¹ I have attempted to apply the methods of analysis set forth by Zellig S. Harris in Methods in structural linguistics (Chicago, 1951; later issued in paperback form as Structural linguistics), in ‘Co-occurrence and transformation in linguistic structure’, Lg. 33.283–340 (1957), and in String analysis of sentence structure (The Hague, 1962). I aim solely at descriptive adequacy; that is, I attempt to provide a framework to describe the syntactic structure of any modern standard Serbo-Croatian utterance which has been or may be produced by a native speaker. I have also operated with transformations, but the paper does not aim to be a transformational syntax. Some of my notational practices deviate from those of strict transformationalists; thus, I sometimes alter more than one element in a single transformation. On the other hand, I refrain from certain procedures considered admissible by orthodox transformationalists, such as the positing of starred strings and the mixing of levels of analysis.

Most previous syntactic studies of Serbo-Croatian have defined syntactic categories, implicitly or explicitly, in semantic terms. Though I have found such traditional works useful as a source of examples—particularly T. Maretić, Gramatika hrvatskoga ili srpskoga književnog jezika ³ (Zagreb, 1963)—the present study does not draw upon any earlier syntactic studies, though I have occasionally commented on the semantic functions of certain types of constituents.

² Examples are partly drawn from written material (always after checking with native speakers), partly transcribed from material overheard in conversation with native speakers, and partly elicited from informants, principally my wife—all speakers of the Eastern variety of the standard language as exemplified in the educated colloquial speech of Belgrade.

³ All Serbo-Croatian examples (except in phonemic transcription) are cited in conventional Latin-letter orthography. In this paragraph hyphens indicate morpheme boundaries.

⁴ A construction at the lowest level has one or more words as constituents. Such a construction may in turn be a constituent of a larger construction. Unless otherwise stated, constructions are assumed to be continuous; that is, all constituents are juxtaposed without intercalation of extraneous elements (except enclitics).

⁵ Though traditional grammars distinguish between dative and locative singular in the noun, I lump them together as dative-locative. According to the grammars (and some linguists, e.g. E. Stankiewicz, M. Samilov), while other cases are distinguished from each other in at least some paradigms by differences in ending, dative and locative always share the same endings, except that in the singular of a few substantives they are distinguished by differences in accent (phonemically, stress). I have never met speakers who made such a distinction, except in a few fossilized expressions. I doubt that it is any longer made by the majority of speakers.

It is significant that both C. T. Hodge and T. F. Magner in their Serbo-Croatian textbooks treat the dative-locative as one unit.

⁶ Resolution here into component morphemes has no historical significance. Historically, ama is not a compound of a plus ma, but a loan from Turkish áma ‘but’.

⁷ The syntax of conditional sentences (in Serbo-Croatian a semantic, not a formal category) may be briefly summarized. In real conditions, the condition clause (protasis) is introduced by ako: Ako živi u gradu, radi u fabrici ‘If he lives in town, he works in the factory.’ Both condition clause and result clause (apodosis) have V in the temporal form or construction required by the semantic time reference: Ako je živeo u gradu, radio je u fabrici ‘If he lived in town, he worked in the factory.’ But if the time reference of the condition clause is future, its V is either in the present (favored mainly in perfective verbs or imperfective verbs with no perfective counterparts) or in the dependent future construction: Ako bude živeo u gradu, radiée u fabrici (or Ako živi u gradu, …) ‘If he lives in town (future), he'll work in the factory.’ The normal (i.e. nondependent) future construction occurs in ako-clauses (Ako če živeti u gradu, …), but much more rarely than the dependent future or present.

An alternative but archaic construction of the condition clause dispenses with the introducer, but inserts the interrogative particle li: Živi li u gradu, on radi u fabrici ‘If he lives, etc.‘

In unreal conditions the V of the result clause is in the conditional construction. The condition clause is introduced by da (i.e. it is daS; see below); in present unreal conditions the V is in the present form, in past unreal conditions the V of daS is in the past construction: Da živi u gradu, on bi radio u fabrici ‘If he lived in town, he'd work in the factory’; Da je živeo u gradu, on bi radio u fabrici ‘If he'd lived in town, he'd have worked in the factory.’ The present unreal condition is also expressed with the V in both clauses in the conditional construction and the condition clause introduced by ako ‘if’ or kad ‘when’: Ako (or Kad) bi živeo u gradu, on bi radio u fabrici ‘If he lived in town, he'd work in the factory.‘

⁸ kA is discussed in the section on kS below.

⁹ Note that D in ND cannot be analysed as a modifier of V in the sentence of which N is a constituent. Take the sentence Mi spremamo ovde ručak za policajca tamo ‘We're preparing lunch here for the policeman over there.’ Not only is the D tamo ‘there’ not in the usual position of D complements to the V, but one does not find such a sequence as V ovde … tamo, except when the two D are linked by C, which is not the case here.

¹⁰ The numerals dva ‘two’, tri ‘three’, and četiri ‘four’ are declined; those from ‘five’ up are indeclinable. In compound numerals of which the last constituent is ‘two’, ‘three’, or ‘four’, that constituent is declined like the simplex. The indeclinables function only as phrase heads; a noun attribute to them is in the genitive plural. The declined numerals function as attributes in oblique cases, but in the nominative and accusative they function, like the indeclinables, only as heads; nouns attribute to them are in the genitive singular. There is a special form of masculine-neuter genitive singular in the adjective declension occurring only in adjectives attribute to such a Ng. Finally, though declined forms of dva, tri, and perhaps četiri do occur colloquially, there is an increasing tendency in colloquial speech to treat these too as indeclinable.

¹¹ In appositional N, D attribute may precede, rather than follow, the head: Goce, već mladič ‘Goce, already a young man’. Note that the placement of već ‘already’ before mladić ‘young man’ reflects the sequence of complements in the sentence of which the appositional N is a transform: Goce je već mladić ‘Goce is already a young man.‘

¹² Possessive adjectives (posA) derived from substantives are replacives for Ng, with appropriate shift in word-order. Possessive adjectives are derived from a subclass of nouns indicating persons; they are derived by addition of a suffix (-ov- to θ/o-class substantives, -in- to a-class substantives). They replace the genitive case of such substantives either as attributes in a noun phrase or as predicate complements in a sentence.

Possessive adjectives derived from pronouns stand to them in suppletive relation and are in complementary syntactic distribution with the pronominal genitive case. The latter occurs only as object of a verb or preposition, the pronominal posA being obligatory as predicate complement or attribute in a noun phrase.

¹³ A small class of words—(koliko ‘how much, how many’ and words that answer koliko) are here called quantifiers. They have the peculiarity that they function as both N and D. Multiple class membership is uncommon in Serbo-Croatian, since morpho-syntactic classes are typically marked by inflexional suffixes; aside from quantifiers there are a few words, traditionally classed as adverbs, which can appear in both D and N slots; e.g. danas ‘today’, sutra ‘tomorrow’, juče(r) ‘yesterday’. The quantifiers include, beside the numerals, koliko ‘how much, how many?’, toliko ‘so much, so many’, ovoliko ‘this much, this many’, onoliko ‘that much, that many’, mnogo ‘much, many’, malo ‘little, few’, dosta ‘enough’, vise ‘more’, suviše ‘too much, too many’, premalo ‘too little, too few’. Except for the numerals dva, tri, and četiri, the quantifiers are indeclinable, though corresponding to most of them there are adjectives formed on the same stem, e.g. viši ‘higher’, mnogi (plural only) ‘many’.

¹⁴ A few D (čak ‘even’, samo ‘only’, bar, barem ‘at least’, skoro ‘almost’, tek ‘only, at the time that’, the negative particle ne, and perhaps some others) function as special phrase modifiers and usually (the negative particle always) stand before the phrase they modify, notwithstanding that D attributes normally follow the head in the noun phrase: čak ovi dobri ljudi ‘even these good people’, samo u Beogradu ‘only in Belgrade’, barem naši prijatelji ‘at least our friends’, ne ta velika flaša ‘not that big bottle’ ne o Jovanu ‘not about John’. They may also function as D constituent on the sentence level, in which case they usually precede a simple verb or, in a verbal construction, are intercalated between auxiliary and main verb, except that the negative particle is always proclitic to the verb (to the auxiliary in a verbal construction): On samo spava ‘He just sleeps (and nothing else)'; On je čak platio naše dugove ‘He even paid our debts'; Jovan ne razume engleski ‘John doesn't understand English’; Neću se vratiti ‘I won't return.‘

When ne is prefixed to the V the whole sentence is negated. This is in keeping with the fact that the verb is the head of the sentence, other constituents being its attributes.

¹⁵ P requires the N following to be in a particular case or cases. A full list of prepositions with their case selection (traditionally termed government or rectio) belongs to the lexicon. The item kao is a preposition, since in conjunction with a following phrase it transforms into a D modifier. It has the peculiarity that it does not govern a particular case, but that the N after it shows case agreement with the N which the kao phrase (PN) modifies. If the kao phrase does not modify a N, the N after kao is in the nominative case.

¹⁶ The Serbo-Croatian orthography is inconsistent in writing proclitics and other unstressed elements: sometimes together with a neighboring word, sometimes separated by a space. In pošto we obviously have the preposition po plus što; cf. po in the meaning ‘after’ in PD popodne ‘afternoon’ and in PN po objedu ‘after dinner’ (Croatian usage), po Uskrsu ‘after Easter’.

¹⁷ Note that when samo što is used in the sense ‘only when, just as’, we are not dealing with an instance of deleted zato (i.e. samo što and samo zato što are not here interchangeable); samo (zato) što occurs solely in the sense ‘only because’.

¹⁸ Nego precedes štoS only in the function ‘than’ after a comparative; nego plus a sentence not introduced by što is adversative: Ja sam ga zvao, nego on nije hteo doći ‘I called him, but he wouldn't come’.

¹⁹ The k-morpheme appears in those elements traditionally called interrogative or relative pronouns, adjectives, and adverbs (here termed k-words). It is so designated because it usually appears as /k/ or as one of the typical morphophonemic alternants of /k/.

²⁰ Normal (neutral) word order for kS is the same as that of any S, except that the phrase containing a k-word is shifted to absolute initial position, preceded only by an introducer (C): Koga si video? ‘Whom did you see?'; O čemu ste govorili? ‘What did you talk about?’ (prepositional phrase containing k-word in initial position); Samo ko je preživeo rat ima pravo da govori o tome ‘Only someone who has been through a war has the right to talk about that’ (k-word as phrase head preceded by special phrasal modifier; k-word phrase in initial position); A šta si ti radio? ‘And what did you do?’ (kS with k-word in initial position preceded by C). As in all sentences, a departure from normal word order in kS signals emphasis; an item to be emphasized is shifted so as to precede the k-word: A ti šta si radio? ‘And you, what did you do?'

²¹ Čiji is a k-posA and as such replaces Ng, specifically in function (a) the genitive of an unknown or variable element and in function (b) the genitive of a constituent of the superordinate construction with which kS is in apposition.

²² Use of qSi in place of daS is most frequent in the speech of the western part of the Serbo-Croatian speech area.

²³ In archaic language an enclitic allomorph of njega ‘him’ occurs with a preposition, which then bears the stress: na nj /náanj/ ‘to him’.

²⁴ In dialect and nonstandard speech of certain regions, particularly the Adriatic coast, enclitics often occur initially in a segment.

²⁵ To avoid a sequence of two enclitics of identical shape /je/ (feminine accusative third singular pronoun and third singular present of biti ‘to be’), the literary language has an allomorph /ju/ for the pronoun which occurs in conjunction with verbal /je/. In the spoken language /ju/ is not used; with verbal /je/, the full, nonenclitic form /njúu/ occurs. As the occurrence of /njúu/ is automatic in this case, there is no connotation of emphasis: thus: On ju je video (literary) ~ On je nju video (colloquial) ‘He saw her.'

²⁶ Curiously enough, dali can be prefixed to daS. The semantic effect is the same as if /↑/ were substituted for /↓/.

²⁷ Cf. the rather formal tag question eliciting agreement, zar plus negative particle, i.e. zar ne ‘nicht wahr?': On je došao, zar ne? ‘He arrived, didn't he?’ A more familiar tag question with approximately the same meaning is jel’, jel' te. Tag questions are independent S, but, like Nv, usually occur under one intonation pattern with the preceding sentence.