A Meta-analysis of Cross-linguistic Syntactic Priming Effects

XIANGYU JIANGLIANG CHEN

University of Georgia, USA/Harbin Institute of Technology at Weihai, ChinaUniversity of Georgia, USA

AMeta-analysisofCross-linguisticSyntacticPrimingEffects

XIANGYU JIANGLIANG CHEN

University of Georgia, USA/Harbin Institute of Technology at Weihai, ChinaUniversity of Georgia, USA

Cross-linguistic syntactic priming (CLSP) refers to the phenomenon that the use of a particular structure (e.g., passives) in one language facilitates or primes the subsequent use of the parallel structure in the other language of a bilingual speaker. Individual studies examining CLSP effects have produced conflicting results. This study aims to estimate the population effects of cross-linguistic syntactic priming effects and lexical boost effect, and to examine how potential moderators (priming direction, L2 learners’ proficiency levels, and structural similarities) influenced cross-linguistic priming, electronic research databases were searched systematically for published studies, which measured CLSP effects. Meta-analytic techniques were used to analyze 20 L1 to L2 cross-linguistic studies with 32 effects sizes and 10 L2 to L1 studies with 19 effect sizes. Most studies involved 20-45 minutes of priming activities. Heterogeneity exists within the crosslinguistic priming but the priming direction (L1 to L2 or L2 to L1), L2 learners’ proficiency levels, experimental modes and structural similarities were not potential moderators. For the L1 to L2 studies, the results turned out to be homogeneous. For the L2 to L1 direction, though the results were heterogeneous, L2 learners’ proficiency level, experimental modes, and structural similarities were not moderators. More detailed information should be given for the cross-linguistic priming studies and more refined meta-analysis containing more potential moderators should be carried out in the future.

cross-linguistic syntactic priming, lexical boost effect, meta-analysis, bilingual

INTRODUCTION

Syntactic priming is the tendency for a speaker to reuse the same structural pattern encountered by listening, repeating or perceiving (e.g., Bock, 1986; Bocketal., 2007). In syntactic priming studies, participants listened to and repeated sentences using particular syntactic structures (e.g., transitive structures). Each of these structures could be presented in one of two alternate forms (e.g., active versus passive). Participants were then asked to describe novel pictures portraying relations that could be described using either form. The results showed that hearing and repeating a particular syntactic form increased the likelihood that participants would use that form to describe pictures involving different lexical items. For example, having repeated aloud a sentence such asTherockstarsoldsomecocainetoanundercoveragent, participants were more likely to describe a picture with the same prepositional-object (PO) construction such asThegirlhandedapaintbrushtothemanrather than the alternative double-object (DO) phrase structureThegirlhandedthemanapaintbrush(Bock, 1986). These results were typically interpreted as evidence for the hypothesis that adults or children represent syntactic forms abstractly.

Within the past decade, a large number of studies have examined syntactic priming in bilingual individuals (e.g., Loebell & Bock, 2003; Bernoletetal., 2007; Meijer & Fox Tree, 2003; Hartsuikeretal., 2004). Cross-linguistic syntactic priming refers to the phenomenon that the use of a particular structure (e.g., passives) in one language facilitates or primes the subsequent use of the parallel structure in the other language of a bilingual speaker. Researchers are interested in four related questions. The first and foundational question is whether there is crosslinguistic syntactic priming, that is, whether the use of a particular structure (e.g., passives) in one language primes the subsequent use of the parallel structure in the other language of a bilingual speaker. This question is crucial for us to understand how knowledge of syntax of the two languages is represented in the bilingual population. For example, it has been assumed that if we find crosslinguistic syntactic priming, then there is evidence for shared syntactic representations between the two languages (Hartsuikeretal., 2004). The second question relates to the potential moderators which would influence the strength of priming, if there’s any. The third question concerns the magnitude of effects of L1-to-L2 syntactic priming versus L2-to-L1 syntactic priming. Finally, there is a question regarding the so-called lexical boost effect, which refers to the observation that syntactic priming occur with structural repetition alone (i.e., in the absence of lexical repetition) but the effect is boosted when the head is repeated in the form of translation equivalence (see e.g., Segaertetal., 2013). The goal of the present study was to investigate these questions through a systematic and meta-analytic examination.

METHODS

The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) (Moher, 2009) statement guided the reporting of this meta-analysis.

Selectioncriteria

To capture all relevant studies on syntactic priming in bilinguals, studies were selected based on the following criteria: (a) Participants were healthy with no history of neurological, cognitive, language, speech, or hearing impairment; (b) participants were randomly selected; (c) studies included information on effect size or sufficient information from which to calculate effect size; (d) studies included information on the duration of priming activities which lasted no longer than 45 minutes; (e) studies tested production rather than comprehension of target syntactic structures; (f) studies included information on participants’ native languages; and (g) studies involved bilingual or crosslinguistic priming. Due to resource limitations, only those studies published in Chinese, English, or translated into English, were included.

Locationandselectionofstudies

We started our search with the search strings “syntactic priming OR structural persistence OR structural priming OR syntactic persistence AND bilingual OR crosslinguistic” in the following electronic databases to locate appropriate studies: Google Scholar, Academic Search Complete, ERIC; JSTOR, Linguistics & Language Behavior Abstracts, PsycINFO, PsycINFO, Web of Science, and CNKI. A manual search of the reference lists of several most recent studies on syntactic priming (e.g., Pickeringetal., 2013; Ruf, 2011) was also subsequently conducted. The search was conducted with no back-dated limitation until the first of November 2013.

A total of 68 articles were obtained from the search procedure. Two researchers reviewed the titles, abstracts, and keywords of these articles and the dissertation for possible inclusion by applying the selection criteria mentioned above. The full texts of the articles were consulted when the abstract did not provide sufficient information to determine inclusion or exclusion. Differences were resolved by agreement between the two researchers. Forty eight articles were excluded for the following reasons: different dependent variables tested (n=28), lack of adequate data to calculate the effect size (n=7), unusual data collection method (n=3), relevant review articles (n=8), unusual syntactic structure applied (n=1), and dissertation (n=1). No additional research articles were found by hand searching the selected references. This resulted in a total of 20 studies for inclusion in the present study.

Studycharacteristics

The 20 studies included in the meta-analysis yielded 32 L1 to L2 crosslinguistic priming effect sizes, and 19 L2 to L1 crosslinguistic priming effect sizes. Mean age of participants ranged from 3.18 to 56 years out of 20 studies, and most of them recruited college student participants for the sake of convenience.

Calculationofeffectsize

Priming effect sizes (ES) were calculated by using the following example. ES of passive priming=mean proportions of passive responses primed by passives—mean proportions of passive responses primed by actives. For the lexical boost effect, it was obtained by subtracting priming effects using different target words or different translation equivalents from priming effects using same target words or translation equivalents.

Selectionandcodingofmoderators

Potential effect size moderators for priming effects were selected and categorized as participant variables (i.e., L2 learners’ proficiency levels) and features of the research design (i.e., priming condition, experimental mode, and structural similarity).

Participantvariable

The participant variable that would influence priming effects is the participants’ L2 proficiency levels, because the study of Bernoletetal. (2013) showed stronger priming effects had been found from the more proficient bilinguals. In the present study, we classified advanced and medium level learners as one group, and bilinguals who were native in both languages as another group.

Featuresoftheresearchdesign

The research design variables that were expected to influence the size of priming effects were priming condition, experimental mode, and structural similarity applied in the study. Even though no crosslinguistic priming was found, Loebell and Bock (2003) showed that a stronger priming existed effect from L1 to L2 than from L2 to L1. Thus with respect to the priming condition, we separated crosslinguistic priming into L1 to L2 priming, and L2 to L1 priming. The contrasted weights are given in Table 1.

In the priming activities, there are mainly two kinds of responses—speaking or writing responses; however, both of them can be regarded as participants’ L2 productive knowledge. In terms of the input, the participants involved in the 20 studies either repeated/recalled or just listened to the experimenters’ descriptions. By repetition or recalling, the input of priming activities can be considered production as well. In contrast, the input by listening to the experimenters’ descriptions is comprehension. It is likely that the priming effect caused by production would be stronger than comprehension, since during the process of production, participants have strengthened the input information. Nevertheless, if comprehension is adequate enough for priming to occur, then the priming effects caused by comprehension and production would not differ much from each other. As a consequence, the studies were coded into two categories according to the experimental modes: comprehension to production and production to production.

Finally, whether the tested structure is similar or not in two languages may associate with the occurrence of crosslinguistic priming. Similar structures may lead to the occurrence of crosslinguistic priming (e.g., Hartsuikeretal., 2004); dissimilar structures may lead to the conflicting results (e.g., Loebell & Bock, 2003). Twenty studies were specified according to structural similarity.

Statisticalanalysis

The mean effect size and the 95% confidence interval (CI) (weighted by their own sample sizes) were computed overall for both priming and word repetition effects by using a random effect model. Heterogeneity of mean effects was tested by using the I2statistic, which indicates the percentage of variance between effects that cannot be explained by sampling errors. Conventionally speaking, I2values of 25%, 50%, and 75% indicate low, moderate, and high heterogeneity between studies respectively. The weighted fail-safe sample (N+) was calculated in order to determine how many effect sizes of zero were still needed to reduce the overall effect size to zero.

Macros in SPSS were used to calculate the overall mean priming effect sizes and to conduct univariate analysis of the four moderators which would influence the magnitude of effect size. Planned contrasts were also carried out and analyzed by MetaReg under each level. For the lexical boost effect sizes, Meta mean had been conducted but not the univariate analysis because they were homogeneous.

RESULTS

Primingandlexicalboosteffects

The mean priming effect size across all 51 effect sizes was 0.09 (95%CI=0.02, 0.16), and the mean effect was moderately heterogeneous across studies,I2=59.45%, which indicates the crosslinguistic priming effect exists and the population variance accounted for 59.45% of the weighted observed variance among effects. The sample size of a single study of no effect needs to be 166 to reduce the significance level to 0.05. The lexical boost mean effect size across all relevant studies was 0.10 (95%CI=0.01, 0.18), and the mean effect was homogeneous across relevant studies.

Moderatoranalysisforprimingeffects

A summary of the univariate priming effect moderator results is presented in Table 1. The means for most levels of each moderator were homogeneous (95%CIfor I2 included zero), which included priming condition from L1 to L2, advanced and medium learners, dissimilar structures, and production to production priming mode. The remaining levels of each moderator were heterogeneous, for priming condition from L2 to L1,Qw(18)=77.65,p<0.01; native bilinguals,Qw(18)=109.19,p<0.01; similar structural similarity,Qw(33)=109.95,p<0.01; for comprehension to production priming mode,Qw(29)=100.45,p<0.01. Priming effect size was unrelated to all 4 potential moderators and correspondingly 2-way or 3-way interactions were not allowed to be conducted. Moreover, for the priming condition moderator, thepvalue forQBwas larger than 0.05, which means L1 to L2 and L2 to L1 crosslinguistic priming were not significantly different from each other.

Within the L1 to L2 priming condition, their effect sizes turned out to be homogeneous (QB(31)=44.43,p=0.06), thus no univariate analysis was carried out. Within the L2 to L1 priming condition, the result illustrated heterogeneity (QB(18)=77.65,p<0.01). However, structural similarity, L2 learners’ proficiency levels, and experimental modes all cannot explain the heterogeneity, and thus no 2-way or 3-way interaction could be conducted as well.

Table 1 Univariate results for priming moderator variables

*pvalue reported is for Q between.

DISCUSSION

The present meta-analysis synthesized data from 20 studies to examine several important questions in syntactic priming in bilingual populations. The results allowed us to answer all four questions put forward in the introduction part. The first question relates to whether the crosslinguistic priming effect exists. Our results demonstrated the existence of crosslinguistic priming effect, with the mean effect size as 0.09. The second question helps us to figure out the potential moderators in the crosslinguistic priming tests. The results showed that all four potential moderators found in the previous research, namely, L2 learners’ proficiency levels, priming condition, structural similarity and experimental modes, cannot explain the heterogeneity in the population. Even when we analyzed within each priming condition (from L1 to L2, and from L2 to L1), no significant results were found. In terms of the third question, the relative strength between L1 to L2 and L2 to L1 syntactic priming, the results showed that the priming effect from L1 to L2 was same as that from L2 to L1.

Finally, in regards to the lexical boost effect in crosslinguistic priming conditions, the repetition of the head or its translation equivalent was found to result in a stronger priming effect. However, the studies reviewed did not differentiate the effect of priming for cognates (i.e., same translations with similar pronunciations) from that for non-cognates. Given that non-cognates share fewer representations at the conceptual level than do cognates (de Groot, 1992), we may expect a differentiated effect of syntactic priming with these two types of translation equivalents. Clearly, further research is required to clarify this critical issue.

The meta-analysis reveals an obvious need for further research in syntactic priming in bilinguals that can better address other questions. First, there were inadequate data across the full range of possible stimuli for us to investigate the role of structural similarity between the two languages in syntactic priming in bilingual speakers. Some of the cross-linguistic studies compared the word order of primed and target sentences, while others did not.

The second area in need of further research is to identify which aspects of bilingual experience seem more important for dictating the manner and extent of syntactic priming in bilingual speakers. For example, a topic for future study is to systematically examine the effect of the relative proficiency of the two languages of the bilingual speaker on the magnitude of syntactic priming. Still in the cross-linguistic priming conditions, for the unbalanced L2 learners, more information about their proficiency levels should be reported, because L2 learners’ proficiency level could predict the magnitude of priming effects. For example, Shin and Christianson (2009) investigated whether structural priming would occur from English (L2) to Korean (L1) with dative structures. Their results showed that cross-linguistic structural priming was limited to proficient L2 speakers. Moreover, while reporting L2 learners’ proficiency levels, standardized scores should be given to classify L2 learners into different proficiency groups.

Another limitation of the literature analyzed is that it has not reported the typology of languages, which could act as a potential moderator. For example, both English and German belong to the Germanic language family, and due to the intensity of language contact, we would expect stronger priming effects between English and German than those between English and Chinese, two languages from different language families. This has been supported by Caietal.’ study (2011), in which the cross-linguistic priming effects between Cantonese and Chinese got the largest priming effect sizes across all studies. Thus, future studies could also report the typology of languages involved in the priming activities to verify whether it is a potential moderator.

Finally, due to the lack of information regarding participants’ knowledge of languages, we could not compare within-language syntactic priming with crosslinguistic priming. According to the shared-syntax account, rules that are the same in the two languages are represented once; whereas the separate-syntax account suggests that there are language-specific stores for a syntactic structure that appear superficially similar in the two languages of the bilingual speaker (Hartsuikeretal., 2004). The shared syntax account would predict that syntactic priming effects when bilinguals switch between languages are just as large as when they produce consecutive utterances in the same language, that is, the crosslinguistic priming effect should be as strong as the within-language priming effect.

Since the majority of studies in this analysis recruited college students who underwent priming activities for 25-40 minutes, these findings should not be generalized to the entire population; thus, additional studies with a variety of groups performing a wider range of priming activities need to be conducted.

CONCLUSION

The present meta-analysis is the first study to quantify and summarize the cross-linguistic priming effects in bilingual speakers. While the study has provided great insight into the lexical boost effect and the factors influencing crosslinguistic priming in bilingual speakers, it remains to be discovered how the features of the relevant syntactic structures and the relative proficiency of the two languages of the bilingual speakers may modulate the effect of syntactic priming. Much larger random samples of participants need to be included to better estimate the population effects of syntactic priming.

marked with an asterisk indicate studies included in the meta-analysis.]

*Bernolet, S., Hartsuiker, R. J., & Pickering, M. J. (2007). Shared syntactic representations in bilinguals: evidence for the role of word-order repetition.JournalofExperimentalPsychology:learning,Memory,andCognition, 33(5), 931-949.

*Bernolet, S., Hartsuiker, R. J., & Pickering, M. J. (2012). Effects of phonological feedback on the selection of syntax: Evidence from between-language syntactic priming.Bilingualism:LanguageandCognition, 15(3), 503-516.

*Bernolet, S., Hartsuiker, R. J., & Pickering, M. J. (2013). From language-specific to shared syntactic representations: The influence of second language proficiency on syntactic sharing in bilinguals.Cognition, 127(3), 287-306.Bock, K. (1986). Syntactic persistence in language production.CognitivePsychology, 18, 355-387.Bock, K., Dell, G. S., Chang, F., & Onishi, K. H. (2007). Persistent structural priming from language comprehension to language production.Cognition, 104(3), 437-458.

*Cai, Z. G., Pickering, M. J., Yan, H., & Branigan, H. P. (2011). Lexical and syntactic representations in closely related languages: Evidence from Cantonese-mandarin bilinguals.JournalofMemoryandLanguage, 65(4), 431-445.

*Chen, B., Jia, Y., Wang, Z., Dunlap, S., & Shin, J. -A. (2013). Is word-order similarity necessary for cross-linguistic structural priming?.SecondLanguageResearch, 29(4), 375-389.de Groot, A.M.B. (1992). Determinants of word translation.JournalofExperimentalPsychology:Learning.MemoryandCognition, 18, 1001-1018.

*Desmet, T. & Declercq, M. (2006). Cross-linguistic priming of syntactic hierarchical configuration information.JournalofMemoryandLanguage, 54(4), 610-632.

*Hartsuiker, R. J., Pickering, M. J., & Veltkamp, E. (2004). Is syntax separate or shared between languages?.PsychologicalScience(Wiley-Blackwell), 15(6), 409-414.

*Hung, Chung-ju. Cross-linguistic syntactic priming effect on EFL learners’ immediate syntactic structure formation.US-Chinaforeignlanguage, 9(10), 634-646.

*Jiang, L. (2012). Cross-linguistic priming of passives and its mechanism.ModernForeignLanguages, 35(1), 54-61.

*Kantola, L. & Gompel, R. (2011). Between- and within-language priming is the same: Evidence for shared bilingual syntactic representations.Memory&Cognition, 39(2), 276-290.

*Lei, L. & Wang, Tongshun. (2009). Bilinguals’ syntactic representation-evidence from Chinese-English unbalanced bilinguals in syntactic priming.Modernforeignlanguages, 32(2), 158-167.

*Loebell, H. & Bock, K. (2003). Structural priming across languages.Linguistics:AnInterdisciplinaryJournaloftheLanguageSciences, 41(5), 791-824.

*Meijer, P. J. A. & Fox Tree, J. E. (2003). Building syntactic structures in speaking: A bilingual exploration.ExperimentalPsychology, 50(3), 184-195.

Moher. (2009). Preferred reporting items for systematic reviews and Meta-Analyses: The PRISMA Statement.PhysicalTherapy, 89(9), 873-880.

Pickering, M. J., McLean, J. F., & Branigan, H. P. (2013). Persistent structural priming and frequency effects during comprehension.JournalofExperimentalPsychology:learning,Memory,andCognition, 39(3), 890-897.

Ruf, H. T. (2011).AnInvestigationofSyntacticPrimingamongGermanSpeakersatVaryingProficiencyLevels. ProQuest LLC.

Segaert, K., Kempen, G., Petersson, K. M., & Hagoort, P. (2013). Syntactic priming and the lexical boost effect during sentence production and sentence comprehension: An fMRI study.BrainandLanguage, 124(2), 174-183.

*Salamoura, A. & Williams, J. N. (2006). Lexical activation of cross-language syntactic priming.Bilingualism:LanguageandCognition, 9(3), 299-307.

*Salamoura, A., & Williams, J. N. (2007). Processing verb argument structure across languages: Evidence for shared representations in the bilingual lexicon.AppliedPsycholinguistics, 28(4), 627-660.

*Schoonbaert, S., Hartsuiker, R. J., & Pickering, M. J. (2007). The representation of lexical and syntactic information in bilinguals: Evidence from syntactic priming.JournalofMemoryandLanguage, 56(2), 153-171.

*Shin, J. -A., & Christianson, K. (2009). Syntactic processing in Korean-English bilingual production: evidence from cross-linguistic structural priming.Cognition, 112(1), 175-180.

*Shin, J. -A. & Christianson, K. (2012). Structural priming and second language learning.LanguageLearning, 62(3), 931-964.

*Vasilyeva, M., Waterfall, H., GMez, P. B., GMez, L. E., Bowers, E., & Shimpi, P. (2010). Cross-linguistic syntactic priming in bilingual children.JournalofChildLanguage, 37(5), 1047-1064.

*Verreyt, N., Bogaerts, L., Cop, U., Bernolet, S., Letter, M. D., Hemelsoet, D., Santens, P., & Duyck, W. (2013). Syntactic priming in bilingual patients with parallel and differential aphasia.Aphasiology, 27(7), 867-887.

10.3969/j.issn.1674-8921.2014.12.002

Correspondence should be addressed to Xianyu Jiang, the University of Georgia, USA. Email: xiangyu@uga.edu