Some word-order effects in Serbo-Croat.

In the SerboCroatian language, the relative order of subject (S), verb (V), and object (a) is flexible. All six of the pe1'mutations of those elements have identical words, meaning, and voice, and all six are grammatically acceptable. Nonetheless, sva is the dominant form. The psychological reality of this dominance was assessed in three tasb. S va was associated with the shortest latencies (and S a forms in general were faster than as forms) when subjects were as"~ed to evaluate the plausibility of a sentence (Experiment 1) or to initiate an utterance (Experiment 2). This advantage did not obtain in lexical decision (Experiments 3 and 4). Results were discussed in the conte;rt of linguistic universals of word order and Forster's (1979) model of the language processor, The relative order of subject, verb, and object is fairly flexible in Serbo-Croat, the major language of Yugoslavia. Of the six possible orders (SVO, SOY, VSO, VOS, OSV, and OVS), all have the same referential meaning and all are considered to be grammatically acceptable (Belie, 1933). Relevant grammatical information is carried by case inflections on nouns rather than word order. For example, in GRANATA POGADA PALATU, GRANATA PALATU POGADA, etc. ("the grenade strikes the palace"), the nominative (grenade) and accusative (palace) are distinct and the direction of action is clear. With respect to what we might term word order "preferences" demonstrated by the world's languages, Greenberg (1966) has noted that the vast majority have several variants but one dominant order. Of the six possibilities, however, all are not equally likely to dominate. In fact, orders in which the object precedes the subject (VOS, OSV, and OVS) are quite rare. Of the three so-called common types, SVO is the most frequent and it is this order that dominates in the Serbo-Croat.ian language (Greenberg, 1966). The issue t.o be addressed is whether or not. t.he universal rarity of OS constructions relative to SO constructions has a parallel in word order preferences of readers of a language in which bot.h constructions are equally acceptable and equally * Language and Speech, 1986, 29, 177-195. ** Universit.y of Belgrade t Stat.e Universit.y of New York at. Binghamt.on t Also Universit.y of Connecticut. Aknowledgment. This research was supported by NICHD Grants HD-08495 and HD-01994 to the Universit.y of Belgrade and Haskins Laborat.ories, respectively. The authors would like t.o thank Kenneth Forster and an anonymous reviewer for helpful suggestions. =HASKINS LABORATORIES: Status Report on Speech Research SR-88 (1986) =