NOM. be-sick.PRES.1PL.‘We are sick’.

As for LFs, the main difference between SFs and LFs in predicate position is that the latter but not the former are markedfor Case, as was illustrated in section 2 (see examples (3a) for adjectives, (7a) for participles). Predicatively used nounsexhibit a more complex pattern with respect to number marking: in some cases, a predicate noun does not have to agreewith the subject in number: example (16a) in section 3.1 illustrates the point. Thus, the full parallel in agreement isobserved only in the case of SFs and past tense verb forms in Russian. This parallel is usually the main reason to suggestthat one of the verbal characteristics that SFs exhibit is their agreement pattern with the subject of a sentence.

These facts are not surprising if we take into account the origin of past tense forms in Russian. Past tense forms, orl-forms, historically come from active past participles which were used in compound tenses like, for example, present orpast perfect, as is well known from historical grammars (e.g., Gorskova and Haburgaev, 1981). In modern Russian l-formsare used as full-fledged (and the only) past tense forms, carrying out all the semantic functions of an absolute tense.Crucially for our purposes here, l-forms agree with the subject of a sentence in number and, in singular, in gender, unlikepresent tense forms, which exhibit a person-number agreement. SFs thus follow the agreement pattern of past l-forms.

The agreement pattern is not the only evidence that points to the verbal character of SFs.22 Another argument I want todiscuss here is the selectional properties of SFs. In particular, Geist shows that SFs of adjectives really impose selectionalrestrictions on their subjects, just like verbal predicates do, whereas LFs do not exhibit strong restrictions. Compare theutterances in (31):

(31) a. Rebenok bol’noj.child.NOM.SG.MASC. sick.LF.NOM.SG.MASC.‘The child is sick’

b. Golos u nego bol’noj.voice.NOM.SG.MASC. at him sick.LF.NOM.SG.MASC.‘His voice is sick’

O. Borik / Lingua 149 (2014) 139--165 151

21 The copula in Russian is always absent in present tense statements and is used only in the past. The presence/absence of the copula verbdoes not influence the morphosyntactic characteristics of SFs.22 For a more extensive summary of the verbal properties of SFs the reader is referred to Babby (1975) or Geist (2010).

c. Rebenok bolen.child.NOM.SG.MASC. sick.SF.SG.MASC.‘The child is sick.’

d. *Golos u nego bolen.voice.NOM.SG.MASC. at him sick.SF.SG.MASC.‘His voice is sick’

As shown in these examples, the SF bolen (sick.SF) admits the nominal phrase ‘child’ but not ‘voice’ as a subject, whereasa LF is fine with both subjects. The restrictions that a SF imposes on the subject coincide with the selectional restrictions ofthe verb bolet’ (to be sick/ill):

(32) Rebenok /*golos bolel.child.NOM.SG.MASC. /*voice.NOM.SG.MASC. be-sick.PST.SG.MASC.

Moreover, some SFs also obligatorily require their internal arguments to be realized (see (33), taken from Geist,2010:248), whereas this is never the case with LFs (Babby, 1975; Geist, 2010). This feature, just like the selectionalrestrictions on subject, groups SFs with verbs.

(33) Eto pal’to *(Kate) dlinno.this.NOM.SG.NEUT. coat.NOM.SG.NEUT *(Katja.DAT). long.SF.SG.NEUT.‘This coat is too long *(for Katja)’.

Thus, the data presented above clearly indicate that SFs exhibit verbal behavior in several respects. However, the parallelbetween SFs and verbs is not absolute. In particular, one of the most important and relevant distinctions between the twoclasses is that verbs but not SFs can express absolute tense. Thus, to ensure that a sentence with the SF of an adjectiveor participle has an unambiguous past interpretation, a copula verb is needed.

Let me briefly review the facts concerning the distribution of the copula verb in Russian. As was already mentioned inrelation to example (28) (see footnote 2), the copula verb is always omitted in present sentences with SFs. In sentenceswith a past reference, the copula is mostly optional with SF participles and obligatory with SF adjectives, as exemplified in(34a) vs. (34b), respectively23:

(34) a. Dom (byl) postroen v proslom godu.house.NOM.SG.MASC. (was) built.SF.SG.MASC. in last year‘The house was built last year.’

b. Rebenok *(byl) bolen v proslom godu.child.NOM.SG.MASC. (was) sick.SF.SG.MASC. in last year‘The child was sick last year.’

The relevant difference between participles and adjectives seems to be that the former, but not the latter, already encodecertain information about temporal reference.24 SF adjectives, however, do not convey any temporal information, hencethe obligatory presence of the copula in (34b) in order to establish any temporal reference at all.

O. Borik / Lingua 149 (2014) 139--165152

23 A reviewer points out that when the reference time is made explicit by, for instance, a temporal subordinate clause, the copula becomesobligatory in the past sentences with SF participles, as in (i):

(i) Dom *(byl) postroen kogda prislo razresenie.house.NOM.SG.MASC. (was) built.SF.SG.MASC. when arrive.PST.SG.NEUT. permit.SG.NEUT.‘The house was built when the permit arrived.’

Note, however, that an explicit temporal adverb does not have the same effect, cf. (34a) in the main text. This points to the conclusion that thedistribution of a copula verb in past sentences with SF participles is affected by rather complex factors and simply specifying the reference time is notenough. Nonetheless, this intriguing issue is not directly relevant to the current discussion and hence will not be further pursued in the present paper.24 The temporal contribution of a past passive participle in Russian could perhaps be compared to that of English past participles in complextense forms like the past or present perfect. The participle points to the fact that the building event itself took place in the past. Similarly, in the caseof (34a) the past passive participle already conveys the information that the building event occurred in the past, but the state of being built holds inthe present. However, to substantiate these claims, much more empirical evidence should be given than this paper can allow for. Since thecontrast exemplified in (34) is largely orthogonal to the main argumentation, I will not elaborate on it any further.

In syntactic terms, I assume that the inability of SFs to express absolute tense can be captured by postulating thatverbs but not SFs can raise to the Tense projection. Technically, this solution can be implemented by postulating that SFslack a necessary feature specification (e.g., [+/" past]), which needs to be checked/valued in TP. Thus, SFs present anexample of a ‘hybrid’ category which displays some verbal characteristics (such as argument selection, agreementfollowing the pattern of past tense verb forms) but which, unlike verbs, lacks temporal specification and hence cannot beused to express absolute tense. Whatever temporal meaning SFs are able to convey (as in the case of the SF participle in(34a)), this information cannot be contributed by the SF category itself. For participles, I assume that the temporalinformation associated with them, along with some additional verbal characteristics such as possible eventivity, comesfrom the verbal functional structure embedded under the SF node, as I will explain below.

With these observations in mind, let us now turn to a syntactic representation of SF adjectives. Consider first thestructure repeated from (20b) in section 3:

(35) [vsP DPi [vs [!P<i>]]]

This is a representation for SF adjectives proposed by Geist (2010:254). The main idea is to account for the verbalbehavior of SFs by introducing a stative verbal head vs in their representation, which will host an argument selected by theroot P in its specifier. However, as I already pointed out in section 3 (see the discussion of (21)), if this analysis is adopted,the adjectival SF bolen (sick.SF) will be indistinguishable from the verb bolet’ (to be sick), which is also stative and,presumably, would have to be associated with the same structure as given in (35). Besides, nothing in (35) can prevent thevs in the SF derivation from further raising to the Aspect and Tense projections, which would form the functional layerabove the vP. This movement, however, must be blocked for SFs, since they do not express tense, whereas verbs do.

To remedy this matter, I propose the following modification of Geist’s analysis. The full structure of a SF adjectiveshould include a SF affix, which is attached to the verbal projection in (35). This will capture the fact that, even though SFsstart projecting similarly to verbs, they will still end up being different from verbs due to the presence of the SF projection.Thus, the full structure of SF adjectives is given in (36):

(36) [SF [SF [vsP DP [vs [!P]]]]

I adopt Geist’s claim concerning the stative character of SF adjectives. This means, in particular, that I assume that thesemantic function of the vs head is to introduce a stative argument (since SFs of adjectives are always stative) which turnsa property denoting root !P into a stative verb. The semantics of the vs head is thus as follows (after Geist, 2010:255)25:

(37) lP lx ls [s:[P(x)]]where P is a property root with a standard semantic denotation

A syntactic function that I attribute to vs is to provide a functional layer that would allow us to introduce the argumentsselected by the predicative root. Such derivation will make sure that the SF adjective bolen (sick) and the verb bolel (was.sick) show the same selectional restrictions with respect to their external argument (see the discussion of (31)--(32)above). The relevant parts of the syntactic derivations for these two elements are given in (38) with the subject rebenok(child.NOM.SG.MASC.) and a predicative root bole- (sick). For the moment, the function of the SF head itself remainsunspecified, but I will get back to it after SF participles have been discussed.

(38) a. The syntactic derivation associated with the SF bolen (sick) from (31a):SFP

vSF s

vDP s’ rebenok

vs !P "P "x "s [s:[P( x)]] "y [ bole-( y)]

P

O. Borik / Lingua 149 (2014) 139--165 153

25 In all the semantic representations used below the abstraction over times is ignored since it is not directly relevant for the exposition. Thus,these representations are slightly simplified.

b. The syntactic derivation associated with the verb boleet (to be sick) from (32):vsP

DP vs’rebenok

vs !P"P "x "s [s:[P(x)]] "y [bole -(y)]

Both structures in (38) are further merged with Aspect and Tense projections (among other necessary functionalelements), but the verb in (38b) moves up to T, whereas the SF in (38a) does not. In the latter case, a copula verb has tobe inserted to lexicalize the part of structure (in particular, to spell out the T features) that cannot be expressed otherwise.The derivation will eventually yield the past tense form bolel (be sick.PST.SG.MASC.) corresponding to the form in (32)above. The SF form will not undergo this movement, in accordance with the assumptions about the SF head specifiedabove.

Now let us turn to SF participles. Participles seem to have a more complex internal structure compared to adjectives.As we have seen above (i.e., examples (9b) and (10) in section 2, (22) in section 3.2, (24b) in section 3.3), SF participlesallow for a wider range of arguments, including by-phrases and instruments, as well as for the event-related modifiers ofduration and manner, whereas adjectives do not. Moreover, participles inherit the perfective/imperfective value of the verbthey are derived from, which signals the presence of the Aspect projection in their syntactic representation. Theseproperties indicate that SF participles have a rather rich internal functional structure, various parts of which (vP, VoiceP,AspP, etc.) are responsible for hosting various types of arguments and modifiers.

Nevertheless, we can obtain a part of the syntactic representation that will be the same for both SF adjectives andSF participles. There are two important points that unify the uses of SF adjectives and participles. The first one is thatthey exhibit the same verbal traits although they do not inflect for tense. Thus, there is always a minimal vP present intheir internal syntax, which also enables them to realize (some of) their arguments. The second is the SF suffix,which in both cases is attached to a verbal structure, be it an elaborate eventive functional structure which potentiallyincludes relative tense and aspect categories, or a simple verbal structure which results from combining the stative vhead with the root. Thus, the part that is common for both SF adjectives and participles can be represented asfollows:

(39) [SF [SF. . .. [vs/e. . .]]

The differences between adjectives and participles will be reflected in the complexity of their internal syntax. Since ourmain purpose here is to capture the commonalities in the behavior of both participial and adjectival SFs, the mostimportant part of the analysis for our present purposes is to identify a syntactic configuration which can be associated withboth SF participles and SF adjectives, and not the details of the syntactic structure embedded under the SF.

Let me now briefly specify the nature of the SF suffix itself. Morphologically, the SF suffix does not have to be realized.Thus, in the case of adjectives, the SF head will mostly be zero, but in the case of participles the SF suffix is always spelledout as --(e)n or --t. Some examples are provided in (40):

(40) a. SFs and LFs of adjectives!vesel- vesel vesel-yjmerry merry.SF.SG.MASC merry.LF.NOM.SG.MASC.!bel- bel bel-yjwhite white.SF.SG.MASC. white.LF.NOM.SG.MASC.

b. SFs and LFs of participles!sdela- sdela-n sdelann-yjmake made.SF.SG.MASC made.LF.NOM.SG.MASC.!sdvinu- sdvinu-t sdvinut-yjmove moved.SF.SG.MASC moved.LF.NOM.SG.MASC.

Given that this suffix can combine with the full range of different syntactic structures, its semantic function must be ratherabstract. One way of unifying the semantic contribution of the SF suffix is to say that it is an operator that provides anexistential closure of a variable in the input derivation. The variable can vary, however. Thus, if we look back at the

O. Borik / Lingua 149 (2014) 139--165154

adjectival SF derivation in (38a) above, the only variable that can be bound by the SF operator is the state variableprovided by the vs head. In the case of stative verbs, this variable will remain unbound until the independent mechanism ofexistential closure applies (Parsons, 1990).26 For participles, the main function of the SF operator is to existentially bind anexternal argument introduced in the extended functional projection associated with SF participles (Borik, 2013). The mainempirical motivation for this last suggestion is that SF participles, being formed on the basis of transitive verbs whichlexically specify a consequent state, inherit all the semantic properties of the verb, including the ability to refer to or modifythe underlying event and a consequent state (see section 3.3 above). The only element that cannot be realized with a pastpassive participle is the external argument of a verb. This suggests that the semantic contribution of the SF suffix incombination with a participial structure should help to explain how an external argument can be suppressed in thesyntactic and semantic representations.

An alternative27 (and in a sense more uniform) treatment of the SF suffix would be to say that the SF morphology isalways zero and the SF suffix itself has a uniform semantic function of binding an event or state variable provided by theverbal structure it combines with. In this case, the analysis of SF adjectives remains intact, but the participial --(e)n/-tmorphology could then be seen as a spell-out of the vs head in the syntactic structure associated with a participle. Thefunction of existentially binding an external argument with the past passive participle can be attributed to the VoicePhead, which we already assumed to be present in the extended syntactic projection associated with participles. Thebiggest advantage of this analysis is, of course, a more parsimonious treatment of the SF head, but its consequenceswith respect to the availability of state/event modification with eventive participles should be carefully considered. Inparticular, we would expect all the relevant state and event modifiers to be hosted by the extended verbal functionalstructure provided by a participle itself, i.e., the structure below SFP. This makes various predictions concerning thescope of the modifiers.

In this paper, I will not commit to either of the alternative treatments of the SF head outlined above.28 However, it shouldbe noted that both potential analyses are viable and the choice between them should eventually be made on the basis of acareful consideration and evaluation of the empirical consequences and predictions made by these analyses.

To sum up, in this section we have examined the behavior of SFs and established a common part of the syntacticstructure for SF participles and adjectives. In particular, I have proposed that SF participles are verbal by virtue of beingderived on the basis of a relatively extended verbal structure, whereas SF adjectives are verbal because their categorydefining head is a stative verb. Being predicative by nature, SFs will freely appear in predicative position. All SFs arederived by merging a particular verbal structure with the SF head, but unlike verbs, they lack the necessary temporalfeatures which would allow them to raise to the TP to express absolute tense. The arguments of SFs are generated by thesame mechanisms as arguments of verbs, although the precise nature of these mechanisms (i.e., which arguments arelicensed by which syntactic heads) does not concern us here. Thus, the analysis provided in this section, common for bothadjectives and participles, captures almost all the properties of SFs as summarized in Table 1 at the end of section 2. Theonly characteristic of SF that at this point remains unexplained is their inability to function as adnominal modifiers. Toexplain this fact we need to rely on the morphosyntactic difference between SFs and LFs, so let me first introduce myanalysis of LFs.

4.2. LFs

I will argue that there are two ways of deriving LFs. One form, which I call a simple LF, involves a minimal derivation thatcombines a LF affix with an adjectival projection. Recall that by assumption, argument realization becomes possible in anextended (verbal) functional structure. In the absence of this structure, simple LFs do not realize any arguments. SimpleLFs are canonical adjectives and freely appear in both adnominal and predicative position.

Another way is to derive LFs on the basis of a more complex SF structure in adnominal positions, which results in thegeneration of what I call complex LFs. This derivation is triggered by the morphosyntactic requirement of Case agreementbetween a head noun and a modifying phrase. The LF affix is precisely the element that enables Case marking (Halle andMatushansky, 2006) and hence full agreement for both adjectives and participles with a modified noun. This part of theproposal derives the phenomenon of ‘argument inheritance’ with the LF forms in adnominal positions.

O. Borik / Lingua 149 (2014) 139--165 155

26 This analysis might also provide an additional reason for why SFs are not inflected for tense. If the SF head binds a stative variable, then,assuming that the structure merges with the temporal layer where, among other things, a state variable is bound by a temporal operator and a run-time corresponding to the state/event is introduced, this operator will not be able to apply if the state variable has already been bound earlier.27 I am grateful to an anonymous reviewer for this suggestion.28 I will not provide here any further details of the semantic analysis, not only for the usual reasons of limited space and time, but also because afull semantic exposition was never meant to be among the main goals of this paper. Thus, though my informal description of the semantic functionof the SF operator is rather sketchy, it will have to wait until another occasion to be turned into a fully developed analysis.

4.2.1. The case of simple LFsI propose that simple LF adjectives are generated in the following configuration:

(41) [LF -oj [aP [a0 [!P]]]]

An adjectival categorial head combines with the predicative base or root to form an adjective, which is further merged with theLF suffix -oj. Compared to the proposal by Geist in (20a), this derivation has an additional element, namely, the LF affix --ojthat attaches to an adjectival projection. In this paper, I adopt the analysis by Halle and Matushansky (2006) who argue thatthe ‘thematic’ element -oj in (41) (and its allomorphs -ij/-aj/-uj) is the actual LF suffix and its main function is to enable Casemarking of an adjective. Halle and Matushansky (2006:359) show quite convincingly that --oj itself cannot be analyzed as anadjectival suffix, given the presence of real adjective-forming suffixes in various adjectival derivations, e.g.:

(42) bed-a -> bed-ın-oj-e -> bednojemisery-NOM.FEM. poor.NOM.NEUT.lom-at’ -> lom-ık-oj-e -> lomkojebreak-INF. fragile.NOM.NEUT

As shown in (42), affixes like -ın-, -ık- and others are the ones that derive adjectives from nouns and verbs.29 These are thesuffixes that remain in the deadjectival derivations (i.e., when a noun or a verb is formed from an adjective), whereas theLF suffix --oj always disappears in such processes. However, the most important part of Halle and Matushansky’s analysisfor the purposes of this paper is the existence of a thematic element -oj, or, as I will call it here, a LF suffix, which enablesCase marking of LF adjectives.

Apart from the role of -oj, there is one more difference between the structure in (41) and the one given in (20a). In heranalysis, Geist has to postulate that the a0 head does not introduce any specifier to account for the absence of argumentswith LFs in a predicate position. I suggest that the absence of arguments simply comes from the absence of verbalfunctional structure, i.e., vP, VoiceP, AspP, etc., where the required arguments and optional modifiers are introduced. Thisprocess of argument realization is independently motivated in some works on nominalization, such as Borer (2003).

Since --oj is treated as a category neutral LF forming affix, the proposal in (41) can be extended for participles. Insection 2.2, it was illustrated that LF participles, just like LF adjectives, are inflected for Case, so it is reasonable to expectthat the role of affix --oj will be the same in the case of LF participles, i.e., it is attached to enable Case marking.

There are, however, some differences between adjectival and participial LFs, which seem to indicate that the LF suffixis attached to a more complex structure in the case of participles, just like in the case of SFs. In particular, example (23) insection 3.2 was given to illustrate that participles express a consequent state of some event and hence presuppose theexistence of a prior event. At the same time, none of the event related modifiers can be used with LF participles, as we sawin section 3.3 (see, for instance, examples (24a) and (26)). To capture these data, I assume that the adjectival head a0

merged with the verbal predicative base !P in (41) existentially binds an event variable associated with the base. Thisexistential binding of the event variable by a semantic operator associated with the adjectival head a0 is a rather commonclaim in the semantic literature on participle formation (see Gehrke, 2012; McIntyre, 2013; Borik, 2013). Crucially, theadjectival head has to combine with the base at an early stage of the derivation, before the functional structure necessaryfor argument realization can be built. Another effect of such an early introduction of the existential closure is that the event,already bound, becomes unavailable for modification (although its existence can still be inferred), as is indeed the casewith LF participles. Thus, an analysis along these lines will capture the differences in the interpretation between the twoexamples in (23), which contrast a LF participle with a LF adjective.

This, however, raises the question of the nature of !P in (41), which up to now has been loosely called a ‘predicativebase’ or a ‘predicative root’. There are various options to specify what a ‘predicative base’ means precisely, and for the

O. Borik / Lingua 149 (2014) 139--165156

29 A reviewer has pointed out that the examples in (42) suggest a possibility that in the SF derivation associated with adjectives (see (36) and(38a) above), the vs head can also combine with a category specified projection, i.e., a0, The reason for this being that the adjectival affixes -ın-and -ık- from (42a) and (42b), respectively, also form part of SFs associated with these adjectives, as shown below:

(i) a. beden b. lomokpoor.SF.SG.MASC. fragile.SF.SG.MASC.

Note that in the case of (ia), the -ın- affix is also a part of the corresponding verbal structure, i.e., bednet’ (to become poor), so it could be treated asa vs morpheme in a structure corresponding to (38a), but the example in (ib) indeed suggests that we might have to allow for vs to merge with a0 aswell as with !P. The issue, however, does not seem to have any dramatic consequences for the analysis proposed. In particular, the semanticsassociated with (38a) remains intact.

purposes of this paper I assume that it is a lexical element whose semantic representation includes lambda-boundarguments, including the event argument, should it be present. Some possible P elements could be represented asfollows:

(43) a. lx[P(x)]b. lxle[P(x,e)]

The representation in (43a) corresponds to a simple property, while the one in (43b) corresponds to an eventive elementwith one argument. Then, if the adjectival head a0 in (41) applies to a P whose denotation is a simple property, the resultwill be an adjective that just denotes a property. In this case, the main function of the adjectival head is simply to provide asyntactic categorization of the predicative root. If, however, it applies to a lexical element whose semantic representationincludes an event variable, it will existentially bind the event and the result of the derivation will be an adjectival constructwith a bound event variable, i.e., a participle.30 The main idea, thus, is that a predicative root or base is a lexical elementwhich is specified for certain arguments, but which cannot introduce them. Once again, I hold to the assumption that thearguments themselves are introduced in the functional syntactic structure.

This being said, the question of semantic distinctions between different types of LF participles and adjectives inRussian is quite a complex issue which still needs more in-depth research. In particular, it seems that perfective pastparticiples are more ‘eventive’ than imperfective ones, which are much closer to adjectives, since perfective participlesalways presuppose the existence of a prior realized event (see again the discussion of example (23a) in section 3.2). Aprecise implementation of this intuition has, to the best of my knowledge, never been proposed. In traditional grammars,imperfective participles are simply referred to as adjectives. In general, imperfective based past passive participles exhibitmany characteristics that are traditionally associated with lexical, not syntactic, derivations such as idiosyncratic meaning,use in idiomatic constructions, etc. Thus, a search in the Russian National Corpus for the participle delannyj (made.LF.sg.Nom) derived from the imperfective verb delat’ (to do/make) gives 39 results, most of which are uses with the figurativemeaning ‘not real, artificial, feigned’ (see (44a)). Only very few examples feature this participle in a meaning clearly relatedto the underlying verb, as in (44b).

(44) a. Capaev smejalsja, no ego smeh byl delannyj.Capaev laugh.PST.SG.MASC. but his laugh.NOM. be.PST.SG.MASC. feigned.LF.SG.NOM.‘Capaev laughed, but his laugh was feigned.’

b. Eto byl portret Anny, delannyj v Italii Mihailovymthis be.PST.SG.MASC. portrait.NOM. Anna.GEN. done.LF.SG.MASC.NOM in Italy Mihailov.INSTR.‘This was a portrait of Anna made in Italy by Mikhailov.’

From the morphological perspective, it is clear that the LF delannyj (done) should be analyzed as derived from the verbdelat’, although semantically, the link between the underlying verb and the participle is weaker. The problems presentedby LF participles would seem to be highly relevant for the syntactic vs. lexical perspective on word formation. At this point, Iwill not go into the details of this discussion, because it will divert us too far from the main subject of this paper, but theproblems just mentioned indicate a need for a theory that provides an independent, principled motivation for distinguishingbetween different levels of word and phrase formation.

To sum up, this section was devoted to the analysis of simple LFs that I propose. I suggest that simple LFs are derivedby means of the LF suffix --oj which is merged with the category defining adjectival affix a0 to enable the Case marking(Halle and Matushansky, 2006). As adjectival elements, simple LFs can be used predicatively or adnominally. None of theverbal functional structure where the arguments and event-related modifiers can be introduced is present in the derivation,hence LFs do not realize any arguments that the corresponding SFs may have.

If we now go back to Table 1 in section 2, which summarized the empirical data to be accounted for, we can see that bynow we have reached the following results: a predicative use and the argument realization pattern of SFs were accountedfor in the previous section, and this section explained how the LFs without arguments are derived and used. Thus, whatremains to be explained is the adnominal uses of LFs with their argument realized and the ban on SFs in the sameposition. I think that these two facts are two sides of the same coin. In the next section, we tackle this remaining issue.

O. Borik / Lingua 149 (2014) 139--165 157

30 A reviewer raises a question as to whether the adjectival head a0 should be of a variable semantic type. I do not see any other option, if the a0

element is taken to be semantically loaded when it combines with both (43a) and (43b). As I suggested in the text, however, a0 can possibly besemantically vacuous in the former case. Then an option of providing a uniform semantics for a0 will depend on further careful examination ofpossible input structures.

4.2.2. The case of complex LFsThe defining properties of complex LFs are that they always appear in adnominal positions and do not have any

restrictions on argument realization. I will argue that these LFs are in fact disguised SFs with the affix --oj attached to themfor reasons of Case agreement, since the rules of Russian morphosyntax require that adnominal phrases headed byadjectives and participles agree with a head noun not only in number and gender (a requirement that could be fulfilled by‘pure’ SFs as well), but also in Case. Crucially, the LF suffix is the morphological element that facilitates Case marking.Thus, the ban on SFs in adnominal position should be interpreted not literally as a prohibition31 but rather as a requirementfor a SF to take a specific morphosyntactic form, namely, a LF.

I assume that syntactically complex modifiers, including adjective and participial phrases with their arguments, aremerged into a specifier position of a functional projection dominated by the DP, following a proposal by Sleeman (2011).The corresponding structure is represented in (45b) below for the English example in (45a)32:

(45) a. the recently sent bookb. [DP the [FP [AspP recently [vP [sent]]] [F’ [book]]]]

On the basis of the data from Dutch and English, Sleeman convincingly argues that prenominal adjectives and participlescan be eventive, contrary to Embick (2004). The Russian data fully supports her conclusion, as illustrated in (46) below, sonot every prenominal participle can be considered stative simply due to its position. Thus, in (46), the participial phrasemodifying the noun vyvod (conclusion) in this example includes both a by-phrase and an adverb nedavno (recently), bothof which are clear indications of the eventive character of the modifying participial phrase.

(46) Polu!ennye resul’taty podtverdili nedavno sdelannyj u!enymireceived.LF.PL.NOM. result.PL.NOM. confirm.PST.PL. recently done.LF.SG.MASC.ACC. scientist.PL.INSTR.vyvod.conclusion.SG.MASC.ACC.‘The obtained results confirmed the conclusion recently made by the scientists.’

I propose that in complex LFs like the one in (46), the LF suffix --oj is merged on top of the SF structure to enable Caseagreement with the head noun. Since the internal syntax of all SFs includes a (minimal or extensive) verbal functionallayer, we expect the arguments to be freely realized in this configuration. The full structure of a complex LF is thus asfollows:

(47) [LF -oj [SF [ØSF. . .. [vs/e. . .]]

This structure can only be generated in a specific syntactic environment, though. Crucially, the LF suffix can merge withthe SF structure only in agreement environments, i.e., only in adnominal positions, since the rules of Russianmorphosyntax require a head noun and its modifier to always agree in Case, in addition to other features. Thus, wecan analyze complex adnominal LFs as a case of context driven coercion, where a SF is simply forced into a LF to facilitateCase agreement in a specific syntactic context, namely, in adnominal position.33 In other words, coercion here describes ashift from a SF to a LF which is required in a given syntactic environment under the requirements imposed by theimmediate context.

An immediate question that arises at this point is why the same derivation cannot occur in predicate position. In otherwords, we should find a way to rule out (47) in predicate position, otherwise we would predict that a complex LF with(potentially) realized arguments can also appear as predicate. Logically, nothing prevents this possibility, i.e., there is noindependent motivation for why (47) should not be allowed in predicate position. However, I do believe that not all logical

O. Borik / Lingua 149 (2014) 139--165158

31 Note that in some other Slavic languages, such as Serbo-Croatian, both short and long forms can appear in adnominal position.32 Two comments are in order here. Firstly, Sleeman provides different analyses (and different structures) for pre- and post-nominal modifiers.The Russian data do not provide any evidence for postulating this difference (see footnote 5), so I will not follow Sleeman in this part of herproposal. Secondly, I will not further specify the nature of the functional projection FP in (45). It could be an agreement phrase (AgrP), but in orderto make a sound decision a much more detailed discussion about the internal DP syntax in Russian is needed, a discussion I will not enter into forreasons of space.33 The term ‘coercion’ is used rather loosely here. In principle, coercion usually describes various types of semantic shifts (e.g., type coercion inPustejovsky, 1995, aspectual coercion in de Swart, 1998, adjective meaning coercion in context in Partee, 2010). Here, I have simply borrowedthe term to describe the same general type of phenomena (i.e., a forced shift) in a particular syntactic configuration.

possibilities can necessarily be realized if different factors interfere. In this case, I suggest that the structure in (47) doesnot occur in predicate position due to economy considerations.

Recall that in section 3.1 I argued against the attributive hypothesis for LFs. In particular, arguments that were providedin that section allowed me to reject the hypothesis that when we see an LF in a predicate position it is, in fact, a modifier ofan empty nominal head. Thus, we do not have any reason to postulate that a LF in predicate position functions as amodifier. If it is not a modifier, it does not have to agree with any noun and hence, it does not have to bear any Casespecification. Predicate position itself does not pose any Case requirements either: as we have seen throughout thispaper, SF adjectives and participles that cannot be marked for Case freely appear in predicate position. Thus, predicateposition is not a kind of syntactic environment which would impose a requirement to turn a SF into a LF. If no suchrequirement exists, the corresponding derivation does not occur simply because it appears unnecessary in a givensyntactic context. This means that a SF appears as a SF in predicate position, and additional LF morphology does nothave to be added to enable Case agreement simply because Case agreement is not required in this position.

Notice that Case is a crucial element here, since, as was shown in section 2, SFs can be inflected for both number and,in singular, gender, so the insertion of the LF affix would not be necessary at all if it was not for Case requirements inadnominal position. Thus, once again, SFs can appear in the predicate positions, just like verbs do, because they do nothave to agree in Case with anything else.

The idea of deriving LF forms in adnominal positions on the basis of complex syntactic structures is, in principle, notentirely new. Bailyn (2012:71) proposes a derivation of the complex modification structure along the same lines. Hisanalysis, however, does not explain what triggers this derivation in adnominal positions, whereas my proposal makes avery clear claim that the addition of the LF suffix in this syntactic environment is due to the Case agreement being imposedon the modifying head.

To summarize the analysis of LFs in general, I have argued that there are two types of adjective and participle LFs.Simple LFs are derived by adding a LF suffix to an adjective phrase which consists of a categorizing head a0 and apredicative base P it combines with. No arguments can be introduced in this configuration due to the absence of the verbalfunctional structure. Complex LFs, on the other hand, are derived on the basis of the SF structures in attributive positionsdue to the Case agreement requirements. The LF suffix --oj is the element that enables Case marking, and it has to be partof the structure where Case agreement is enforced between a modifying phrase and a head noun. The arguments of acomplex LF can be expressed because this form is built on the basis of SF structures, which allows for argumentrealization. Thus, a complex adnominal LF is in fact a SF in disguise. Both simple and complex LFs can appear in modifierpositions, but there is no reason to derive a complex LF in a predicative position. Thus, we have by now accounted for allthe SF and LF occurrences as summarized in Table 1 of section 2.

4.3. Two types of LFs

In the previous two sections I argued that there are two types of LF adjectives and participles: a simple LF and acomplex LF. The distinction is reminiscent of two types (or two sources) of modification postulated by Cinque (2010):direct and indirect modifiers. The purpose of this section is to briefly compare two types of LFs with different types ofCinque’s modifiers.

In Cinque’s proposal, direct modifiers are different classes of adjectival phrases which merge directly in the functionalstructure of an NP and behave like regular modifiers. They are not derived from more complex structures. Indirectmodifiers are those which are derived from relative clauses, so that the indirect modifier itself is an instance of a reducedrelative clause. A number of different interpretive properties are associated with each class of modifiers. Thus, directmodifiers are usually individual level, non-restrictive, modal, non-intersective etc.34

As I already briefly mentioned in footnote 6 of section 2.1, adnominal LFs with arguments can indeed be viewed asreduced relative clauses. However, and crucially, in a full relative clause counterpart of a complex LF we would still have touse a corresponding SF in predicative position. The pattern is exemplified once again in example (48) below, with acomplex LF in pre- and postposition illustrated in (48a) and (48b), respectively, vs. a full relative clause in (48c). I take thiscontrast to reflect a structural difference between predicate position in a clausal structure and a complex DP-internalmodifier. Note that these data fit the analysis that I proposed for complex LFs treated as modifiers derived from SFs, whichwould otherwise be used in predicative position.

(48) a. Razrusennyj stroiteljami fasad budet vosstanovlen.destroyed.LF.NOM.SG.MASC. builder.PL.INSTR. facade.NOM.SG.MASC. will reconstructed.SF.SG.MASC.

O. Borik / Lingua 149 (2014) 139--165 159

34 For a full list and a detailed summary of properties distinguishing direct modifiers from indirect ones see Cinque (2010, chapters 2 and 3).

b. Fasad, razrusennyj stroiteljami, budet vosstanovlen.facade.NOM.SG.MASC. destroyed.LF.NOM.SG.MASC. builder.PL.INSTR. will reconstructed.SF.SG.MASC.‘The facade destroyed by the builders will be reconstructed’.

c. Fasad, kotoryj byl razrusen/ *razrusennyj stroiteljami,facade.NOM.SG.MASC. which was destroyed.SF.SG.MASC/ destroyed.LF.NOM.SG.MASC. builder.PL.INSTR.budet vosstanovlen.will reconstructed.SF.SG.MASC.‘The facade which was destroyed by the builders will be reconstructed’.

The pattern in (48) can also serve a straightforward illustration for Cinque’s analysis of indirect modifiers, where a SFwhich appears as the predicate of a full relative clause turns into a reduced structure (i.e., a relative missing a C and a Vhead) with a LF in adnominal position. Thus, at least at first sight and from a purely structural viewpoint, the parallelbetween complex LFs and indirect modifiers is quite clear. Now let me demonstrate that various properties of direct andindirect modification are indeed found in simple vs. complex LFs, respectively.

To start with, Cinque notes that only direct modifiers can acquire metaphorical or idiomatic meanings. In Russian, onlysimple LFs can give rise to idiomatic interpretations, LFs with realized arguments (i.e., our complex LFs) lose these non-literal meanings. This can be illustrated by means of the examples based on (31) in section 4.1, where it was shown thatLFs do not impose selectional restrictions on their arguments in the same sense that SFs and corresponding verbs do.Consider (49). A simple LF in (49a) can modify both nouns mal’!ik (boy) and golos (voice), but a complex LF with anargument in (49b) can only be used as a modifier of the first noun, since the second one is not selected by thecorresponding SF (cf. the ungrammaticality of (31d) in section 4.1). Moreover, an additional idiomatic interpretation isavailable for the phrase bol’noj mal’!ik (sick boy) in (49a), meaning something like ‘an insane boy’, i.e., a boy that behavesout of the ordinary. This interpretation disappears in (49b) with a complex LF.

(49) a. Bol’noj mal’!ik/ golossick.LF.NOM.SG.MASC. boy.NOM.SG.MASC./ voice.NOM.SG.MASC.‘a sick boy/an insane boy/a sick voice’

b. Bol’noj anginoj mal’!ik/ *golossick.LF.NOM.SG.MASC. quinsy.INSTR. boy.NOM.SG.MASC./ voice.NOM.SG.MASC.‘a boy/*a voice sick with quinsy’

Cinque also notes that when different types of modifiers are combined, the relative order is always fixed with indirectmodifiers preceding direct ones. In Russian the order of modifiers is usually not so strict, but if the same adjective is used, itis clear that the complex LF must precede the simple one:

(50) a. bol’noj anginoj bol’noj mal’!iksick.LF.NOM.SG.MASC. quinsy.INSTR. sick.LF.NOM.SG.MASC. boy.NOM.SG.MASC.

b. *bol’noj bol’noj anginoj mal’!iksick.LF.NOM.SG.MASC. sick.LF.NOM.SG.MASC. quinsy.INSTR. boy.NOM.SG.MASC.‘a sick/insane boy sick with quinsy’

In principle, an LF without any arguments in adnominal position will be systematically ambiguous between a simple and acomplex type, and hence, we can expect these LFs to display properties of both direct and indirect modifiers, which will notallow us to unequivocally classify them in Cinque’s sense.35 The ambiguity disappears only if a LF is used with itsarguments, in which case it can only be interpreted as a complex LF. Some contrasts can indeed be brought out bycomparing LFs with and without arguments in the same context, as in, for instance, example (51) below:

(51) a. vezlivyj mal’!ikpolite.LF.NOM.SG.MASC. boy.NOM.SG.MASC.‘a polite boy’

b. vezlivyj s roditeljami mal’!ikpolite.LF.NOM.SG.MASC. with parent.INSTR.PL. boy.NOM.SG.MASC.‘a boy polite with his parents’

O. Borik / Lingua 149 (2014) 139--165160

35 Cinque (2010, section 6 of the Appendix) suggests the same with respect to Russian LFs, namely, that they can have two sources, a directmodifier and an indirect modifier.

With respect to these two examples, it should be observed that the adjective vezlivyj (polite) in (51a) can have either anabsolute (i.e., polite as an absolute value) or a relative interpretation (i.e., polite with respect to other kids), with an absolutereading being slightly more dominant if no further contextual information is available. In Cinque’s proposal, an absolutereading is available for direct modifiers but not for indirect ones. In accordance with this claim, a complex LF in (51b) canonly have a relative interpretation.

Moreover, the LF in (51a) can get either a restrictive (i.e., polite for a boy) or a non-restrictive (i.e., it is a boy and he ispolite) interpretation (see Cinque, 2010; Larson 1998), whereas in (51b) a restrictive reading is much more difficult toobtain, if possible at all.36 Once again, the complex LF in (51b) patterns with indirect modifiers, whereas the LF in (51a) isambiguous and can be associated with both a direct and an indirect modification source.

Thus, as I hope to have illustrated in this section, there are substantial reasons to believe that the two types of LFcorrespond to the two modification sources postulated by Cinque (2010). The precise extent to which the observedparallels could and should be extended depends on what other properties of (in)direct modification can be found in simple/complex LF. This question, however, requires a much more detailed and thorough analysis than can be carried out withinthe limits of this paper, although it is clear that the issue would be well worth exploring.

4.4. A problem of depictives

A reviewer has pointed out that the following construction is problematic for the hypothesis that complex LFs arise onlyas a result of the Case agreement requirement:

(52) Petja prisel gotovym k drakePeter.NOM.SG. come.PST.SG.MASC. ready.LF.INSTR.MASC. to fight.DAT.‘Peter came ready to fight.’

The example above features a subject-oriented depictive predicate gotovym (ready) in Instrumental Case, whereas thesubject of the sentence itself is in Nominative. The depictive adjective agrees with the subject in number and gender butnot in Case, although its argument ‘fight.DAT.’ is nevertheless realized. If the same adjective occurs in predicative positionin a copular sentence, the argument is not licensed, in accordance with the expectations:

(53) Petja byl gotovym (*k drake)Peter.NOM.SG. be.PST.SG.MASC. ready.LF.INSTR.MASC. (to fight.DAT.)‘Peter was ready to fight.’

As the reviewer stated it, ‘‘there is no way of analyzing it (i.e., the LF adjective in (52)) as resulting from agreement with anominal head’’. I think that this and similar examples appear to present a critical problem for the analysis of complex LFadjectives developed in this paper, but a solution can still be found. In what follows, I would like to indicate some possibleways of approaching the puzzle of complex depictive LFs in Russian,37 although I do not claim to provide acomprehensive solution. In particular, I would like to emphasize two points in the discussion below. The first one isthat the agreement fails to obtain only in a restricted set of cases, whereas in the majority of examples a depictiveobligatorily appears in exactly the same Case as the antecedent. Thus, Case agreement can be viewed as a defaultoption, whereas Instrumental form appears only when certain additional conditions obtain (see Timberlake, 1986,Hinterhölzl, 2001).38 The second point is that more then one factor may be involved in determining the Case marking of thedepictive predicate in (52) and until these factors are fully understood we cannot simply state that ‘‘there is no way ofanalyzing it as agreement’’.

To start with the first point, it should be noted that in Russian (as in many other languages: see Marusi! et al., 2004 andthe literature cited therein), depictive predicates are not only subject-oriented or object-oriented but can be hosted by otherarguments as well, in particular, by indirect objects and prepositional phrases. In this case, a depictive adjective must

O. Borik / Lingua 149 (2014) 139--165 161

36 In my own judgment, a restrictive reading in (51b) is simply not available. However, given that this is not an easy judgment to give, I think asolid generalization can only be made after consulting with many more native speakers.37 The issue is very poorly understood, which is why it remains a puzzle. Note that the pattern exemplified in (52) has also always been a problemfor Babby’s attributive hypothesis, discussed and rejected in section 3.1.38 Nonetheless, this is a position not shared by everybody. For example, Bailyn (2001) argues that Case agreement (or the ‘Sameness of case’,in his terminology) is an epiphenomenon, whereas Instrumental is assigned to all secondary predicates in Russian as default. A brief discussion ofBailyn’s account is provided below.

bear the same Case as the antecedent and no variation is ever observed or allowed, as illustrated in the examplebelow39:

(54) Ona prisla ko mne bol’nomu/ *bol’noj/ *bol’nogo/she.NOM.SG. come.PST.SG.MASC. to me.DAT. sick.LF.DAT.MASC./ sick.LF.NOM.MASC/ sick.LF.ACC.MASC./*bol’nymsick.LF.INSTR.MASC

‘She came to me and(/when) I was sick.’

Secondly, even in the case of subject- or direct object-oriented depictives, the Case form of the depictive predicate itselfcan vary between Nominative and Instrumental, or Accusative and Instrumental, respectively. Thus, in the example in (52)above, the depictive can also have Nominative Case, as illustrated in (55):

(55) Petja prisel gotovyj k drakePeter.NOM.SG. come.PST.SG.MASC. ready.LF.NOM.MASC. to fight.DAT.‘Peter came ready to fight.’

These two facts seem to provide evidence against the hypothesis that secondary predicates in Russian are assignedInstrumental case by default, as proposed, for instance, by Bailyn (2001). Bailyn’s account is based on a central claim thatin Russian, as opposed to some other Slavic languages like Polish where secondary predicates normally bear Nominativecase, Instrumental is assigned to predicative AP or NP by the Pred(icative) head, which is lexically specified as Instr caseassigner, as in the following configuration (Bailyn, 2001:8):

(56)

Pred’

PredP

Spec

Pred 0 NP/AP

Instr assigned

Even stronger empirical objection against Bailyn’s analysis than what is described above come from the fact that inprimary predication, which is presumed to have the same structure involving PredP, the following pattern is observed:

(57) a. Ivan durak/ *durakomIvan fool.NOM.SG./ *fool.INSTR.SG.‘Ivan is a fool.’

b. Ivan byl durak/ durakomIvan was fool.NOM.SG./ fool.INSTR.SG.‘Ivan was a fool.’

Bailyn’s claim is that overt elements occupying the Pred head position can absorb Instrumental Case, in which case thecomplement NP/AP will be assigned a structural Nominative Case. Clearly, this does not explain the pattern in (57). In(57a), a present sentence with the null copula, there is no overt element in the Pred position and the predicative noun is inthe Nominative Case, whereas we would expect Instrumental, which turns out to be ungrammatical. In (57b), by contrast,the expectation is to have Nominative on the predicative noun, since the overt copula in the past occupies the Predposition, but we can also have Instrumental, which is not predicted by Bailyn’s account in this configuration. Thus, Iconclude that this particular analysis, as well as the general idea that the Instrumental case is assigned by default toprimary and secondary predicates in Russian, does not seem to fully agree with the empirical evidence.

With this conclusion in mind, let me now outline some theoretical considerations concerning depictive predicates.In general, I assume that depictives are a type of secondary predicate for which a small clause (SC, Williams, 1975)

analysis can be adopted. In one of the first analyses of SCs, Stowell (1983) suggested that they should be represented asmaximal projections of a lexical category corresponding to the depictive predicate, i.e., AdjP in the case of the examples

O. Borik / Lingua 149 (2014) 139--165162

39 See also a nice selection of examples with depictive predicates related to the complements of various prepositions in Babby (2010:79).

provided above. However, as has been shown in many subsequent works on secondary predication, the representation ofSCs must include some (minimal or extended) functional structure, which, among other things, would help to integrate aSC into the main sentence. In some analyses of secondary predication it was argued that the functional structure of a SChas to incorporate Agreement projections (see Guéron and Hoekstra, 1994, Legendre, 1997). Thus, the structure in (58b)can be attributed to the SC (italicized) part of the sentence in (58a)40:

(58) a. We consider Bill foolish.b. [Billi Agr [AP ti foolish]]

Clearly, the agreement facts of the Russian depictives will follow if the structure with the AgrP as part of the SCconfiguration is adopted. Moreover, if the head noun and a secondary predicate originate in the same syntactic category(see Marusi! et al., 2004 for a specific proposal involving a Dep(ictive) Phrase), this provides an additional reason toexpect agreement between the two elements in all the relevant morphosyntactic characteristics, including Case.However, the sentences where the Case marking on the main subject and the depictive adjective or participle do notcoincide (like in example (52) above) remain a problem. In what follows, I would like to indicate some possible ways ofsolving it. Note that this discussion does not claim to be anything but suggestive. However, I believe that the generalreasoning I propose below is plausible and hence could be developed into an analysis which would explain the behavior ofdepictive predicates.

First of all, notice that the Case discrepancy only arises when the antecedent of a depictive appears in subject or directobject position, i.e., those positions where a structural Case is obligatorily assigned. It could be the case, then, thatstructural Case configurations ‘overrule’ the initial Case assigned to the head noun (which would be Instrumental to beginwith) before it has been raised from the SC. If this line of reasoning is adopted, a proposal to be developed would have toboth specify a precise mechanism of Case reassignment and provide an independent empirical motivation for such aprocess to take place.

Another possible way is to explore systematic semantic differences that nouns and adjectives exhibit in predicateposition depending on which Case they bear, Nominative or Instrumental. One of the hypothesis (Pereltsvaig, 2001;Bogatyreva, 2011) is that Nominative Case corresponds to an identity interpretation contributed by a predicative element,whereas Instrumental corresponds to an interpretation of ascribing a property to the subject. It could be argued that theInstrumental Case on a depictive head appears under specific semantic requirements, i.e., it is there to render aninterpretation that a predicate in the Nominative Case is unable to express (Timberlake, 1986; Hinterhölzl, 2001). In thiscase, Case agreement between the head noun and a depictive head could be overruled due to the interpretational needs.This option seems most plausible to me, at least at first sight, especially since it builds on already established semanticdifferences between Nominative and Instrumental Cases in predicate position.

Thus, as should be clear from this short exposition, additional factors that could play a role in determining the Casemarking options for both a depictive predicate and an antecedent noun are both the structural position of the noun itselfand a semantic interpretation of the depictive. In any case, my general idea concerning depictive predicates would be thatat least at some point of the derivation, when both an antecedent noun and a depictive adjective or participle still form partof the same SC configuration and possibly the same maximal projection, agreement between the two elements can (andshould) obtain. The choice between the options for an analysis of depictives which I have listed here (plus other potentialoptions that I have probably overlooked) has to be made on the basis of empiricial and theoretical considerations that Icannot further explore here. A detailed analysis of depictives is, thus, left for a future occasion.

5. Concluding remarks

In this paper, I have provided an analysis which explains how short and long forms of participles and adjectives inRussian are used and realize their arguments. We started off with a set of data to be accounted for, which wassummarized in Table 1 in section 2, and the analysis developed in section 4 gradually explained all the data from that set.

The main points of the analysis are as follows. SFs display a number of verbal characteristics and are able to realizetheir arguments due to the presence of the verbal functional layer, which is either minimal in the case of SF adjectives ormore extended in the case of SF participles. The main difference between SFs and verbs is that the latter, but not the

O. Borik / Lingua 149 (2014) 139--165 163

40 Other proposals for the exact composition of the functional layer of a SC include full sentence structure, including TP and CP (Starke, 1995), aspecial predication phrase (Bowers, 2001), etc. Moreover, a standard (at least until recently) analysis of SC also postulates a PRO as a subject ofa SC (see Stowell, 1983; Bowers, 2001 and many others). The presence or absence of PRO does not influence the argumentation in this section:in the case of PRO, which is coindexed with the main subject, the agreement is established between the main subject, PRO and the head of theAP or PartP phrase. However, see Marusi! et al. (2004) for (convincing, in my view) arguments against the PRO analysis of depictives.

former, can express absolute tense. Syntactically, this means that SFs, being deprived of temporal features, cannot raiseto the T projection, hence a copula verb has to be used in sentences with SFs to render absolute tense. As for LFs, I haveargued that there are two possible derivations that can be associated with them. Simple LFs do not have argumentstructure and do not have any restrictions on the syntactic positions they can appear in. LFs derived on the basis of SFs, orcomplex LFs are morphologically indistinguishable from their simple counterparts, but the complex forms only arise if a SFstructure needs to be used adnominally. In this case, a LF suffix needs to be attached to a SF to enable this form to carryCase morphology, which is obligatorily required to ensure full agreement between a head noun and its modifier. LFsformed in this way can always realize their arguments because underlyingly they are SFs. Since the predicate positiondoes not pose any Case agreement requirement, complex LFs can only appear adnominally. I have also shown that thereare good reasons to believe that the two types of LFs, namely, simple vs. complex LFs, can be viewed as an instantiationof Cinque’s (2010) distinction between direct and indirect modification sources, respectively.

To return once again to the set of data we started with, although Table 1 at first sight seems to exhibit a certainasymmetry, with SFs which appear to be banned from adnominal position and LFs realizing their arguments in the sameposition, the analysis developed in section 4 allows us to account for these data in a different way. In particular, it could besaid that the distribution of SFs and simple LFs is actually fully symmetrical: both can appear in adnominal and predicativeposition, but SFs consistently realize their arguments, whereas (simple) LFs do not. However, when a SF is usedadnominally, as a modifier, it is forced to take a LF form to facilitate Case agreement with a head noun. This leads to aderivation of what has been called here a complex LF, which is, in essence, a SF with an additional LF marker that enablesCase marking.

Acknowledgments

Part of this paper was presented at the Workshop on Aspect and Argument Structure of Adjectives and Participles(WAASAP) held in June 2012 at the University of Greenwich, London. I would like to thank the organizers of the workshopand the editors of this volume for putting it all together, and the audience at the workshop for their questions andcomments.

I am very grateful to three anonymous reviewers for their criticism, questions and suggestions which I believe have ledto a much improved version of the paper. All remaining errors are, of course, my own.

This research has been funded by a research grant awarded by the Spanish Ministerio de Ciencia e Innovacion(FFI2011-23356) and by a grant awarded by the Generalitat de Catalunya to the Centre de Linguística Teòrica (2009SGR-1073).

References

Alexiadou, A., 2001. Functional Structure in Nominals: Nominalization, and Ergativity. John Benjamins, Amsterdam.Alexiadou, A., Anagnostopoulou, E., 2008. Structuring participles. In: Chang, C., Haynie, H. (Eds.), Proceedings of the 26th West Coast

Conference on Formal Linguistics, Cascadilla Proceedings Project, Somerville, MA, pp. 33--41.Alexiadou, A., Anagnostopoulou, E., Schäfer, F., 2006. The properties of anticausatives crosslinguistically. In: Frascarelli, M. (Ed.), Phases of

Interpretation. Mouton, Berlin, pp. 187--212.Arad, M., 1999. On ‘little v’. In: Arregui, K., et, al. (Eds.), Papers on morphology and syntax. Cycle one, MITWPL 33, Cambridge, pp. 1--25.Babby, L.H., 1975. A Transformational Grammar of Russian Adjectives. Mouton, The Hague.Babby, L.H., 1999. Adjectives in Russian: primary vs. secondary predication. In: Dziwirek, K., et al. (Eds.), Formal Approaches to Slavic

Linguistics (FASL-7). The Seattle Meeting 1998, vol. 35. Michigan Slavic Materials, Ann Arbor, pp. 1--30.Babby, L.H., 2009. The Syntax of Argument Structure. Cambridge, CUP.Babby, L.H., 2010. The syntactic differences between long and short forms of Russian adjectives. In: Cabredo Hofherr, P., Matushansky, O.

(Eds.), Adjectives. Formal Analyses in Syntax and Semantics. John Benjamins, Amsterdam/Philadelphia, pp. 53--84.Bailyn, J., 1994. The syntax and semantics of Russian long and short adjectives: an X’-theoretic account. In: Toman, J. (Ed.), Formal Approaches

to Slavic Linguistics (FASL-1) 1992. The Ann Arbor Meeting, Ann Arbor, pp. 1--30.Bailyn, J., 2001. The syntax of slavic predicate case. ZAS Pap. Linguist. 22, 1--23.Bailyn, J., 2012. The Syntax of Russian. Cambridge University Press, Cambridge.Bogatyreva, L., 2011. Russian Instrumental Nominals as a Counterpart of Bare Nominals in Romance (M.A. thesis). Barcelona, Universitat

Autònoma de Barcelona.Borer, Hagit, 2003. Exo-skeletal vs. endo-skeletal explanations: syntactic projections and the Lexicon. In: Polinsky, M., Moore, J. (Eds.), The

Nature of Explanation in Linguistic Theory. CSLI, Stanford, pp. 31--67.Borik, O., 2013. Past participle formation and the eventive/adjectival passive in Russian. In: Chemla, E., Homer, V., Winterstein, G. (Eds.),

Proceedings of Sinn und Bedeutung, vol. 17. pp. 115--132.Bowers, J., 2001. Predication. In: Baltin, M., Collins, C. (Eds.), The Handbook of Contemporary Syntactic Theory. Cambridge, MA, Blackwell.Cinque, G., 2010. The Syntax of Adjectives: A Comparative Study. LI Monographs, MIT Press, Cambridge, MA.Dixon, R.M.V., 2004. Adjective classes in typological perspective. In: Dixon, R.M.V., Aikhenvald, A. (Eds.), Adjective Classes. A Cross-linguistic

Typology. Oxford University Press, Oxford, pp. 1--49.

O. Borik / Lingua 149 (2014) 139--165164

Embick, D., 2004. On the structure of resultative participles in English. Linguist. Inq. 35.3, 355--392.Gehrke, B., 2012. Passive states. In: Demonte, V., McNally, L. (Eds.), Telicity, Change, and State: A Cross-categorial View of Event Structure.

Oxford University Press, Oxford, pp. 185--211.Geist, L., 2010. The argument structure of predicate adjectives in Russian. Russ. Linguist. 34, 239--260.Gorskova, K.V., Haburgaev, G.A., 1981. Istori!eskaja grammatika russkogo jazyka. Vyssaja skola, Moskva.Guéron, J., Hoekstra, T., 1994. The temporal interpretation of predication. In: Cardinaletti, A., Guasti, M.T. (Eds.), Syntax and Semantics: Small

Clauses, vol. 28. Academic Press, San Diego.Halle, M., Matushansky, O., 2006. The morphophonology of Russian adjectival inflection. Linguist. Inq. 37 (3), 351--404.Hinterhölzl, R., 2001. Semantic constraints on case assignment in secondary adjectival predicates in Russian. ZAS Pap. Linguist. 22, 99--112.Kratzer, A., 2000. Building Statives. Ms. University of Massachusetts, Amherst.Krifka, M., 2000. Alternatives for aspectual particles: semantics of still and already. In: Conathan, L., Good, J., Kavitskaya, D., Wulf, A., Yu, A.

(Eds.), Proceedings of the Twenty-sixth Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on Aspect.University of California, Berkeley, CA, pp. 401--412.

Larson, R., 1998. Events and modification in nominals. In: Strolovitch, D., Lawson, A. (Eds.), Proceedings from Semantics and Linguistic Theory(SALT) VIII. Cornell University, Ithaca, NY, pp. 145--168.

Legendre, G., 1997. Secondary predication and functional projections in French. Nat. Lang. Linguist. Theory 15 (1), 43--87.Löbner, S., 1989. German schon -- erst -- noch: an integrated analysis. Linguist. Philos. 12, 167--212.Maienborn, C., 2009. Building ad hoc properties: on the interpretation of adjectival passives. In: Riester, A., Solstad, T. (Eds.), Proceedings of Sinn

und Bedeutung, vol. 13. University of Stuttgart, pp. 35--49.Marusi!, F., Marvin, T., Zaucer, R., 2004. Depictive secondary predication with no PRO. In: Zybatow at, G., et al. (Eds.), Formal Description of

Slavic Languages. Peter Lang, Frankfurt, pp. 423--434.McIntyre, A., 2013. Adjectival passives and adjectival participles in English. In: Alexiadou, A., Schäfer, F. (Eds.), Non-canonical Passives. John

Benjamins, Amsterdam, pp. 21--42.Nichols, J., 1981. Predicate Nominals. The Partial Surface Syntax of Russian, Linguistics, vol. 97. Berkeley, University of California Press.Parsons, T., 1990. Events in the Semantics of English: A Study in Subatomic Semantics. MIT Press, Cambridge, MA (Current Studies in

Linguistics: 19).Partee, B., 2010. Privative adjectives: subsective plus coercion. In: Bäuerle, R., Reyle, U., Zimmermann, T.E. (Eds.), Presuppositions and

Discourse: Essays offered to Hans Kamp. Emerald Group Publishing, Bingley, UK, pp. 273--285.Paslawska, A., von Stechow, A., 2003. Perfect readings in Russian. In: Alexiadou, A., Rathert, M., von Stechow, A. (Eds.), Perfect Explorations,

Walter de Gruyter. Berlin, pp. 307--362.Pereltsvaig, A., 2001. On the Nature of Intra-Clausal Relations: A Study of Copular Sentences in Russian and Italian (Ph.D. diss.). McGill

University, MontréalPustejovsky, J., 1995. The Generative Lexicon. MIT Press, Cambridge, MA.Richardson, K., 2001. What secondary predicates in Russian tell us about the link between tense, aspect and case. ZAS Pap. Linguist. 26,

152--177.Sakhmatov, A.A., 1941. O!erk sovremennogo russkogo literaturnogo jazyka. U!pedgiz, Moskva.Schoorlemmer, M., 1995. Participial Passive and Aspect in Russian (Ph.D. diss.). Utrecht UniversitySiegel, M., 1976. Capturing the Russian adjective. In: Partee, B. (Ed.), Montague Grammar. Academic Press, New York, pp. 293--309.Sleeman, P., 2011. Verbal and adjectival participles. Lingua 121 (10), 1569--1650.Starke, M., 1995. On the format for small clauses. In: Cardinaletti, A., Guasti, M.T. (Eds.), Syntax and Semantics: Small clauses, vol. 28.

Academic Press, San Diego.Stowell, T., 1983. Subjects across categories. Linguist. Rev. 2, 285--312.Svedova, N.J., et al., 1980. Russkaja grammatica. Nauka, Moskva.Swart, H. de, 1998. Aspect shift and coercion. Nat. Lang. Linguist. Theory 16 (2), 347--385.Timberlake, A., 1986. The semantics of case in Russian predicate complements. Russ. Linguist. 10, 137--165.Williams, E., 1975. Small clauses in English. In: Kimball, J. (Ed.), Syntax and Semantics, vol. 4. Academic Press, New York, pp. 249--273.

O. Borik / Lingua 149 (2014) 139--165 165