Serbo-Croatian (SC) is a language with a rich case morphology and a complex and specific lexical prosody.1 Its case system consists of three main declension classes with seven cases, each interacting with the value of number. It displays pitch-accent lexical prosody: a system combining tone and stress (Zec 1999). SC morphology and prosody interact in complex ways, as discussed in Arsenijević 2010 and Simonović and Arsenijević 2014.
Vocatives of certain SC nouns occur with two different prosodic patterns: one faithful to the base, in the sense that it corresponds to a prosodic pattern displayed by at least one other case form, and another that receives the default prosodic pattern (a falling accent on the initial syllable). The former is restricted to indefinite contexts, and the latter to definite ones. I propose an analysis in terms of lexicalization, whereby definite vocatives of common nouns lexicalize into the equivalent of proper names. As a result, they are associated with the respective declension classes of proper names and take their prosodic pattern as well. The indefinite form, which is very rarely used, is derived productively and thus displays the prosodic pattern of the base.
2 Vocatives and Definiteness in Serbo-Croatian
The vocative is one of the seven forms making up the SC case system: the case used to address the hearer(s). It generally has its own distinctive ending: -u/-e in the masculine gender and -o/-e in the feminine, depending on the declension (sub)class, except in a few small classes where it is syncretic with the nominative, as in (1c).
A number of SC nouns, including but not limited to those with a vocative-nominative syncretism, show two different vocative forms distinguished by their ending and often also by their prosodic shape.
All such pairs exhibit two general asymmetries. One member of the pair is rather regular in behavior: it shares the prosodic pattern of the base and has a vocative ending rather than being syncretic with the nominative. The other member exhibits irregularity in one or both of these dimensions. The regular member is always limited to indefinite uses of the vocative, and the irregular member to definite ones.
Nouns with the vocative-nominative syncretism are typically -a declension nouns, either proper names (Tea, Mia, Ivana, Jasmina) or kinship terms (tata ‘dad’, mama ‘mom’, deda ‘grandpa’, baba ‘grandma’). While for those with three or more syllables the conditions for the syncretism are rather complex, in disyllabic nouns of the -a declension the syncretism can only occur if the first syllable bears a falling tone. Most of the nouns in this class are highly frequent, typically definitely used nouns, often denoting kinship and other relations. I argue that due to frequent definite use, certain definite vocatives lexicalize into suppletive forms; that is, they develop a suppletive form reserved for definite use (instead of their compositional use that is unrestricted regarding definiteness). The falling initial syllable, representing the default prosodic shape, reflects a lack of lexical prosodic specification, a property that plays a role in specifying the declension subclass a noun belongs to. The role of the falling initial syllable is thus to detach the lexicalized vocative from the declension subclass of its base.
The vocative is used to address the collocutor, and the collocutor is likely to be definite—that is, both unique and familiar. However, there are uses that involve arguably indefinite nominal expressions, such as the predicative part in the so-called expressive use (Arsenijević 2007). Feminines that do not exhibit syncretism with the nominative (those that are not kinship terms, like s2 However, those that do exhibit syncretism, like t3), where t
Examples like (3) have been analyzed in Arsenijević 2007 as vocative small clauses, whose subject (here, Jovane) is definite, but whose predicate (e.g., izjelice jedna ‘you eater’) is indefinite, as marked by jedna ‘one’, which here expresses a combination of indefiniteness and intensification (see Arsenijević 2018).3
Another class of SC nouns that is curious in this regard shifts stress from the penultimate syllable of the base in other case forms to the initial syllable in vocatives (for lexical prosody shifts in SC, see Zec 1999, Arsenijević 2010, Simonović and Arsenijević 2014). They are illustrated in (4), which also shows that default prosody vocatives are allowed only in definite contexts, while base stress vocatives are restricted to the indefinite use.
The noun gospoda ‘gentlemen’ in (4b) is a partial exception to these restrictions, as its base stress vocative can be used in both definite and indefinite contexts, while its default stress vocative, like that of the other nouns in (4), is restricted to the definite use.5
As specified above, in the indefinite use, these nouns display the lexically specified prosody of the stem, which is judged ungrammatical in the normal, definite use. This shared sensitivity to definiteness puts these nouns in a macroclass with the tetka-type nouns discussed above.
One might suggest that it is predicativity rather than definiteness that conditions the behavior of these nouns in the vocative case. However, other contexts where their vocative form is conditioned, such as example (6), where the noun’s plural reference allows indefinite addressing without predication, confirm that indefiniteness is the relevant factor. The example assumes a context in which the speaker does not know if there are any gentlemen among the addressees—for instance, the speaker enters a dark room and utters it (the first sentence establishes the speaker’s agnosticism about the existence of an adequate referent, thus ensuring an indefinite, nonpredicative addressing use).
This indefinite, nonpredicative use also allows only the form prosodically faithful to the base.
To sum up: In nouns with a vocative-nominative syncretism in SC, the syncretic form is restricted to the definite use, while in the indefinite use a marked vocative with its own ending is derived. Feminine nouns of this class are characterized by the default initial falling accent.
3 Lexicalized DPs
I propose to treat the irregular vocatives (those that are syncretic with the nominative and those that mismatch the prosody of the base) as lexicalized DPs.6 Due to their almost exclusive use in definite contexts, the DPs of these vocatives undergo lexicalization: that is, their internal structure is collapsed into a simplex lexical item, and they are memorized on a par with the paradigm of the base. For reasons of space, instead of developing a formal syntactic model I will refer to some applicable models in the literature and focus on the more informative lexical and prosodic issues, discussed in section 4.
Regardless of whether one adopts a lexicalist or nonlexicalist view of syntax, the choice between the vocative forms discussed in section 2 must be determined in syntax. The reason is that the choice between two phonological spell-out options is conditioned by a semantic notion: definiteness. Any other solution would involve bypassing syntax in the interaction between LF and PF regarding the syntactically active notion of definiteness, which goes against Minimalist assumptions regarding the architecture of grammar. The simplest way to capture this dependency is to include [def(inite)] in the inventory of syntactic features, as a syntactic representation of the definite reference of the relevant vocatives. This feature needs to be spelled out both at LF, where it is interpreted as a property of reference, and at PF, where it is interpreted as a licensor of a particular ending and/or prosody.
In this section, I sketch an analysis for the phenomena presented above, arguing that nouns frequently used in definite contexts are subject to a lexicalization process whereby the entire DP projected by such a noun and free of any other lexical material is structurally collapsed into the near equivalent of a proper name.7 Through an interaction between prosody and morphology, these items may end up in different declension subclasses from those of their bases, thus acting as suppletive forms in their slots in the paradigm. Let me provide more details.
In SC, a declension class may have more than one subclass,8 related to different realizations of one or more case forms. For instance, nouns of the feminine -a declension take three different genitive plural endings: -ii, -aa, and -aa with an additional -aa epenthesized between the sonant and the obstruent at the right edge of the base. As illustrated in (7), some nouns are subject to variation and belong to different subclasses in different varieties of SC (see Simonović and Baroni 2014 for an analysis of the -aa epenthesis).
Few -a declension nouns vary among three subclasses as in (7), some vary between two, and most are invariant.
Declension subclasses are sensitive to the prosodic and phonotactic shape of the base. The noun školjka ‘shell’ has a long falling accent on its initial syllable and a sequence of a sonant and an obstruent (ljk) stem-finally, which—together with the fact that it is a native item in SC—results in its access to multiple declension subclasses. Bisyllabic proper names ending in -a with a falling initial syllable, such as S
Vocatives are typically used in the domain of proper names: interlocutors address one another using this case form of their (sur)names, which are definite by default.
Relatively few common nouns are used in the vocative with considerable frequency. These are human-denoting nouns most typically involving a social relation or function: kinship terms, societal positions, and so on.
Apart from infrequent indefinite uses such as those discussed in section 2, vocatives of these nouns are definite. They are prototypically used in contexts where exactly one familiar individual satisfies the predicate denoted by the noun (consider the nouns učiteljica ‘teacher’ and kapetan ‘captain’, typically used in the vocative form in a classroom or in a military or naval context, respectively).
Due to its frequent bare definite use, the bare DP of these nouns switches from a productively derived form to a form that is memorized and retrieved from memory. Exemplifying with učiteljica ‘teacher’, I represent this process formally as the lexicalization of a DP, as in (10a).
Lexicalization makes the internal morphological structure of the DP opaque, turning the noun effectively into a proper name: a memorized, definite, internally noncompositional DP.9 The lexically stored DP then acts as a suppletive form, which inhibits its ability to derive a definite DP by these nouns (Kiparsky 1973): in definite contexts, the stored DP is retrieved from the lexicon. Such is the case with the addressing use of these nouns, as schematically presented in (10b), where the numeration entry with a lexically stored DP and the vocative feature is simpler for computation than the numeration entry with a noun and the definite feature selected independently and then merged into the DP structure.10
Technically, the process of lexicalization collapses two syntactic heads into one, much like the morphosyntactic process that has been modeled by Marantz (1988) as Morphological Merge, by Embick and Noyer (2001) through the operation Lower, and by Harley (2005) as conflation, and that is most directly derivable by Matushansky’s (2006) m-merger. I do not discuss the technical implementation as that would dramatically expand the range of relevant empirical observations.
4 Lexical and Prosodic Realignments
Common nouns that satisfy the conditions for this type of lexicalization, like učiteljica ‘teacher’, include a subclass crucially bearing the default stress pattern (falling accent on the initial syllable) across the case paradigm; examples are t11 Being lexically unspecified for prosody, they are not bound to a declension subclass. As lexicalization effectively turns vocatives of these nouns into vocatives of proper names, they change their declension subclass from that of common nouns lexically unspecified for prosody to that of their counterparts among proper names. And, as mentioned above, these proper names have vocatives syncretic with the nominative, as illustrated in (11).
This explains the syncretism in the regular, definite use of the -a class nouns with a falling accent on the initial syllable, illustrated in (12).
The nonsyncretic vocative in -o/-e in the indefinite use is the nonlexicalized vocative of these nouns, productively derived according to their declension subclass.
The behavior of masculine nouns like kapetan ‘captain’ and general ‘general’, which display a shift in prosodic pattern in the vocative, is explained by the same mechanism of DP-lexicalization (plausibly mediated by m-merger), which includes flattening of their internal compositional structure. Lexicalization induces opacity regarding the prosodic structure of these nouns, as discussed in Arsenijević 2010 and Simonović and Arsenijević 2014. This causes them to take the default prosodic pattern with a falling accent on the initial syllable (Zec 1999) and (in certain varieties) to switch declension subclass, as illustrated in (13).
Let me present this analysis in more detail.
A number of derivational suffixes in SC appear to be available in two versions: a stress-attracting one and one that goes with a prosody faithful to the base. The former goes hand in hand with lexicalized, often idiomatic meanings, while the latter always preserves the transparent, compositional interpretation. Consider the deadjectival nominalizing suffix -ost and the gerundial suffix -je in (14).
Simonović and Arsenijević (2014) provide arguments that the stress-attracting behavior of these suffixes derives from the structural opacity of the base. To simplify somewhat, they argue that the default stress in SC as described by Zec (1999)—the falling accent on the first syllable—surfaces only when the base is not prosodically specified in the lexicon and the affix cannot bear prosody. Affixes that can bear prosody attract stress in contexts where the base is prosodically unspecified (in other words, rather than stress-attracting, they should be called last-resort-stressable). In such cases, the affix receives a high tone, which is then mapped onto a surface form in which the high tone occurs on two syllables: the one from the input form and the one left-adjacent to it (Zec 1999, Zsiga and Zec 2013). On these grounds, the stress patterns in (14) are explained as follows in Arsenijević 2010 and Simonović and Arsenijević 2014.
Compositional nominalizations are based on full-fledged predications (i.e., PredPs), as in (15a) (Arsenijević 2010, Roy 2010). The base is base-generated within the PredP and, due to its structural transparency, remains accessible at PF, together with its lexical prosodic specification. Lexicalized nominalizations involve a base derived by collapsing the PredP into a lexically stored item that behaves as an adjectival head, as in (15b). Access to the original base with lexically specified prosody is lost, and the stressable suffix receives a high tone (H), resulting in a rising accent on the left-adjacent syllable (Zec 1999, Zsiga and Zec 2013).
I argue that a similar process is responsible both for the general stress pattern of nouns like kapetan ‘captain’ and general ‘general’ and for the stress pattern of their vocative forms. Let me first discuss their general prosodic pattern. These nouns involve loan bases, which tend to be prosodically unspecified in SC (Simonović 2015). The base combines with an accented zero suffix. Because the suffix lacks phonological material in the nominative, its high tone is realized on the final syllable of the base. The end result is a surface form with a rising accent on the left-adjacent syllable.12
In other cases the suffix has a phonological realization due to the case marking. Here, the high tone spreads from the case ending to the final syllable of the base, as in the genitive form in (17), resulting in a longrising accent on the penultimate syllable.
Let me now outline the derivation of the lexicalized vocative prosody of these nouns. As already argued regarding examples (14a–b), lexicalized DPs are structurally flattened, as in (18a). In this particular case, this means that analogously to the PredP in (15), the DP consisting of the base and the zero suffix is collapsed into an atomic chunk. As a result, the stress-attracting suffix is no longer accessible to the phonology. The entire chunk is treated as lexically unspecified, and it is assigned the default prosody: a falling accent on the first syllable (Zec 1999). This change in the item’s prosody also changes its declension subclass, and even in the variety of SC that normally uses the ending -u for nouns matching the prosodic shape of the transparent vocative, it surfaces with the ending -e, as in (18b).
The question remains whether feminine nouns with lexically specified prosody such as sèstra ‘sister’ and učitèljica ‘teacher’ are lexicalized too—it is just that their lexicalized version involves the nonsyncretic ending because their lexically specified prosody binds them to a subdeclension—or whether they do not even get lexicalized. An extreme version, pointed out by an anonymous reviewer, would be that all nouns develop both lexicalized and nonlexicalized genitives, but only some of them can be attested, because the others either are rarely ever used (proper names in indefinite contexts) or have the same ending as the productive form (feminines in -ica and many other classes). This would make the role of frequency only an illusion. As I could not design an empirical test of the lexicalized status of regularlooking vocatives, this issue remains open.
As the proposed view stipulates that proper names with a falling initial syllable form a declension subclass with a syncretism between vocative and nominative forms, and derives the behavior of tetka-type nouns in terms of their joining this subclass as a result of DP-lexicalization, it also makes the following prediction: vocative forms of proper names with a short(-falling) initial syllable will not be sensitive to definiteness. This prediction is borne out. Imagine a context in which a prophecy exists that a dead prince will be brought back to life by a kiss from a girl called Sanja. Five thousand people are gathered in the public square and the king asks:
Ima li neke Sanje među vama? Sanja / *Sanjo?
has Q some Sanja among you Sanja.VOC
‘Is there a Sanja among you? Sanja?’
I must note, however, that while the ill-formedness of the form with -o is uncontroversial, the use of the form syncretic with the nominative tends to be elusive in terms of being indefinite and at the same time vocative. The definite interpretation (where a presupposition is locally accommodated, for addressing purposes, that a person called Sanja is present) and the nominative interpretation (where the existential predicate is elided:
Ima li nekaSanja? ‘Is there a Sanja?’) cannot easily be suppressed.
The analysis in terms of lexicalization is supported by the fact that the set of masculine nouns with two vocatives is highly idiosyncratic, and even among the feminine nouns with a falling initial syllable there are certain exceptions that do not display the vocative-nominative syncretism in the definite use. Such idiosyncrasy is indeed expected if lexicalization is part of the process leading to the described situation.13
To sum up: Frequently used DPs undergo lexicalization, which flattens their internal structure and effectively turns them into proper names. Vocatives refer to the hearer and are typically definites. In kapetan-type nouns, this blocks access to the stress-attracting suffix and leaves the default prosody as the last resort. With respect to tetka-type nouns, once they turn into proper names, their lack of lexical prosodic specification classifies them in terms of declension with prosodically equivalent proper names. We arrive at an analysis in which the vocative-nominative syncretism is limited to proper names, and common nouns show it only when they are lexicalized into proper names. The remaining question—why certain proper names show the vocative-nominative syncretism—is left for future research.
I presented a curious behavior of SC vocatives, wherein two different forms occur depending on whether the nominal expression is definite or indefinite. I analyzed definite vocatives as vocatives of DPs that have undergone lexicalization, turned into a kind of proper name, and thus changed their prosodic pattern—thereby also changing their declension subclass. The latter process is conditioned by the vocatives’ lack of lexical prosodic specification.
1Serbo-Croatian is used here as the politically neutral name traditionally used in linguistics to refer to the language spoken by Bosniaks, Croats, Montenegrins, Serbs, and a number of other ethnicities of the former Yugoslavia.
3Jedna ‘one’ should not be linked with the English pronominal one, as in you auntie one; it is a different element. In fact, a direct counterpart of the English construction is ungrammatical in SC (note also that jedan would have to agree with Jovan).
*Jovane (ti) tetkasti jedan!
Jovan.M.VOC you auntie.M.SG one.M.SG
Moreover, note that in an overt predicative expression, jedna retains the same interpretation (the inverse order is due to focalization of izjelica ‘eater’ in (3), somewhat dispreferred in (ii) because a clitic is present).
Jovan je (jedna) izjelica.
Jovan.M.SG is one.F.SG eater.F.SG.VOC
‘Jovan is (such) an eater.’
However, when, instead of an adjective, a noun with the vocative-nominative syncretism is used in the expression in (i), a curious asymmetry emerges. Consider (iii).
Jovan.M.VOC aunt.F.SG.VOC one.F.SG
‘Jovan, you auntie!’
The vocative form syncretic with the nominative in (1c), which is normally used for this noun, becomes ungrammatical, and only a form with the distinctive vocative ending -o can be used, matching the declension subclasses in (1b). The syncretic form yields the interpretation that two vocatives are being used to address someone (i.e., that the speaker is calling both Jovan and his aunt), and the use of jedna becomes impossible.
4 Genitives are provided instead of nominatives when they better illustrate the prosodic shape of the base.
6 In this squib, for simplicity, I choose to model the data using a nominal structure with a projected DP, leaving aside the debate about the presence of this projection in articleless languages like SC (for arguments in favor of a DP, see Progovac 1998, Caruso 2012, Stanković 2014; for arguments against it, see Corver 1992, Zlatić 1997, Bošković 2005). Although some conclusions regarding the issue of DP in SC are explicitly suggested in the following paragraph, apart from this the analysis does not depend on this issue and could be formulated in terms of the definiteness feature (or any head carrying it) rather than the DP projection.
7 A straightforward way to account for these facts would be to invoke impoverishment rules triggered in tetka-nouns by definiteness combined with a short-falling accent on the initial syllable (Andrew Nevins, pers. comm.). While correctly capturing the data, such rules would only reformulate the question: namely, why do definiteness and the relevant prosodic shape trigger impoverishment? I try to provide an analysis that answers this question in either formulation.
I also do not address the debate about whether vocatives are NPs, per Longobardi’s (1994) rather conceptual argument that only arguments have the DP layer and that vocatives are not arguments, or whether they can also be DPs, as argued by Szabolcsi (1994) and evidenced by data from a number of languages where vocatives may include a definite article. For expository purposes, I assume the latter position, but all the arguments made here can be reformulated in terms of the former.
8 I assume that a declension class is marked on nouns in syntax as a feature. At the PF interface, this feature is interpreted as an index that links to a set of exponents of number and case, and at LF it triggers certain gender-related presuppositions (for more details, see Arsenijević 2016, 2017).
9 As an anonymous reviewer points out, the lexicalized item does not need to be treated as a DP; it is a proper name and may have anyone’s favorite proper-name status. I stick to the label DP for ease of exposition—more precisely, for ease of linking the lexicalized item with the structure that was lexicalized.
10 As with other instances of lexicalization, the present one should be seen as a historical process yielding a state of the lexicon that is then learned in the course of language acquisition rather than reproduced each time a noun from this class is used in a definite context. I thank an anonymous reviewer for pointing out the need to take a position regarding this issue.
11 In support of the role of the short-falling initial syllable, consider the nouns báka ‘grandma’ and séka ‘sissy’, with a long-rising accent on the initial syllable. As the prosodic shape predicts, these nouns have the regular vocatives bako and seko, in both the definite and the indefinite use. Among children, probably due to frequent definite use, these nouns have also developed a vocative syncretic with the nominative, b
12Zec (1999) derives the surface form in terms of spreading of the high tone. Zsiga and Zec (2013) formulate an analysis that derives phonetically more accurate lexical surface forms without resorting to spreading. To spare the reader the details of the prosodic mappings, I do not spell out the mechanism that leads from the input with a high tone on one syllable to the surface form with an accent on that syllable’s left-adjacent neighbor.
13 I am grateful to two anonymous reviewers for pointing out these two supporting arguments for the analysis: the prediction about the indefinite use of proper names, and the idiosyncratic nature of the sets of nouns showing the relevant behavior.
I thank the anonymous reviewers for valuable suggestions, as well as the Squibs and Discussion editors and Marko Simonović for fruitful discussion.