Achieving specificity in an articleless language: Specificity markers in Croatian

Abstract

The main goal of the paper was to detect specificity markers in Croatian. As specificity is in a close relation to (in)definiteness, markers of (in)definiteness such as articles are stan- dardly used as indicators of specificity as well. However, detecting specificity in articleless languages becomes less straightforward. Regardless of the lack of the article system, other (in)definiteness markers exist in Croatian. Here, we focused on the following markers of (in)definiteness in order to find specificity markers: i) the adjectives, ii) the numeral jedan 'one', and iii) the demonstrative (definite) and indefinite pronouns. As specificity markers in Croatian we discuss the numeral jedan 'one', the indefinite pronoun neki 'some' and the construction taj + neki 'that some'. Our study also lead to the formation of three categories of specificity based on the properties of these markers: a) contextual specificity, b) semi-lexicalized specificity and c) constructional specificity. [PUBLICATION ABSTRACT]

Headnote

The main goal of the paper was to detect specificity markers in Croatian. As specificity is in a close relation to (in)definiteness, markers of (in)definiteness such as articles are stan- dardly used as indicators of specificity as well. However, detecting specificity in articleless languages becomes less straightforward. Regardless of the lack of the article system, other (in)definiteness markers exist in Croatian. Here, we focused on the following markers of (in)definiteness in order to find specificity markers: i) the adjectives, ii) the numeral jedan 'one', and iii) the demonstrative (definite) and indefinite pronouns. As specificity markers in Croatian we discuss the numeral jedan 'one', the indefinite pronoun neki 'some' and the construction taj + neki 'that some'. Our study also lead to the formation of three categories of specificity based on the properties of these markers: a) contextual specificity, b) semi-lexicalized specificity and c) constructional specificity.

Izrazavanje specificnosti u jeziku bez clanova: Pokazatelji specificnosti u hrvatskom jeziku

Glavni cilj ovoga rada bio je odrediti pokazatelje specificnosti u hrvatskom jeziku. Definicija specificnosti odnosi se na mogucnost govornikove identifikacije jedinstvenog referenta imenske fraze i na taj nacin funkcionira kao primarno pragmaticki utemeljena kategorija, koja svoju funkciju ostvaruje u govornom, odnosno komunikacijskom cinu. Takoder, specificnost se kao sintakticko-semanticka kategorija u literaturi cesto veze uz pojmove odredenosti i neodredenosti te se obicno izrazava clanovima u jezicima poput engleskoga. Na taj se nacin specificnost u literaturi définira kao kategorija koja presijeca kategorije odredenosti i neodredenosti te je njezino izdvajanje za potrebe jezicne analize izazovan zadatak. Takoder, jezici bez sustava clanova, poput hrvatskoga, koriste raznorodne strategije za izrazavanje neodredenosti i odredenosti, kao i njima srodne kategorije specificnosti. U radu se tako raspravlja s kojim se pokazateljima odredenosti i neodredenosti u hrvatskome moze povezati i ostvarivanje specificnosti. Tri kategorije za izrazavanje odredenosti i neodredenosti koje se obraduju jesu: a) kratki i dugi oblici pridjeva, b) broj jedan i c) pokazne i neodredene zamjenice. Na primjerima iz Hrvatskoga nacionalnog korpusa te Hrvatske jezicne riznice pokazuje se nestanak opreke neodredenosti i odredenosti u kratkim i dugim oblicima pridjeva te u skladu s time i nedostatnost ove tradicionalne gramaticke podjele za opis kategorije specificnosti. Posebna se paznja posvecuje funkcijama broja jedan i procesu gramatikalizacije kojime jedan u odredenim kontekstima gubi svoju funkciju broja te preuzima funkciju neodredenoga specificnog clana. Nadalje, kao dodatni pokazatelj specificnosti u hrvatskome istice se slozena konstrukcija taj+neki, odnosno konstrukcija pokazne zamjenice i neodredene zamjenice u poziciji modifikatora imenske fraze. Na temelju korpusne analize primjera i sredstava kojima se izrazava specificnost u hrvatskome ustanovljena je klasifikacija jezicnih sredstava za izrazavanje specificnosti koja se sastoji od kategorija a) kontekstualne specificnosti, b) poluleksikalizirane specificnosti i c) konstrukcijske specificnosti. Kontekstualna specificnost se définira kao pragmaticki ostvarena kategorija, dok se poluleksikalizirana specificnost ostvaruje gramatikalizacijom broja jedan. Konstrukcijska je specificnost izrazena slaganjem zamjenica taj+neki.

Key words: specificity in language, definiteness in language, articleless languages, Croatian language

Kljucne rijeci: specificnost u jeziku, odredenost u jeziku, jezici bez clanova, hrvatski jezik

1. Introduction

Specificity is a semantic property of noun phrases closely related to the notions of indefiniteness and definiteness, as it refers to the ways of marking noun phrases as expressions with a referring function (see Heusinger 2002a). The particular referring function of specificity is to point out the semantic uniqueness of an entity in the mind of the speaker (Haspelmath 1997, Pavey 2008). In this regard, the dominant characteristics of specificity can be subsu- med under the notions of referentiality and identifiability, thus meaning that the entity denoted by the specific NP is a) existentially presupposed and b) the speaker is certain of the identity of the referent (Heusinger 2002a). Dis- eussions related to specificity (e.g. Lyons 1999, Heusinger 2002a, Haspelmath 1997) provide various examples of specificity that illustrate how specificity can be expressed by different lexical or grammatical means, e.g. included within the article system in English, or simply inferred from the context of the utterance itself.

Languages without article systems have been noted to express the notions of indefiniteness/definiteness in various ways, e.g. through the use of prono- uns, affixes and so forth (see Haspelmath 1997, Ionin 2010), and thus require a different approach to investigating specificity with regard to the mechanisms employed to express (in)definiteness. Many of these languages are sometimes referred to as 'determinerless' languages (Boskovic 2005, 2008, 2009), and some of these, such as Croatian, Serbian and Russian, provide the basis for novel syntactic and semantic analyses of determinacy and (in)definiteness.1 However, few of these analyses have focused on specificity marking in these languages2.

What is of interest then is to explore the language specific means that such 'determinerless' languages use to express specificity. Since Croatian is such a language, the goal of this paper is to provide a comprehensive account of the linguistic structures used in expressing specificity in Croatian. As to our knowledge, there have been few studies focusing on specificity marking in Croatian (see Aljovic 2002), so we believe that the present paper will make a contribution to the linguistic description of this phenomenon in terms of a) describing the specificity markers in Croatian, b) expanding the descriptions of specificity cross-linguistically, and c) providing new data on which to base an account of specificity and its syntactic, semantic and pragmatic properties.

Since limited data on specificity marking existed for Croatian, the primary task was to find the linguistic elements Croatian employs for marking speci- ficity. For this reason corpus data were consulted3 and a study of the results was conducted, which yielded the findings of specificity markers as well as their classification according to the morphosyntactic and semantic properties of these markers. The NP constructions most commonly used for expressing spe- cificity in Croatian are a) the constructions with the numeral jedan 'one'4, and b) the Definite + Indefinite Pronoun constructions, e.g. taj neki 'that some'.

1.1 Specificity and (in)definiteness

Specificity is often considered as cross-cutting the notions of definiteness and indefiniteness. As Heusinger (2002a) points out, specificity is spread over the two categories in that both specific and non-specific features can combine with indefinite NPs, and the two categories do not yield contradictory inter- pretations of the NP in question (as shown in example (1)).

(1) I bought a car yesterday.

The NP ? car gives a specific reading of the NP because the speaker knows the identity of the car and also states that the car is a unique referent, i.e. it is a single specific automobile that is now the property of the speaker. Contextual inference is the predominant factor in the specific reading of the NP, since it is expected that the buyer of the car knows exactly which car he or she is talking about. In this respect, specificity is often discussed as being a subcategory of (in)definiteness, which functions as a fine-grained distinction usually conditioned by the context of the utterance. For this reason many studies conducted on languages with article systems, such as English, focus mostly on specificity as coded within the article system itself (e.g. Fodor and Sag 1982).

It is also often considered that all definite NPs are at the same time spe- cific by default, in that the definiteness of the entity denoted by the NP is a necessary consequence of the identifiability and referentiality of the referent. So for instance, in the example (2), the blue book has a specific referent that both the speaker and the hearer recognize, and as such is both definite and necessarily specific.

(2) Give me the blue book on the table over there.

Heusinger (2002a)5 presents this intersection of the two categories in the following table 1:

It is worth noting that the most interesting and the most studied cate- gory in the table is the one in the middle, the specific indefinite NPs, mostly because this is the category in which the two notions, that of specificity and indefiniteness are most clearly separated by the speaker/hearer knowledge of the identity of the referent. In our description of the Croatian data, we will also focus on the indefinite - specific markers as the starting point for our analysis.

However, it has been pointed out by Heusinger (2002a) that specificity can be distinguished from indefiniteness based on its focus on "speaker knowled- ge", whereas indefiniteness in examples such as (1) has its focus on the dis- course as a whole. In other words, the function of the article a is to mark the novelty of the NP car within the discourse in terms of indefiniteness. The indefinite article thus has a pragmatic-discourse function different from specificity, in terms of establishing links between parts of the discourse and marking novel topics of the discourse (see Heusinger 2002a).

Additionally, it has been noted by Ionin (2006) and Heusinger (2002a) that both definite and indefinite NPs can have specific as well as non-specific rea- dings. In her study on the English demonstrative this, Ionin (2006) shows how the demonstrative pronoun has a specific indefinite reading when it is used in its unstressed, phonologically reduced form as in example (3):

(3) Mary read this exciting book the other day.

Such distinctions are important to note because they imply that the inve- stigation of specificity in different languages has to take into account possible different functions of specificity and definiteness that could prompt languages to use different linguistic forms for the two categories. More importantly, specificity may be in some cases a property relevant enough to be expressed formally, rather than only through context. Many languages can draw from existing definiteness markers to express specificity as well (such is the case of English, for example).

In his analysis on the ways of expressing specificity in various languages, Heusinger (2002b) calls these linguistic forms language-specific grammatical means of expressing specificity and includes primarily grammatical constructi- ons that exploit the existing (in)definiteness markers by combining them into specificity markers.

The clearest example of specificity as a standalone category comes from languages such as Maori or Samoan that have particles that function as speci- ficity markers instead of (in)definiteness markers. In a study of Samoan con- ducted by Mosel and Hovdhaugen (1992) (in: Lyons 1999) the distributional patterns of two Samoan particles le and se lead the authors to conclude that these are specificity particles marking 'specific' and 'non-specific' NPs, res- pectively. The example they provide in (4) illustrates how these distributional properties guide the usage of the 'specific particle' le regardless of the disco- urse functions performed by the indefinite and definite articles in its English translation.

(4) Sa i ai le ulugali'i 'o Papa le tane a 'o

PST exist ART couple PRS Papa ART husband but PRS

Eleele le fafine.

Eleele ART woman

'There was a couple, Papa, the husband, and Eleele, the wife.'

'There was a couple, Papa, the husband, and Eleele, the wife.'

As Lyons (1999) points out, the distributional properties of the Samoan particles depend only on the specificity or non-specificity of the NPs, and not on the novelty of NPs as topics in the discourse. In other words, the particles have a referring function, and not a discourse pragmatic one. In the example it is clear how the knowledge of the narrator about a particular couple determi- nes the usage of the specific particle, and not the fact that the NP 'a couple' is introduced for the first time into the discourse. The English translation shows how different principles guide the usage of its articles as opposed to Samoan. Similarly, the Siouan language Lakhota distinguishes the indefinite specific article and the indefinite non-specific article (Van Valin 1995: 41-42).

Although languages such as Samoan and Lakhota provide clear evidence of the existence of specificity as a linguistic category in its own right, many languages that fall between the examples of Samoan and English still remain to be investigated. Such is the case of Croatian, as we will argue below.

2. Indefiniteness in Croatian

To examine the relation between (in)definiteness and specificity in Croa- tian, we must first examine the ways of expressing definiteness in Croatian. Croatian does not have an article system, so definiteness is often inferred from the context of the utterance, e.g.

(5) Ptica je odletjela sa stabla.

bird-NOM.SG AUX-3.SG fly-PST from tree-GEN.SG

'A / The bird flew offthe tree.'

The most common means of explicitly expressing definiteness according to Croatian grammars (Silic and Pranjkovic 2007; Baric et al. 2005) are a) via indefinite and definite adjective inflection, b) definite and indefinite pronouns, and c) the numeral jedan 'one' in the function of an indefinite article.

2.1 The definite and indefinite adjectival inflection

Adjectives in the attributive position in Croatian, as in the majority of Slavic languages, can appear in two different forms - the definite and the in- definite form (e.g. Aljovic 2002, Silic i Pranjkovic 2007: 134), also referred to as the long and the short form of the adjective, or the adjectival aspect (Katicic 1991: 385, Znika 2006: 16). For example, the nominative masculine singular of the adjective meaning 'big' is velik if indefinite, and veliki if definite. It is usu- ally claimed, as in Silic and Pranjkovic (2007), that the indefinite form of the adjectives is used to modify a noun phrase that is discourse-new and unknown to the speech act participants, while definite adjectives function as modifiers in NPs that are known and have been previously mentioned in the discourse.

The distinction between the definite and the indefinite adjectival inflec- tion is not so clear-cut in language use anymore. Silic and Pranjkovic (2007: 381) state that the lack of this distinction is characteristic of the language of administration, and Brozovic (1988, as cited in Znika 2006) believes that the opposition has become optional in language and is stylistically marked. Any study of the meaning and use of long as opposed to short adjectival forms in the attributive position should thus take into account the neutralization of this distinction in contemporary language.

Examples that we found in the Croatian Language Repository confirm the claim that both long and short adjectival forms can be used equally in definite and indefinite noun phrases, as shown in (6) and (7):

(6) Taj velik skup, na kojem ce sudjelovati

that[NOM.SG] big[INDEF.NOM.SG] assembly[NOM.SG] on which-LOC.SG AUX-3SG participate-INF

stotinjak likovnih kriticara iz citavoga svijeta,

hundred art-GEN.PL critic-GEN.PL from entire-GEN.SG world-GEN.SG

a pratit ce ga izlozbe u Zagrebu i Dubrovniku, najavili su u petak

u Muzeju »Mimara« predsjednik hrvatske sekcije AICA-e Berislav

Valusek i tajnica Leonida Kovac.

'This big assembly, which around a hundred art critics from all around the world will take part in, and which will be accompanied by exhibitions in Zagreb and Dubrovnik, was announced on Friday in the "Mimara" Museum by the president of the Croatian section of AICA, Berislav Valusek, and the secretary, Leonida Kovac.'

(7) Svi se ra|amo, svi zivimo i umiremo, ali neka umiremo tako, kako je rekao

as AUX-3SG say-PST

jedan veliki pjesnik, da svijet iza sebe ostavimo

one[NOM.SG] big-DEF.NOM.SG poet[NOM.SG]

malo boljim, nego sto smo ga nasli.

'We are all born, we all live and die, but let us die in such a way, as a great poet said, that we leave the world behind a bit better than we found it.'

In (6), an indefinite adjective occurs in an NP whose definiteness is si- gnaled by the demonstrative pronoun taj, while in (7) the numeral / article jedan 'a; one' indicates the indefiniteness of the NP (see Section 2.3. below) in which, nevertheless, a "definite" adjective is used as modification. This clearly shows that the distinction between long and short adjectival forms, if it ever did carry the definite/indefinite meaning, should be reinterpreted as to how it functions in the contemporary language, possibly as a specificity marker.

Aljovic (30-31) uses very similar examples to show that the two adjectival inflections are not used to mark the definiteness/indefiniteness distinction, but rather to trigger a specificity reading. In her view, the short form of the adjective is always used in non-specific NPs, while the long form signals that the noun phrase should be interpreted as specific, and not necessarily definite, as, for example, in (8):

8) jedan vrijedni student

'one diligent student'

While we agree with Aljovic that long and short adjectival forms are not used to signal the definiteness or indefiniteness of the NP, in this case we are more inclined to attribute the specificity reading of (8) to the numeral / article jedan, 'a; one' which we believe to be an instance of semi-lexicalized specificity (see section 3.).

2.2 Indefinite and definite pronouns

As to the pronoun system, Croatian grammars often classify indefinite and definite pronouns as specific subtypes of the pronoun word class. Definite pronouns are termed demonstratives, and their primary function is defined as indexical, pointing to the proximity of a certain entity to the speaker, hearer or a third party (Baric et al. 2005).

(9) Evo ti ova knjiga, ti meni daj tu,

here 2SG-DAT DEM.PROX-NOM.SG book-NOM.SG 2SG.NOM 1SG-DAT give-IMP.2SG DEM.MED-ACC.SG

a ona na stolu mi ne treba.

and DEM.DIST-NOM.SG on table-LOC.SG 1SG-DAT NEG need-PRS.3SG

'Take this book, you give me that one, and that one on the table I do not need.'

It is also important to note that this three-way distinction is slowly being replaced by the dyadic opposition ovaj - taj 'this-that', especially in spoken language and substandard registers6.

The function of demonstratives to express definiteness is not uncommon among languages without an article system, as a way of compensating for the communicative needs of expressing definiteness (see Lyons 1999, Haspelmath 1997). In this way Croatian demonstratives follow the pattern observed in other languages, e.g. Russian, Czech, etc. (Haspelmath 1997)7. Furthermore, it can be shown that examples such as:

(10)...

with the demonstrative tu 'that', can be translated felicitously both with the definite article the or the demonstrative this/that.

Indefinite pronouns form a large subclass of pronouns with both simplex and complex forms, e.g. tko 'who, someone' > ne-tko 'someone, somebody', ni-tko 'no one, nobody', i-tko 'anyone, anybody', sva-tko 'everyone, everybody', etc. Their main function is to express indefîniteness (Silic and Pranjkovic 2000; Baric et al. 2005). Among indefinite pronouns one has to distinguish between those that take the modifier position in the NP8, e.g. neki 'some', nekakav 'some kind' and neciji 'someone's', and those that take the head position, e.g. netko 'someone, somebody', as in examples (11) and (12):

(11)...

(12)...

In the example (11) the pronoun neki 'some' establishes the indefîniteness of the NP covjek 'man' and thus acts as an indefîniteness marker.

As to the correlation between the demonstrative (definite) and indefinite modifier pronouns, it can be stated that they perform the discourse-pragma- tic functions in a very similar way to articles in English, e.g. the correlation between the pronouns neki 'some' and taj 'that' in example (13):

(13)...

2.3 The article jedan

Although Croatian lacks articles as a separate word class, the numeral je- dan 'one' in some cases acts like an indefinite article (see Silic and Pranjkovic 2007), e.g.

(14)...

(15)...

According to Silic and Pranjkovic (2007), the use of jedan 'one' as an article is restricted only to cases where the adjective of the NP is not inflecti- onally marked as either a definite or an indefinite adjective, e.g. majcino srce 'mother's heart'. However, due to the fact that definite and indefinite adjecti- val flection is becoming obsolete, the use of jedan 'one' as an article9 seems to be more facultative and influenced by specificity marking as well (see discussi- on below). As the following corpus examples10 show, jedan 'one' can be used as an article either with a definite or an indefinite adjective:

(16)...

(17)...

3. Specificity markers in Croatian

Though there is much more that can be said about definiteness marking in Croatian, for the purposes of our research we decided to focus on the afo- rementioned strategies of marking (in)definiteness to determine the factors that guide specificity readings in Croatian. Based on the fact that markers of specificity are usually closely tied to those of (in)definiteness, our corpus search focused on these markers and their syntagmatic surroundings within a discourse. To determine the specificity markers we used the criterion of sub- stitution, according to which specificity markers are substituted in a sentence with 'a certain' (see Heusinger 2002a, Lyons 1999). This simple test allowed us to establish two prominent specificity markers, the article jedan 'one' and the construction taj+neki 'that+some', i.e. Definite + Indefinite Pronoun con- struction.

(18)...

(19)...

Based on the corpus overview we also established a classification of the ways specificity is expressed in Croatian. The classification was based on the following factors influencing specificity readings:

a) because of contextual inference, indefinite articles and pronouns can have both specific and non-specific readings

b) some linguistic expressions yield only specific readings, and for that reason must be analyzed as having a systematic specificity function, not simply being subjected to contextual modification

c) a specific NP has to be a) identifiable and b) carry felicity conditions on its use (see Ionin 2010)

The contrast between contextual modification and systematic meaning is the key opposition in determining specificity markers, i.e. the language specific means Croatian uses to express specificity11. The felicity conditions of specific NPs are discussed at length by Ionin (2010) and refer to the fact that certa- in pragmatic conditions have to be met for the utterance to sound perfectly acceptable. For instance, in the following example (20) the use of this would be odd if the speaker did not know anything about the movie (genre, title or the like), as compared to (21):

(20)...

(21)...

The classification we propose based on these factors deals with three ca- tegories of specificity:

1) contextual specificity - zero marked, indefinite pronoun neki

2) semi-lexicalized specificity - article jedan

3) constructional specificity - taj+neki construction

3.1 Contextual specificity

The first category, contextual specificity, is the most straightforward pra- gmatically motivated way of establishing specificity. In Croatian it is encoun- tered in examples such as:

(22)...

The example lacks any prenominal (in)definiteness markers and can have both a specific and a non-specific reading depending on the context which can be expanded as follows:

(23)...

(24)...

When the indefinite pronoun neki 'some' is added to the existing senten- ces they are still both grammatically acceptable, however the example (25) seems pragmatically odd.

(25)...

(26)...

The differences in co-reference (rieka - koja / rieka - nijedna) between the two sentences are often referred to as scope properties of specific NPs (see Pa- vey 2008, Fodor and Sag 1982, Heusinger 2002a, Lyons 1999). As Ionin (2010) points out, specific NPs have wide scope properties, meaning that the specific NPs remain identifiable as a constant through the utterance and can co-refe- rence as such with other elements within the utterance. In an example taken from Lyons (1999) such is the scope relation between the NP 'a merchant banker' and the co-referenced pronoun 'her':

(27)...

Non-specific NPs, on the other hand, have different scope properties since the non-specific NP is a variable that cannot be co-referenced in the same way:

(28)...

Thus the scope properties are also a good indicator of specificity, and can point to specificity markers alongside the so-called substitution test with 'a certain'.

Although we list the indefinite pronoun as an example of contextual specificity, it seems that without postmodification it tends to yield mostly non-specific reading (see e.g. (33) and (34)). However, the frequency analysis that we have performed to establish its tendency to lexicalize non-specific reading showed that the distinction between the specific and non-specific neki is approximately only 52% for the non-specific reading. We believe that the reason why the corpus data is inconclusive may be attributed to the properties of corpus texts in general, which consist mostly of complex sentences including postmodifications of the pronominal construction such as a relative clause in example (29).

(29)...

3.2 Semi-lexicalized specificity

If we add the article jedon 'a; one' to our examples, we will get a wide scope specific NP:

(30)...

On the other hand, although the examples such as (31) are grammatically acceptable, they seem pragmatically odd:

(31)...

Furthermore, if we contrast examples with jedan 'a; one' and non-marked or pronoun marked indefinite NPs, there is a clear contrast in the specific versus non-specific interpretation of these NPs, as in the examples:

(32)...

(33)...

Note the oddness of the example:

(34)...

However, the following is fully acceptable:

(35)...

Thus it seems that jedan 'a; one' is used predominantly as a specific indefinite marker, and instantiates the category of lexicalized specificity in Croatian. This is not to say that jedan 'a; one' has to be a specific marker in every instance of it's usage, but only that it shows a strong tendency towards specific reading. Based on the opposition between contextual modification and systematic meaning, it serves to show that the lexicalization of specificity in the example of jedan 'a; one' is a gradable phenomenon. However, it is a phenomenon worth noting since it expands the description of the use of jedan 'a; one' beyond the current descriptions. On the other hand, the same can be said for the indefinite pronoun neki 'some', which shows a tendency towards indefinite non-specific readings when the NP is not postmodified, as in example (32). To further support this claim, we performed a frequency analysis13 of the specific versus non-specific occurrences of jedan 'a; one' in the Croatian National Corpus (henceforth CNC).

The analysis showed the following results:

Column 1 represents the sum of all the tokens of jedan in the CNC, columns 2 and 3 represent the token number of numeral jedan 'one' and the article jedan 'a; one', respectively. Columns 4 and 5 show the frequencies of specific vs non-specific occurrences of the article jedan 'a; one'.

3.3 Constructional specificity

The category of constructional specificity is perhaps the most clear-cut case of language specific means Croatian uses to express specificity. In all the corpus examples of the construction taj+neki 'that+some' the readings were specific14, i.e. had a wide scope and were substitutable with 'a certain'15. It is interesting to note that the construction itself is composed of two notionally opposing parts, the demonstrative pronoun taj 'that' and the indefinite pronoun neki 'some'. Although the two units of the construction have opposing semantic features, those of definiteness and indefiniteness, it seems that the end result is a construction with the features 'Indefinite and Specific'. From the examples it is clear that the criterion of speaker knowledge is very prominent, e.g.

(36)...

(37)...

In both examples taken from the CNC it is implied that the speaker knows the identity of 'new names on the music scene' and 'the structured spa- ce', respectively. Thus it seems that the construction Def. + Indef. pronoun serves a particular function within the discourse which adds the notion of speaker knowledge as a relevant notion of the utterance. In her study on spe- cificity marking in English and Russian, Ionin (2006) called indefinite specific markers, such as English this and Russian odin 'one' markers of epistemic specificity. The function of such markers, according to Ionin (2006), is to stress the importance of the identity of the indefinite NP, as opposed to cases with a articles, where the identity is not of importance. To illustrate this point, Ionin used the following examples for English, where the identity of the referent is necessary for a felicitous interpretation of the sentence (38) and (39).

(38)...

(39)...

The Croatian examples follow the felicity conditions established by Ionin (2006) for English:

(40)...

(41)...

Also, the examples below show the non-specific readings of neki 'some'.

(42)...

(43)...

Based on the tendency of neki 'some' to have a non-specific reading in many non-postmodified NPs, it seems that the demonstrative taj serves to modify this reading into a specific one, thus creating a construction with novel specificity properties. The whole construction being indefinite, we can assume that the original indexical function of taj 'that' is of greater salience in crea- ting the construction than its function as a definiteness marker. Thus it loses its definiteness meaning in the taj neki 'that some' construction and triggers a specific interpretation of the referent of the NP, creating a specific NP.

4. Conclusion

Specificity is a category that often cross-cuts the category of (in)definiteness. In languages such as Croatian, where no article system exists, these categories have various ways of being expressed and serving their communicative func- tions. Distinguishing specificity as opposed to (in)definiteness is not an easy task, since the categories often use the same or somewhat related markers. However, since specificity is connected mainly with the speaker knowledge of the identity of the referent, it obviously serves a function that requires its expression in certain linguistic forms, such as the numeral jedan 'a; one' and taj neki 'that some' in Croatian. Specificity is often a matter of contextual inference, and as such carries various pragmatic information about an utteran- ce. Because of its pragmatic nature, it is often difficult to establish clear-cut boundaries between specific and non-specific markers, as we have shown with the example of jedan 'a; one'. Although there are, in Heusinger's terms, langu- age-specific means of expressing specificity, we believe that the phenomenon of specificity as a linguistic category is gradable, and for that reason we propose a threefold classification of specificity in Croatian, a) contextual specificity, b) semi-lexicalized specificity, and c) constructional specificity. In our classifica- tion we focused on single lexical units or constructions, which can serve as specificity markers, but also showed that the only category with a continuous specificity reading in all cases is the construction taj neki 'that some'.

LIST OF ABBREVIATIONS

1 - first person

2 - second person

3 - third person

ACC - accusative

ART - article

AUX - auxiliary

COND - conditional

COP - copula

DAT - dative

DEF - definite

DEM - demonstrative

DIST - distal

GEN - genitive

IMP - imperative

INDEF - indefinite

INF - infinitive

INS - instrumental

LOC - locative

MED - medial

NEG - negation

NOM - nominative

PL - plural

PROX - proximal

PRS - present

PST - past

PTCP - participle

REFL - reflexive

SG - singular

Footnote

1 As we here do not discuss the functional-lexical status of the DP from the perspective of formal/generative approaches, we refer the reader to the following literature. For a general discussion of the topic see Alexiadou, Haegeman and Stavrou 2007; for the discussion of splitting the DP where projections FinP and TopP host [+definite] and [+specific] features respectively see Ihsane & Puskás 2001, Aboh et al. 2010. For the discussion of the topic for Croatian data see two competing proposals: for the proposal of the complete omission of the DP layer in favor of a simple NP analysis see Zlatic 1998, 2009, Trenkic 2004, Boskovic 2005, 2008, 2009; for the proposed of elaborated DP structure analysis see Progovac 1998, Leko 1999, Dimitrova-Vulchanova and Giusti 1998, Basic 2004, Caruso 2011a, 2011b; for the evidence of the D head, hence DP, in Croatian see Caruso 2011a, b. For the discussion of the topic for other Slavic languages see Rutkowski 2002, Pereltsvaig & Franks 2004, Pereltsvaig 2007. We also note that the analysis proposed for the Croatian data in this article is in accordance with the elaborated DP structure analysis, and in particular with the analysis and arguments presented in Caruso 2011a, b.

2 Some exceptions are Heusinger's (2002a) study on specificity markers in Turkish, as well as Ionin's (2006) analysis of specificity with respect to the Russian article odin 'one'.

3 Corpora used in our research are the Croatian National Corpus, http://www.hnk.ffzg.hr/, and the Croatian Language Repository, http://riznica.ihjj.hr/. Our analysis of the corpus data is primarily qualitative, focused on describing the main properties of specificity markers. Frequency analysis was applied in some cases in order to confirm our hypotheses about the properties of specificity markers.

4 Which is, as will be argued in the paper, also functionally an indefinite article.

5 In his paper Heusinger (2002a) goes on to argue that a fourth category, that of definite non-specific markers, can be established, e.g. Someday I will find the man ofmy dreams.

6 See also Zic Fuchs (1996) for a discussion on spatial deixis in Croatian.

7 This is a very common definiteness marking strategy is languages without definite articles, demonstratives are used in the same way in e.g. Mandarin Chinese (Hedberg 1996).

8 In this paper we will only focus on indefinite pronouns as modifiers, since they are the ones that can carry an "article-like" function in Croatian, contrary to indefinite pronouns in the head position, e.g. netko 'somebody'.

9 Use of the numeral 'one' as an indefinite article is not uncommon among (especially Slavic) languages, see Ionin (2006) for the description of the Russian odin 'one'. Diachronically, many languages have developed articles by grammaticalizing numerals, e.g. French and English (see Lyons 1999). Turkish seems to be very similar to Croatian in this respect, since it uses bir 'one' to mark specific indefinite NPs in the accusative case (Schaaik 1996: 40-41).

10 http ://riznica.ihjj.hr/.

11 Within the framework of usage-based models of language (Kemmer and Barlow 2000, Evans and Green 2004), the term contextual modification refers to meanings of lexical units that are established through the surrounding context of the utterance and are not part of the encyclopedic knowledge pertaining to the language system, more specifically the conceptual structure of the lexical unit itself. Evans and Green (2004) give examples such as This toy is safe for children, vs The girl is safe with, her grandmother, where the meaning of safe is in both cases ruled by the surrounding context. In other words, such modifications are pragmatically guided occurrences, which depend on knowledge structures wider than the lexical meaning itself. Such is the example in (1) ? bought a car yesterday', where the specificity reading is a matter of contextual modification. Some frequent instances of contextual modifications can become more and more entrenched in time, and this process can be observed as a tendency of some language units to appear more frequently with the same meaning; such is the process between the numeral jedan 'one' and the grammatical- ized article jedan 'a, one'.

12 http://riznica.ihjj.hr

13 The Croatian National Corpus has 100 million tokens, and it is compiled from various literary and newspaper texts. The analysis was conducted on a random sample of 200 concordances, by determing the numeral, article, and specific and non-specific article readings of jedan, and then multiplying these percentages with the total number of tokens of jedan, thus obtaining an estimation of the distribution of various meanings of jedan.

14 All the inflectional variants for gender and number were examined, e.g. ta neka, ti neki, te neke, taj neki.

15 Also note that the order of the elements is reverse than that of the English 'a certain' construction, which has the indefinite marker first.

References

References

Aboh, E. O., M. Dyakonova, M. van Koppen (2010) "DP-internal information structure: Some introductory remarks". Lingua 120: 782-801. Special issue on DP-internal information structure

Alexiadou A., L. Haegeman, M. Stavrou (2007) Noun phrase in the generative perspective. Berlin: Mouton de Gruyter.

Aljovic, N. (2002) "Long Adjectival Inflection and Specificity in Serbo-Croatian". Recherches Lin- guistiques de Vincennes 31: 27^i2.

Baric et al. (2005) Hrvatska gramatika [Croatian Grammar]; Skolska knjiga, Zagreb.

Basic, M. (2004) Nominal Subextractions and the Structure of NPs in Serbian and English. MA Thesis: University of Troms0.

Boskovic, Z. (2005) "On the locality of left branch extraction and the structure of NP". Studia Lingüistica 59: l^i5.

Boskovic, Z. (2008) "What will you have, DP or NP?", Proceedings of NELS 37.

Boskovic, Z. (2009) "More on the no-DP analysis of articleless languages". Studia Lingüistica 63 (2): 187-203.

Caruso, D. Z. (2011a) "In support of a DP-Analysis of Nominal Phrases in Croatian". Workshop on Languages With and Without Articles, Université de Paris 8 & UMR 7023, March 3-4, 2011.

Caruso, D. Z. (2011b) "Nominal Phrases in Croatian as DPs". Online Proceedings of GLOW in Asia Workshop for Young Scholars 2011.

Dimitrova-Vulchanova, M., G. Giusti (1998) "Fragments of Balkan Nominal Structure". In: A. Alexiadou and C. Wilder (eds). Possessors, Predicates and Movement in the Determiner Phrase. 333-60. Amsterdam/Philadelphia: John Benjamins.

Evans, V., M. Green (2004) Cognitive Linguistics: An Introduction. Cambridge University Press: Fodor, J. D., I. Sag (1982) "Referential and quantificational indefinites". Linguistics and Philo- sophy 5: 355-398.

Haspelmath, M., (1997) Indefinite Pronouns. Oxford Studies in Typology and Linguistic Theory. Oxford: Clarendon Press.

Hedberg, N. (1996) "Word Order and Cognitive Status in Mandarin Discourse". In: Fretheim, T., J. Gundel (eds). Reference and Referent Accessibility. Pragmatics and Beyond Series. John Benjamins: 173-192.

von Heusinger, K (2002a) "Specificity and Definiteness in Sentence and Discourse Structure". Journal of Semantics 19: 245-274.

von Heusinger, K (2002b) "The Cross-Linguistic Implementations of Specificity". In: Jaszczolt Κ, Κ Turner (eds.). Meaning through Language Contrast. Vol. 2. Amsterdam; Philadelphia: Benjamins: 405^421. (Pragmatics and Beyond NS 100)

Ihsane, T., G. Puskás (2001) "Specific is not Definite". Generative Grammar in Geneva 2: 39-54.

Ionin, T. (2006) "This is definitely specific: specificity and definiteness in article systems". Natural Language Semantics 14: 175-234.

Ionin, T. (2010) "The scope of indefinites: an experimental investigation". Natural Language Semantics 18.3: 295-350.

Katiäc, R. (1991) Sintaksa hrvatskoga knjizevnog jezika: nacrt za gramatiku. Zagreb: HAZU and Globus.

Kemmer, S., M. Barlow (eds.) (2000) Usage-Based Models of Language. Cambridge University Press.

Leko, N. (1999) "Functional Categories and the Structure of the DP in Bosnian". In Dimitro- va-Vulchanova, M., L. Hellan (eds): Topics in South Slavic Syntax and Semantics. John Benjamins Publishing Company, Amsterdam: 229-252.

Lyons, Ch., (1999) Definiteness. Cambridge University Press.

Pavey, E. (2008) "Predication and reference: the interpretation of noun phrases in specificational constructions". In: Van Valin, R. Jr. (ed). Investigations of the Syntax-Semantics-Pragmatics Interface. John Benjamins. 305-317.

Progovac, L. 1998. "Determiner Phrase in a language without determiners". Journal of Linguistics 34(1): 165-179.

Pereltsvaig, A. (2007) "On the Universality of DP: A View from Russian". Studia Lingüistica. 61(1): 59-94.

Pereltsvaig, A., S. Franks (2004) "Functional Categories in the Nominal Domain". In Proceedings ofFASL 12. The Ottawa Meeting 2003, ed. by Arnaudova, Olga, Wayles Browne, Maria Luisa Rivero, and Danijela Stojanovic. University of Michigan Press. Ann Arbor, MI: 109-128.

Rutkowski, P. (2002) "Noun/Pronoun Asymmetries: Evidence in Support of the DP Hypothesis in Polish". Jezikoslovlje 3.1-2: 159-170.

van Schaaik, G. (1996) Studies in Turkish Grammar. Wiesbaden: Harrassowitz.

Silic, J., I. Pranjkovic (2007) Gramatika hrvatskoga jezika za gimnazije i visoka ucilista. (Second edition) Zagreb: Skolska knjiga.

Trenkic, D. (2004) "Definiteness in Serbian/Croatian/Bosnian and some implications for the ge- neral structure of the nominal phrase". Lingua 114: 1401-1427.

Van Valin Jr., R. D. (1995) "Toward a functionalist account of so-called 'extraction constraints'". In: B. Devriendt (ed.) Complex structures: A functionalist perspective. Berlin: Mouton de Gruyter: 29-60.

Zlatic, L. (1998) "Slavic Noun Phrases are NPs not DPs". To appear in Harves, S., J. Lavine (eds) (2013): Comparative Slavic Morphosyntax. Slavica Publishers, Bloomington.

Zlatic, L. (2009) "Definiteness and Structure of NPs in Slavic". The Eighth European Conference on Formal Description of Slavic Languages, December 2-5, 2009, Potsdam, Germany.

Znika, M. (2006) "Kategorija odredenosti i predikatno ime". Jezik: casopis za kulturu hrvatskoga knjizevnog jezika 53, 1; 16-25.

Zic Fuchs, M. (1996) "Here" and "There" in Croatian: a case study of an urban standard vari- ety"; in: Dirven, R. (ed.) (1996) Construal of Space in Language and Thought. Walter de Gruyter.