What Turkish Acquisition Tells Us about Underlying Word Order and Scrambling

Natalie Batman-Ratyosyan & Karin Stromswold
Rutgers University

Purpose. This study investigates whether young Turkish speaking children treat Turkish as a non configurational free word-order language or whether they are initially predisposed to prefer certain word orders that are structured hierarchically.

About the language. The pragmatically unmarked word order in Turkish is SOV, however, a basic sentence with three constituents can have six possible orderings. While the subject initial sentences are the most natural the verb initial sentences are the least natural1. It is an agglutinative language with rich case morphology. The nominative case is null and the accusative case is overt but it can also be dropped in which case there are only two possible word-orders, SOV and OVS.

Subjects. Subjects were 31 monolingual Turkish-speaking children between the ages of 2;10 and 5;8. Subjects were divided into 3 groups with mean ages of 3;3, 4;2 and 5;2, respectively.

Stimuli. The stimuli comprised of sentences with a grammatical ordering (SOV and OVS) with unmarked direct objects and with an ungrammatical ordering (OSV and SVO) with unmarked direct objects.


1)  Siyah kar  Inca-lar   yem   topla-dI-lar.            (SOV) 
      Black    ant-Plur   food  gather-Past-3rdPlur. 
    'The Black ants gathered food.'

2)  Mektup   yaz-dI        sarIsIn   kIz.                (OVS)
    Letter   write-Past    blond     girl. 
    'The blond girl wrote letter/letters.'

3)  *El    neseli  yolcu-lar       salla-dI-lar.         (OSV)
    Hand  happy   passenger-Plur  wave-Past-3rd Plur.	
    'The happy passengers waved their hand/hands.'

4)  *Tembel  kapIcI   boya-dI     duvar.                 (SVO)
      Lazy   genitor  paint-Past  wall. 
    'The lazy genitor/doorman painted wall/walls.'

Procedure. Children imitated 32 sentences in total (see Table 1). Sentences varied in length such that half were long and half were short. Each sentence comprised of a subject NP that contained an adjective, an object NP which was an uninflected noun and the verb. Measures were also taken to control for order of presentation, choice of child-appropriate words, etc. Half of the children in the oldest group judged rather than imitated the sentences. Twenty-six adult Turkish speakers rated on a 5 point scale, the acceptability of the sentences used.

Table 1. The linguistic stimuli used in experiment 1.

        32 Sentences with Scrambled Constituents
          16 Grammatical      16 Ungrammatical
           8 SOV  8 OVS        8 SVO  8 OSV
           4 PL   4 PL         4 PL   4 PL
           4 SG   4 SG         4 SG   4 SG

Results. A 3 (Age) x 2 (Length) x 4 (SOV, OVS, OSV, SVO) NOVA with correct vs. incorrect imitation as the dependent variable revealed the following significant main effects and interactions. (Planned pair-wise comparisons were also performed and the results used to determine the source of main effects and interactions). There was a significant main effect for sentence type (mean correct imitation for SOV = 72%, OVS = 60%, SVO = 46%, OSV = 43%), F(3,84) = 23.55, p<.0005. There were also significant main effects for Age (F(2,28)=21.12, p<.0005, older children were more successful imitators than younger children) and length (F(1,28)=69.29, p.<.0005, short sentences were imitated better than long sentences). There were significant interactions between word order and age (F(6,84)=2.65, p<.05, older children were more successful imitating ungrammatical word orders than younger children) and between word order and length (F(3,84)= 10.94, p<.0005, children imitated long, ungrammatical word order sentences very poorly). There was also a significant 3-way interaction among word order, length, and age (F(6,84)= 3.87, p<.005. The youngest children did poorly on all but the short SOV sentences; and the oldest children did well (85% correct or better) on all of the short sentences and the long grammatical sentences, and less well on the long ungrammatical sentences (69% for SVO and 44% correct for OSV).

The results of our study suggest that, even though SOV and OVS orders are equally good in adult Turkish, young Turkish speaking children treat SOV word order as being the primary word order in Turkish. Consistent with Otsu's (1994) for Japanese and Slobin and Bever's (1982) results for case-marked constructions in Turkish, these acquisitional data indicate that even when children are exposed linguistic input which indicates they are learning a free word order language, they are initially predisposed to treat one of these orders as being the word order for their language. These data also provide support to linguistic theories which argue that Turkish has an underlying word order and the apparent free word order of Turkish is the result of scrambling (Kornfilt 1994, Kural 1992).

1 Slobin and Bever (1982) analyzed the relative frequency of the different word orders in children's and adults' speech. Looking only at sentences which contained inflected NPs, they found that the relative frequency of NNV, NVN, and VNN were similar for adults and children. All of children's and adult's VNN sentences were VSO. Eighty-seven percent of children's NNV utterances were SOV, and 86% of adults' NNV utterances were SOV. Of the NVN utterances, 36% of the children's and 66% of the adults' were SVO.


< Back to List of Abstracts | Back to PLC23 Home Page >

About the PLC23 Committee
Previously held Penn Linguistics Colloquium: PLC22 (1998), PLC21 (1997)

Penn Department of Linguistics
University of Pennsylvania