Do non-native languages have an effect on word order processing in first language Turkish?

Abstract

Aims and Objectives/Purpose/Research Questions:

Existing studies on sentence processing in bi-/multilinguals are typically centred on the first language (L1) influence on second language sentence processing. However, there is almost no evidence of influence in the other direction. The aim of this study is to find out whether being mono-, bi-, tri- or plurilingual has an effect on reading times (RTs) in the native language. To this end, Turkish native speakers’ RTs are measured when processing Turkish canonical subject–object–verb sentences, subject–verb–object (SVO) sentences where constituents move to post-verbal positions and SVO–ki sentences where post-verbal constituents are base generated.

Design/Methodology/Approach:

A non-cumulative self-paced reading task is used in order to measure the RTs of a sentence. The area of interest contains (i) the critical verb, (ii) the verb of the complement clause and (iii) the argument or adjunct of the complement clause (32 sentences + 12 filler sentences). All elements are matched according to their frequency of occurrence and their syllable structure.

Data and Analysis:

Analyses of variance are performed on RTs of the area of interest.

Findings/Conclusions:

One of the main findings in this study is that all three sentence types are processed significantly slower by the monolingual group than by the bi- and multilingual groups. We infer that non-native languages have a positive effect on processing the word order in the L1, which might lead to a faster processing in the three sentence types. The findings are discussed in terms of working memory and the “gap-driven strategy”.

Originality and Significance/Implications:

The results are interpreted from psycholinguistic and syntactic points of view.

Keywords

Multilingualism sentence processing Turkish canonical and non-canonical word orders

Introduction

Existing studies on sentence processing in bi-/multilinguals are typically centred on the influence of the first language (L1) on sentence processing in the second language (L2) (Flege & Frieda, 1997; Liszka, 2004; Song & Andrews, 2009). However, there are almost no data on the influence of any non-native languages on the L1. The purpose of this study is to fill in this gap by investigating whether there is a difference in sentence processing between monolinguals, bilinguals and multilinguals when processing their L1. The study examines this question from two different perspectives: psycholinguistic and syntactic.

We want to explore the effect of knowing two or more languages on on-line word order (WO) processing in the L1 and to find out whether knowing languages can be counted as experience (in the sense of Bialystok, 2009) and exerts, consequently, influence on sentence processing in the native language. For this purpose, using a self-paced reading-task, the study compares the reading times (RTs) for Turkish sentences of four groups of Turkish native speakers (monolinguals, bilinguals, trilinguals and multilinguals) to find out whether there are differences in RTs when processing canonical subject–object–verb (SOV), scrambled subject–verb–object (SVO) and SVO–ki sentence orders.

Accordingly, our first research question is as follows.

Does knowing language(s) other than the L1 cause differences in the RTs of the participants?

One hypothesis concerning this research question is that languages in the mind co-exist, so that the non-native languages have an interfering effect when processing WO in the L1. If so, this would cause slower RTs in the bi- and multilingual groups compared to the monolingual group. On the other hand, bilingual experience (Bialystok, 2009, p. 3) might have an impact on processing WOs in different languages, which can lead to an enhancing effect on working memory (WM) capacity. In this case, RTs would be expected to be shorter in the bi- and multilingual groups when compared to the monolingual group. A further hypothesis is that each additional language might have an impact on WO processing. If so, there would be a difference among all groups in RTs. We hypothesize that the monolinguals would exhibit longer RTs for all sentence types, which would indicate that knowing languages has an enhancing effect on WM capacity during reading.

The second research question is related to the impact of WO variation on processing.

Will the WO variation (canonical SOV, scrambled SVO, SVO–ki) cause a difference in RTs of the participants?

The first possible hypothesis concerning this research question is that due to the distance between the scrambled Noun Phrase (NP) and its trace (i.e. an empty category from where the NP has moved), processing cost might increase, so that RTs for SVO sentences would be longer. Since there is no scrambling in the SVO–ki sentences, RTs would be shorter than in the SVO sentences. The second hypothesis, on the other hand, is that the parser predicts a finite verb immediately after the subject of a sentence, which would result in shorter RTs for SVO sentences. Regarding the SVO–ki sentences, since the finite verb occurs immediately after the subject, RTs would be shorter than in the canonical SOV sentences.

Background to the study

Online sentence processing and WM capacity

WM is considered to be a specialized memory system that allows a small amount of information to be simultaneously maintained and processed for a short period of time during the performance of a task (Baddeley, 1986, 2003). Related to the processing of a sentence, this means that the reader has to understand the structural relations among its words and phrases. The syntactic processing system reconstructs the structure of the sentence incrementally by assigning the perceived words to phrases as quickly as possible and by determining hierarchical relations among the different phrases, a process often referred to as syntactic structure building (Fiebach, Schlesewsky, Lohmann, von Cramon, & Friederici, 2005). This is important evidence for the involvement of WM during sentence processing.

WM is typically measured by tasks such as Daneman and Carpenter’s (1980) reading-span task; few studies have investigated the relationship between on-line language processing and verbal WM. Syntactic processing often involves relating words to one another over a distance (Waters & Caplan, 2004). The distance over which these integrations occur and the complexity of the integration processes are claimed to cause differences in local WM load, due to differences in the demands sentences place on syntactic processing. Although sentence processing is largely unconscious and apparently effortless, there are measurable effects of increases in the processing load. RTs in self-paced word-by-word reading tend to increase at points in a sentence where models of sentence processing predict an increased processing load. Results of some studies have been said to show that these on-line measures of syntactic processing load differ as a function of WM (Waters & Caplan, 2004). In a more recent study, Caplan and Waters (2013) argue that WM supports retrieval in points of high processing load, which are identified by regressive saccades and longer self-paced RTs that enable better comprehension. Nicenboim, Vasishth, Gattei, Sigman, and Kliegl (2015) state that “there is a wealth of evidence showing that increasing the distance between an argument and its head hinders underlying memory processes in some way” (p. 2).

Taken together, these findings indicate that WO processing requires attention and the involvement of the WM and the differences in RTs during sentence processing reflect the processing load and cognitive effort.

Processing main WO and scrambled sentences

Studies on sentence processing: SVO or SOV?

Weyerts, Penke, Münte, Heinze, and Clahsen (2002) investigate the on-line comprehension of correct and incorrect WO in main and embedded clauses in German. According to Weyerts et al. (2002), the parser predicts a finite verb after having received the subject of a sentence. They add that intervening material (i.e. the object) between the point where the verb is predicted to appear and its actual occurrence leads to extra memory costs (see also Gibson, 1998). Assuming that ungrammatical sentences require more parsing effort than corresponding grammatical ones, RTs for sentences with incorrect verb placement should be significantly longer than RTs for sentences with correct verb placement. The overall results indicate that native speakers of German prefer to process finite verbs immediately after the subject and before the object rather than at the end of the sentence. They account for this SVO preference in terms of differences in processing costs.

Unlike German, Serbo-Croatian has canonical SVO order, but scrambling is unrestricted with respect to the position of the verb, making it possible to create numerous WO configurations. Stojanović (1999) reports a series of experiments on the processing of different WOs in Serbo-Croatian. Using a self-paced scrambling experiment, she tests four types of ambiguous sentences. Stojanović’s findings reveal shorter RTs for SVO order than for SOV order (4560 ms versus 4904 ms) and shorter RTs for object–verb–subject (OVS) order than for object–subject–verb (OSV; 4744 ms versus 5024 ms).

Schneider and Indefrey (2007) investigate the influence of L1 WO on L2 sentence processing. In a self-paced reading experiment, in which L1 Dutch speakers and German and Turkish learners of Dutch participate, they report significantly longer RTs for simple declarative sentences in SOV compared to both SVO and verb–subject–object (VSO). German learners of Dutch read SOV sentences more slowly than SVO sentences but not more slowly than VSO sentences. Turkish learners of Dutch do not show any difference between SOV and the other WOs. Schneider and Indefrey conclude that the L1 has an influence on L2 processing, since German speakers show partly the same RT pattern as the L1 Dutch control group, while Turkish speakers show a RT pattern that is both different from German learners and the control group. According to Schneider and Indefrey, these results cast doubt on the assumption of SVO as a universal WO preference for L2 sentence processing.

On the other hand, there are some studies that examine WO preferences in a language that has canonical SOV order, for example, Japanese. Mazuka, Itoh, and Kondo (2002) found that scrambled sentences take longer to read than comparable sentences in the canonical SOV order, but Yamashita (1997) and Nakayama (1995) do not report such a difference. In an overview, Sekerina (2003) concludes on the basis of behavioural studies that sentence processing is sensitive to WO, at least in declarative sentences. In these studies, the canonical WO is processed faster and with greater ease, presumably because it involves a simpler computation than derived ones. Even in highly inflected and free WO languages like Basque, there is clear behavioural and electrophysiological evidence that canonical order is easier and faster to process. Erdocia, Laka, Mestres-Missé, and Rodriguez-Fornells (2009) interpret their finding as a universal design property of a language, despite differences in case-marking, verb agreement and other variable specifications of a given grammar.

We conclude that some studies point to a preference for SVO order, while some studies state that the canonical WO is preferred. Although in Turkish the canonical WO is SOV, it is also possible to build correct Turkish sentences with a scrambled WO. For this reason, Turkish provides an opportunity to investigate the relation between WO preferences and processing ease (Aydın & Cedden, 2010).

Explanations for processing scrambled sentences

In the literature, it is claimed that associating a scrambled constituent with the corresponding gap, referred to as its “trace”, becomes more difficult as the distance between the scrambled constituent and its trace increases¹ (Gibson, 1998; Kaan, Harris, Gibson, & Holcomb, 2000; Miyamoto & Takahashi, 2002, 2004; Nakano, Fesler, & Clahsen, 2002). According to these studies, the distance between the scrambled object and its trace might exceed the memory span (Nakano et al., 2002) or, under certain conditions, antecedents held in WM may be lost or difficult to retrieve during the processing of filler–gap dependencies (McElree, 2000). In case of leftward movement, the preceding filler strongly signals the presence of a gap, so that the parser actively predicts a possible gap position. The Active Filler Strategy (Frazier, 1987; Frazier & Clifton, 1989) describes this filler-driven dependency in the processing of leftward scrambled sentences in Turkish. The parser, after identifying the filler, will actively look for a gap position to assign the filler. When a potential gap position appears, the parser immediately associates it with the filler. However, in rightward scrambled sentences, in which the sentential object moves to the right of the main verb, the filler follows the gap; therefore, the parser has to seek out a filler in the incoming material, selecting the first available candidate without waiting to see if there are other or better ones. This seeking can occur via an “Active Gap Strategy” analogous to the Active Filler Strategy (Ng, 2008). On the other hand, a filler-driven explanation can be constructed following Lin (2006), who proposes the Incremental Minimalist Parser, which builds syntactic structures incrementally from left to right, with operators searching to bind their variables downwards in a syntactic tree. All these approaches yield the prediction that the RTs of canonical SOV sentences and SVO–ki sentences would be shorter than the RTs of SVO sentences in Turkish.

Word order properties of the main three languages involved in the study

In this study, the participants’ L1 is Turkish, and English and German are their L2s. This section presents some relevant properties of the languages involved.

Although the canonical WO in Turkish is SOV, Turkish allows for scrambling, that is, the optional linear reordering of verbal arguments, as seen in example (1).² In German and English, on the other hand, in main clauses with simple verb constructions the finite verb stays in the second position ((2a) and (3a)). Unlike English with a relatively fixed SVO order (see (3)), German also allows for scrambling. Contrary to German, constituents in Turkish can move both to the sentence-initial position as in (1b) and to the post verbal position as in (1c) and (1d); that is, Turkish has rightward scrambling. Example (1c) shows that the post-verbal scrambled structures in Turkish have similar WO to English and the simple verb constructions of German main clauses. In this study, we will examine the effects of canonical WO, as in (1a) and of post-verbal scrambled structures in Turkish, as seen in (1c).

(1) a. Ali kitab-ı oku-du SOV

Ali book-acc read-past

‘The child read the book’

b. Kitab-ı_i Ali t_i oku-du OSV

c. Ali t_i oku-du kitab-ı_i SVO

d. t_i kitab-ı oku-du Ali_i OVS

(2) a. Johann liest das Buch SVO

Johann read-past book.acc

b. das Buch_i liest Johann t_i OSV

(3) a. John read the book SVO

b. *The book_i read John t_i OVS

As for embedded clauses, verbal elements appear in clause-final in German embedded clauses (4b). In generative analyses of German, SOV order in embedded clauses is generally taken to be the base order from which main clause verb-second is derived by movement (e.g. Vikner & Schwartz, 1996). This shows that Turkish and German are head-final languages ((4a) and (4b)), while English is a head-initial language (4c) – that is, the complement occurs after the head in verb phrases (VPs) in both main and embedded clauses in English.

(4) a. Siz [ Ali-nin kitab-ı oku-duğ-un-u] san-ıyor-sunuz SOV

you Ali-gen book-acc read-nomin-3sg-acc think-pres-3pl

b. Sie glauben, [dass Johann das Buch gelesen hat] SOV

you think that Johann the book read has

c. You think [that John has read the book] SVO

Turkish has another structure, rather limited in use, which was borrowed from Persian and has a SVO order without rightward scrambling. In these structures, the complement clause is fully tensed, the agreement morphology is that found in main clauses and the complement clause is introduced by the complementizer ki. In contrast to scrambled SVO sentences (cf. (4a) and (5a)) these ki–clauses, as in (5b), are not adjoined to clauses, but rather are base-generated in their surface position. Because of this, the complement clause introduced by ki cannot precede the matrix verb in ki–clauses (5c) (Veld, 1998):

(5) a. Siz t_i san-ıyor-sunuz [ Ali-nin kitab-ı oku-duğ-un-u]_i scrambled SVO

you think-pres-2pl Ali.gen book-acc read-nomin-3sg-acc

b. Siz san-ıyor-sunuz ki [ Ali kitab-ı oku-du] SVO–ki

you think-pres-3pl that Ali.nom book-acc read-past

‘You think that John has read the book’

c. *Siz ki [ Ali kitab-ı oku-du] san-ıyor-sunuz

you that Ali.nom book-acc read-past think-pres-3pl

The study

Participants

The data were collected from 117 participants whose ages ranged from 18 to 46 (M = 24.5), and who were enrolled in graduate programmes or worked at Ankara University or Middle East Technical University in Turkey. All the participants were native speakers of Turkish. The participants were divided into four groups: a monolingual group (L1 group), a bilingual group (L2 group) of Turkish-English speakers and two multilingual groups: a third language (L3) group (trilinguals) who were Turkish-English-German speakers and a L3+ group (plurilinguals) who spoke more than these three languages (Table 1). All the participants had Turkish as their L1 and learnt English or German as their L2 or L3 after the age of 12. Twenty-five participants stated that they do not know any other language than Turkish.

Table 1.

Age and gender of participants in each group.

Groups	n	Mean age	Female	Male
L1 group	25	32.5	9	16
L2 group	47	20.3	34	12
L3 group	19	21.8	14	5
L3+ group	26	21.6	19	7
Total	117	24.5	76	40

The bilingual participants (L2 group) started learning English in secondary school at the age of 12–13. Being students in language departments, they had to take a language proficiency exam before they could start their departmental education. The trilingual participants (L3 group), on the other hand, started learning German as their L2 in secondary school at the age of 12–13 and started learning English at Middle East Technical University, an English medium university. All the trilingual participants attended a two-semester intensive English course. The additional language of the participants in the L3+ group was Arabic, Chinese, French, Greek, Italian, Japanese, Russian or Spanish. Due to the diversity of the languages involved, it is not easy to assess the proficiency level of the participants in this group. Participants’ statements indicated that their level of proficiency in these languages varied from intermediate to advanced. We accepted these self-reports as valid.

Task

A non-cumulative self-paced reading task (Aydın & Cedden, 2010; Cedden & Aydın, 2008) was used in order to measure the RTs via SuperLab 5.0. At the beginning of a trial, all the words appeared masked by underscores. The participants had to press the space bar to reveal the next word or phrase. As the next word or phrase appeared, the previous one was masked again by underscores. The time between key-presses was recorded as the RT for the word. Before the task, the participants were presented with a preparation task, which included three sentences. The sentences in (6)–(8) are examples of the type of test sentences that were used for the three experimental conditions.

(6) SOV–canonical WO

Genel müdür kadro-nun açıl-acağ-ın-ı müjdele-di.

general director staff-gen expand-part-3sg-acc tell-past.3sg

‘The general director said that the staff will be expanded’

(7) SVO–scrambled sentences

Çalışan-lar isti-yor başkan-ın istifa et-me-sin-i.

worker-pl want-pres.3sg president-gen resignation aux-inf-3sg-acc.

‘Workers want the president to resign’

(8) SVO–ki sentences

Siz-ler düşün-ün ki kaçakçı-lar yakala-n-dı-lar.

you-pl think-imp that smuggler-pl be arrested-PASS-PAST-3PL

‘You assume that smugglers are arrested’

Only the italicized field in (6)–(8) was calculated by SuperLab 5.0. This field contains (i) the critical verb, (ii) the verb of the complement clause and (iii) the argument or adjunct³ of the complement clause. All the elements shown by the examples in italics are matched according to their frequency of occurrence and their syllable structure. The length (in characters) of the region of interest (i.e. the italicized field) was balanced across conditions: 28.75 (SE = 0.37) for canonical SOV, 29.91 (SE = 0.61) for scrambled SVO and 30.25 (SE = 0.39) for SVO–ki sentences.

There were 32 experimental sentences: 12 for each syntactic type. None of the content words in the experimental sentences were ever repeated.⁴ Twelve filler sentences were constructed covering various other syntactic constructions.

Data analysis

Table 2 gives descriptive statistics of the mean RTs of each group for each sentence type. When comparing the four groups by collapsing the sentence types, all three sentence types are shown to be processed significantly slower by the L1 group, whereas no difference is found among the other three groups. The Tukey honest significant difference (HSD) post-hoc test (multiple comparisons test) reveals that the L1 group differs significantly from the L2 group (p < 0.000) and the L3+ group (p < 0.000) in all three sentence types.

Table 2.

Mean values of reading times in the interest areas analysed (standard errors in parenthesis).

Groups	n	SOV	SVO	SVO–ki
L1 group	25	897.65 (68.56)	901.96 (65.83)	932.69 (70.99)
L2 group	47	547.69 (30.88)	550.36 (31.44)	548.84 (32.93)
L3 group	19	684.94 (45.24)	713.92 (52.48)	693.26 (49.84)
L3+ group	26	561.22 (36.11)	609.23 (46.89)	693.26 (49.84)
All group (total)	117	647.76 (25.29)	665.13 (26.06)	661.63 (27.30)

When comparing the three sentence types by collapsing the four groups as one, an analysis of variance (ANOVA) shows that the effect of RTs between groups is significant for all sentence types (Table 3). There is a significant difference between the L1 group and the L3 group in SVO–ki sentences (p = 0.015) and canonical SOV sentences (p = 0.021), whereas no difference is found between the L1 group and the L3 group in scrambled SVO sentences (p = 0.070). This might be due to sample insufficiency.

Table 3.

Analysis of variance of the reading times of the three sentence types between and within groups.

Source	df	SS	MS	F	p
SVO–scrambled sentences
Between groups	3	2,147,797	715,932	11.43	0.000
Within groups	113	7,073,886	626,00
Total	116	9,221,683
SOV canonical WO
Between groups	3	2,252,713	750,904	13.19	0.000
Within groups	113	6,431,693	56,917
Total	116	8,684,407
SVO–ki sentences
Between groups	3	2,619,592	873,197	13.15	0.000
Within groups	113	7,501,833	66,387
Total	116	10,121,425

When the mathematical differences between the mean RTs for all sentence types are analysed, the split-plot ANOVA reveals a significant difference (Table 4). Tukey HSD post-hoc analysis indicates that the mathematical difference (–30.73 ms) between the mean RTs for SVO and SVO–ki sentences in the L1 group differs significantly from the mathematical difference (+27.49 ms) between the mean RTs for SVO and SVO–ki sentences in the L3+ group.⁵ That is, RTs for the SVO–ki sentences are longer than the RTs for the SVO sentences in the L1 group, whereas in the L3+ group, RTs of the SVO sentences are longer than those for the SVO–ki sentences.

Table 4.

Analysis of variance of the mathematical differences of the mean reading times of all sentence types.

Source	df	SS	MS	F	p
Group	3	6,982,785	2,327,595	12.89	0.000
Condition	2	25,459	12,729	4.68	0.010
Group × Condition	6	37,319	6219	2.28	0.037
Error	226	614,462	2718
Total	237	7,660,026

Furthermore, we compared the mean RTs of the three sentence types of all the participants (Table 3). The repeated measures ANOVA reveals a significant difference: F(2,232) = 3.512, p = 0.031. In order to examine this difference, a LSD post-hoc test was carried out. The results reveal that SVO and SVO–ki sentences were processed more slowly than the canonical SOV sentences (p = 0.030 for the interaction between SOV and SVO, p = 0.032 for the interaction between SOV and SVO–ki sentences). There is no statistical difference between the RTs for SVO and SVO–ki sentences.

Discussion

The main findings can be summarized as follows.

When comparing the four groups, the results reveal that all three sentence types are processed significantly more slowly by the monolingual participants, but there are no significant differences in RTs among the multilingual groups: L1 > L2 = L3+.

When comparing the three sentence types, the results reveal that the canonical SOV sentences are processed significantly faster by all participants than the SVO and SVO–ki sentences. There is no statistical difference between the RTs for SVO sentences and SVO–ki sentences: canonical SOV sentences < SVO sentences = SVO–ki sentences.

When comparing the monolingual and multilingual (L3+) groups, the results for the SVO–ki sentences reveal that in the monolingual group, RTs are longer for the SVO–ki sentences than for the SVO sentences. On the other hand, the RTs for the SVO–ki sentences are significantly shorter in the multilingual group than in the monolingual group. L1: SVO sentences < SVO–ki sentences; L3+: SVO–ki sentences < SVO sentences.

One of the main findings of this study is that all three sentence types are processed significantly more slowly by the monolingual group than by the bi- and multilingual groups. Although sentence processing of the native language might appear effortless, these differences suggest that monolinguals have a higher processing load and more processing effort when they are processing the structure of the sentences. The difference in processing load and effort might be explained in terms of WM, since in the WM system information must be sustained while manipulations are performed on that information. Moreover, the judgement of the semantic and structural relations among the words being processed requires effortful attention while building the structure of the sentence and interpreting its meaning. On the other hand, the frequent processing of diverse types and structures of sentences in different languages presumably leads to an experience in WO processing. Kroll and Bialystok (2013) claim that some types of pervasive experience could have an impact on human brain structure and function and that bilingualism is one such experience. Based on this hypothesis, we assume that the processing of different sentence types of all the languages a bi- or multilingual person knows can be counted as such a pervasive experience improving the WM system during syntactic structure building.

Processing SVO sentences take longer compared to the other sentence types if all participants are collapsed as one group (Aydın & Cedden, 2010). In Turkish rightward scrambled sentences, the gap precedes the filler. Therefore, the RT differences between the experimental sentences may be explained using a “gap-driven strategy” rather than a filler-driven strategy. Considering the gap-driven strategy, the question arises as to how the parser identifies the gap. Since the gap has no phonological content, other indicators are needed to determine a potential gap position. In cases where the filler follows the gap, the gap position is the internal argument position in sentences where the object has moved to the post-verbal position. The indicator for the parser mentioned above is the main verb, which is evidence for assigning the theta-role to the internal argument. In short, the Theta Criterion triggers the process in which the parser will actively look for a filler position. So, the parser prefers to adopt that potential DP without waiting to see whether there are other or better alternatives later on in the sentence (see example (9)). Constituting a true syntactic chain would only be possible if the parser recognizes the first DP. This recognition will occur in the head position of the Determiner Phrase (DP), as in (9). Only then will the requirement of a semantic or referential interpretation in the gap be satisfied. Given that both the search for the filler and the integration take time, this might be the cause for the increase in RTs.

As indicated above, the participants in all groups tend to read SVO–ki sentences more slowly than the canonical SOV sentences, while there was no statistical difference between SVO and of SVO–ki sentences, as long as all participants are taken together. Contrary to SVO sentences, the sentential object is base-generated in SVO–ki sentences in the sentence-final position, that is, there is no rightward movement operation in SVO–ki sentences. The explanation of the fact that RTs of SVO–ki sentences are longer compared to those of canonical SOV sentences might be that only a very limited number of verbs select ki–clauses in Turkish (compare (10) and (11)). In other words, the low frequency of occurrence of SVO–ki sentences causes slower RTs of SVO–ki sentences in Turkish. As seen in (10) and (11), the root verb um- (hope) selects both nominalized clauses and ki–clauses (compare (10b) and (11b)), while the root verb öner- (suggest) selects only nominalized clauses (compare (10a) and (11a)).

(10) SVO canonical WO

a. Siz [yarın akşam kaç-ma-mız-ı] öner-miş-ti-niz

you tomorrow evening escape-inf-1pl-acc suggest-perf-past-2pl

‘You suggested us to escape tomorrow evening’

b. Biz [ mantıklı davran-ma-nız-ı] um-uyor-uz

we rationally behave-inf-2pl-acc hope-pres-1pl

‘We hope that you behave rationally’

(11) SVO–ki sentences

a. *?Siz öner-miş-ti-niz ki [yarın akşam kaç-alım]

you suggest-perf-past-2pl that tomorrow evening escape-opt.1pl

‘You suggested that we’ll escape tomorrow evening’

b. Biz um-uyor-uz ki [mantıklı davran-ır-sınız]

we hope-pres-1pl that rationally behave-aor-2pl

‘We hope that you behave rationally’

As the data show, the participants in the L1 group tend to read the SVO sentences faster than the SVO–ki sentence. On the other hand, the L3 group reads the SVO–ki sentences faster than the SVO sentences. The reason for the shorter RTs for the SVO–ki sentences in the multilingual group is probably that SVO–ki sentences are similar to the structure of embedded sentences in their L2 or L3 (12). These results may indicate that multilinguals have an advantage over monolinguals in processing particular sentence types.

(12) a. Turkish

Biz um-uyor-uz ki [ mantıklı davran-ır-sınız]

we hope-pres-1pl that logical behave-aor-2pl

b. English

We hope that [you behave rationally]

c. German

Wir hoffen, dass [Sie sich rationell verhalten]

we hope-pres-1pl that you yourself rational behave

Conclusion

This study has examined whether there is a difference in sentence processing between mono- and multilinguals when processing sentences in their L1. The results suggest that non-native languages have a positive effect on processing the WO in the L1, as it leads to faster processing in all the three sentence types explored in our study. We therefore conclude that additional languages represented in the mind have an improving effect on the WM system, which seems to enable the reader to rapidly judge over diverse sentential structures in different languages in order to relate structure and meaning. We assume that due to the constant processing of miscellaneous structural relations among lexical items in different languages, the WM system works in a more effective way when it comes to WO processing in the native language. This assumption supports the view “that bilingual minds are different not because bilingualism itself creates advantages or disadvantages, but because bilinguals recruit mental resources differently from monolinguals” (Kroll & Bialystok, 2013, p. 498).

Furthermore, the results of this study indicate that in Turkish rightward scrambled sentences are more difficult to process. The longer RTs for these sentences can be interpreted as reflecting the search for the filler of the gap left behind after the movement of the object, which is a step needed to arrive at an integrated semantic representation. In this case, the memory span might be taxed, which can be the reason for why rightward scrambled sentences are processed slower than the canonical SOV sentences.

Finally, the finding that SVO–ki sentences tend to trigger longer RTs than canonical SOV sentences might be due to the low frequency of occurrence of SVO–ki sentences in everyday Turkish.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author biographies

Gülay Cedden is an associate professor working as a linguist at the Middle East Technical University, at the Department of Foreign Language Education. Her areas of interest include aspects of psycholinguistics (multi-lingualism, mental lexicon, WO processing, WM).

Özgür Aydın received his MA and PhD degree in the Turkish Teaching Department from Ankara University. He is currently a professor of linguistics at Ankara University. He has also worked at the Department of Turkish Teaching at Duisburg-Essen University, the Department of Special Education at Ankara University and the Departments of Language and Speech Therapy at Anadolu University and Hacettepe University. His research interests include syntax, psycholinguistics and clinical linguistics.

References

Aydın

Cedden

. (2010). Sözdizim işlemlemesinde sağa taşıma etkisi. Dilbilim Araştırmaları, 1, 1–13.

Baddeley

(2003). Working memory and language: An overview. Journal of Communication Disorders, 36, 189–208.

Baddeley

A. D.

(1986). Working memory. Oxford, UK: Clarendon Press.

Bialystok

(2009). Bilingualism: The good, the bad, and the indifferent. Bilingualism: Language and Cognition, 12(1), 3–11.

Caplan

Waters

(2013). Memory mechanisms supporting syntactic comprehension. Psychon Bull Rev, 20, 243–268.

Cedden

Aydın

. (2008). Is there a difference between multilinguals’ and monolinguals’ word order processing in their L1? In Arnándiz

O. M.

Jordà

ve M. P. S.

(Eds.), Achieving multilingualism: wills and ways, Proceedings of the 1st International Conference on Multilingualism (ICOM) (pp. 403–421). Castelló de la Plana: Publicacions de la Universitat Jaume.

Daneman

Carpenter

(1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450–466.

Erdocia

Laka

Mestres-Missé

Rodriguez-Fornells

(2009). Syntactic complexity and ambiguity resolution in a free word order language: Behavioral and electrophysiological evidences from Basque. Brain and Language, 109, 1–17.

Fiebach

C. J.

Schlesewsky

Lohmann

von Cramon

D. Y.

Friederici

A. D.

(2005). Revisiting the role of Broca’s area in sentence processing: Syntactic integration versus syntactic working memory. Human Brain Mapping, 24, 79–91.

10.

Flege

J. E.

Frieda

E. M.

(1997). Amount of native-language (L1) use affects the pronunciation of an L2. Journal of Phonetics, 25, 169–186.

11.

Frazier

(1987). Syntactic processing: Evidence from Dutch. Natural Language & Linguistic Theory, 5(4), 519–559.

12.

Frazier

Clifton

C. J.

(1989). Successive cyclicity in the grammar and the parser. Language and Cognitive Processes, 4(2), 93–126.

13.

Gibson

(1998). Linguistic complexity: Locality of syntactic dependencies. Cognition, 68, 1–76.

14.

Kaan

Harris

Gibson

Holcomb

(2000). The P600 as an index of syntactic integration difficulty. Language and Cognitive Processes, 15, 159–201.

15.

Kroll

J. F.

Bialystok

(2013). Understanding the consequences of bilingualism for language processing and cognition. Journal of Cognitive Psychology, 25, 497–514.

16.

Lin

C. C.

(2006). Grammar and parsing: A typological investigation of relative-clause processing (Unpublished doctoral dissertation). University of Arizona, USA.

17.

Liszka

S. A.

(2004). Exploring the effects of first language influence on second language pragmatic processes from a syntactic deficit perspective. Second Language Research, 20, 212–231.

18.

Mazuka

Itoh

Kondo

(2002). Costs of scrambling in Japanese sentence processing. In Nakayama

(Ed.), Sentence processing in East Asian languages (pp. 131–166). Stanford, CA: CSLI Publications.

19.

McElree

(2000). Sentence comprehension is mediated by content-addressable memory structures. Journal of Psycholinguistic Research, 29, 111–123.

20.

Miyamoto

E. T.

Takahashi

(2002). Sources of difficulty in the processing of scrambling in Japanese. In Nakayama

(Ed.), Sentence processing in East Asian languages (pp. 167–188). Stanford, CA: CSLI Publications.

21.

Miyamoto

E. T.

Takahashi

(2004). Filler-gap dependencies in the processing of scrambling in Japanese. Language and Linguistics, 5(1), 153–166.

22.

Nakano

Fesler

Clahsen

(2002). Antecedent priming at trace positions in Japanese long distance scrambling. Journal of Psycholinguistic Research, 31, 531–571.

23.

Nakayama

(1995). Scrambling and probe recognition. In Mazuka

Nagai

(Eds.), Japanese sentence processing (pp. 257–273). Hillsdale, NJ: Erlbaum.

24.

Nicenboim

Vasishth

Gattei

Sigman

Kliegl

(2015). Working memory differences in long-distance dependency resolution. Frontiers in Psychology, 6, 312.

25.

(2008). An active gap strategy in the processing of filler-gap dependencies in Chinese. In Chan

M. K. M.

Kang

(Eds.), Proceedings of the 20th North American Conference on Chinese Linguistics (Vol. 2, pp. 943–957). Columbus: The Ohio State University.

26.

Schneider

Indefrey

(2007). L1 Word order influences in L2 sentence processing. Talk presented at PiF – Psycholinguistics in Flanders, 28.-29.6.2007, Antwerp, Belgium.

27.

Sekerina

I. A.

(2003). Scrambling and processing: Dependencies, complexity, and constraints. In Karimi

(Ed.), Word order and scrambling (pp. 301–324). Malden, MA: Blackwell Publishers.

28.

Song

Andrews

(2009). The L1 in L2 Learning - Teachers’ Beliefs and Practices. Lincom Studies in Language Acquisition 24. Muenchen: LINCOM Europa.

29.

Stojanović

(1999). Parsing and acquisition: Evidence from Serbo-Croatian (Unpublished doctoral dissertation). University of Ottawa, Canada.

30.

Vikner

Schwartz

B. D.

(1996). The verb always leaves IP in V2 clauses. In Belletti

Rizzi

(Eds.), Parameters and functional heads (pp. 11–62). New York: Oxford University Press.

31.

Veld

(1998). Postverbal clausal constituents. In Johanson

(Ed.), The Mainz meeting: Proceedings of Seventh International Conference on Turkish Linguistics (pp. 183–196). Wiesbaden: Harrossowitz.

32.

Waters

G. S.

Caplan

(2004). Verbal working memory and on-line syntactic processing: Evidence from self-paced listening. The Quarterly Journal of Experimental Psychology, 57A(1), 129–163.

33.

Weyerts

Penke

Münte

Heinze

H. J.

Clahsen

(2002). Word order in sentence processing: An experimental study on verb placement in German. Journal of Psycholinguistic Research, 31, 211–268.

34.

Yamashita

(1997). The effects of word order and case marking information on the processing of Japanese. Journal of Psycholinguistic Research, 26, 163–188.