Lexico-syntactic interactions during the processing of temporally ambiguous L2 relative clauses: An eye-tracking study with intermediate and advanced Portuguese-English bilinguals.

Ana Paula Soares,Montserrat Comesaña,Isabel Fraga,Marisa Ferreira,António Filipe Macedo,Helena Oliveira,Juan Hernández-Cabrera,Carlos Acuña-Fariña,Pilar Ferré

doi:10.1371/journal.pone.0216779

Abstract

There is extensive evidence showing that bilinguals activate the lexical and the syntactic representations of both languages in a nonselective way. However, the extent to which the lexical and the syntactic levels of representations interact during second language (L2) sentence processing and how those interactions are modulated by L2 proficiency remain unclear. This paper aimed to directly address these issues by using an online technique (eye-tracking) that is highly sensitive to the lexical and syntactic processes involved in sentence reading. To that purpose, native-speakers of European Portuguese (EP) learning English as L2 at intermediate and advanced levels of proficiency were asked to silently read temporally ambiguous L2 relative clause (RC) sentences disambiguated with a High-Attachment (HA) or Low-Attachment (LA) strategy while their eye-movements were monitored. Since EP and English native speakers differ in the way they process and comprehend this syntactic structure (EP: HA, English: LA), HA preferences were used as a marker of L1 RC syntax interference. Additionally, the cognate status of the complex NP that preceded the RC was manipulated to further analyze how the lexical co-activation of both languages would also affect the syntactic representations of the non-target (L1) language. Results showed cognate facilitation in early and late reading time measures regardless of L2 proficiency, and also that the cognate status of the complex NP impacted L2 reading performance, particularly at lower levels of L2 proficiency. These findings provide compelling evidence for a bilingual reading system that seems to be highly dynamic and interactive not only within each level of processing, but, importantly, across levels of representation. They also suggested that, as the level of L2 proficiency increases, L1 RC syntax interference becomes stronger, in a syntactic parser that seems to operate in a more integrated and nonselective way, with both strategies being equally available to guide L2 reading comprehension. Results are discussed attending to the current models of bilingual syntactic processing.

Highlights

We live in an increasingly globalized world, where using two or more languages to communicate is becoming the rule rather than the exception [1]
Evidence from the processing of relative clauses (RC) preceded by a complex noun phrase (NP) in sentences like “Someone shot the servant of the actress who was on the balcony” ([3]), have shown that native speakers differ considerably in the way they resolve this syntactic ambiguity across languages
The length of the N1, N2, and N3 words in the sentences was equivalent across conditions, we decided to adopt this procedure to reduce error variance by making different items roughly comparable with each other within and across sentence regions

Summary

Introduction

We live in an increasingly globalized world, where using two or more languages to communicate is becoming the rule rather than the exception [1]. While European Portuguese (EP) native speakers are likely to interpret that the servant was on the balcony, since in this language an HA attachment preference is observed [5], English native speakers would be likely to interpret that it was the actress who was on the balcony, since an LA strategy is preferred in English instead (e.g., [3], [6], [7]) Studying this grammatical structure offers an excellent opportunity to analyze how the human mechanism responsible for assigning a grammatical structure to a sentence or phrase (parser) works in situations in which individuals master languages whose principles lead to different syntactic and semantic representations, as EP and English. Since previous works have shown that words that share meaning and form (orthography and/or phonology) across languages (i.e., cognate words, such as actress and actriz in English and EP, respectively) are processed faster and more accurately than noncognate words (i.e., words that share meaning, but not form across languages, such as servant and criada in English and EP, respectively; see [8] and [9] for reviews), embedding these words in the complex NP of the RC structures can provide an excellent opportunity to further explore how the lexical and syntactic levels of representation interact in the bilingual mind during L2 sentence processing, a largely unexplored issue in the bilingual and L2 acquisition literature

Objectives

Methods

Results

Conclusion