Being a case-based language, Lithuanian is very flexible with respect to word order (Klimas, 2002). For example, the sentence „The girl threw the ball“, which in English can only be expressed as presented, in Lithuanian can be stated in all possible permutations: subject-verb-object (SVO, SOV, VSO, VOS, OSV, & OVS). The goal of this study was to explore the relationship between the different syntactic arrangements and memory for sentence content and reconstruction of syntactic word order.
Two hypotheses were taken into account with respect to memory for sentence content. The first may be called the word order proficiency hypothesis. Since Lithuanian shows a preference for SVO and SOV word order sentences (Ambrazas, 1994), one would expect these sentences to occur more frequently, appear less complex, and possibly involve a reduced memory load (see Clifton & Duffy, 2001; Hemforth & Konieczny, 2000). If that is the case, SVO and SOV sentences should be recalled significantly better than VSO, VOS, OSV, and OVS sentences. The second is the semantic identity hypothesis. Research has consistently shown that sentences are recalled mainuly in terms of their semantic content, whereas surface information is lost very rapidly (Brewer & Hay, 1984; Sachs, 1967). Since in Lithuanian the different word order sentences carry the same underlying semantic content, their recall should thus be the same. Regarding reconstruction of word order, subjects tend to reconstruct sentences according to prevailing forms (Brewer & Hay, 1984). Hence, the prediction was made that the non-canonical syntactic order sentences (VSO, VOS, OSV, & OVS) would tend to mirgate over to the canonical forms (SVO & SOV) during the process of recall.
The participants in the study were 72 native Lithuanian speakers who had recently immigrated to the United States. They were told that they would be asked to read a list of sentences and afterwards try to remember as many of them as possible. During the course of individual testing, the participant was presented a list of 24 sentences printed on separate cards. Six of the sentences served as a permanent buffer, three at the beginning and three at the end of the list. The 18 critical sentences that stood in the middle were divided into six groups of three sentences. The respective six syntactic word orders (SVO, SOV, VSO, VOS, OSV, and OVS) were counterbalanced across the six groups of three sentences. The participants had 5 sec. to look at each card, read the sentence, and then flip over to the next sentence. After seeing the whole list, the participant was asked to recall as many of the sentences as possible by writing them down on a lined sheet of paper. Following free recall the participant was given a cued recall test. Each participant was handed an 18-page booklet with one of the critical sentence words: subject, verb, or object, printed in either the first, second, or third position with the task to write in the two missing words. Each of the three sentence words was counterbalanced across the three positions.
For both free recall and cued recall, a repeated measures ANOVA of correct recall revealed no significant difference amongst the six types of presented sentences, F(5, 355) < 1 for both. Hence, the results support the semantic identity hypothesis over the word order proficiency hypothesis. Additional support for the semantic identity hypothesis came from the results of cued recall by word type. Apparently, subject, verb, and object were equally effective in facilitating recall of the remaining sentence, F(2, 142) = 1,22, p > .05. However, as to the kind of syntactic word order used in recalling the sentences, the difference amongst the six sentence orders was highly significant, F(5, 355) = 20.51, p < .001. Specifically, the two subjectbefore-object syntactic forms (SVO & SOV) were used far more often than the non-canonical forms (VSO, VOS, OSV, & OVS). The results of cued recall revealed a similar pattern.
In conclusion, different syntactic arrangements do not seem to affect the semantic retention of sentences. However, as they are being retrieved, the terms of the non-canonical syntactic order sentences are recast into their natural thematic relations, thus giving prominence to the subject as the agent of the underlying argument structure.