Research on the acquisition and processing of nonnative languages is of broad social relevance because the ability to communicate in multiple languages is becoming increasingly valued in today's global economy. In this project, we investigated the interplay between the implicit knowledge learners have of a nonnative language and their processing of that language by examining two types of relative clauses (RCs) in English: gapped RCs (e.g., "the man that they hired") and resumptive RCs (e.g., "the man that they hired him"). Resumptive RCs are considered ungrammatical in English (i.e., they are disallowed by the implicit mental grammar of the native speaker of English), and yet they are not uncommon in nonnative speech/writing, even from learners whose first language (L1) also disallows them. Phenomena of this type represent a particularly interesting area of second language (L2) research because they cannot be straightforwardly traced to either L1 influence or target-language input, which raises questions about why they emerge in the first place.

Drawing on methods honed over the last two decades in the psycholinguistics literature on mature native speakers, we investigated three key questions concerning resumptive RCs in adult L1/L2 English: First, is resumption a licit option for relativization in the implicit mental grammar and/or is it a means of managing cognitive load during real-time RC processing? Second, if resumption does aid in RC processing, are facilitation effects observed only in language production or also in language comprehension? Finally, for L2 learners (L2ers), do performance patterns vary as a function of English proficiency?

Our participants -- English native-speaker controls, L1-Korean L2ers of English, and L1-Mandarin L2ers of English -- completed three main tasks: an oral elicited production task (EPT) probing processing during RC production, a self-paced reading task (SPRT) probing processing during RC comprehension, and an acceptability judgment task (AJT) testing the offline acceptability of the sentence types in the other tasks. Each task had three conditions targeting increasingly complex RC environments (short-distance, long-distance, and "WH-island"). L2ers undertook two versions of the AJT -- one in English, one in their L1 -- so that L1 transfer effects, if present, could be identified; note that of English, Korean, and Mandarin, only Mandarin allows resumption in any of the RC environments tested. Also administered to all participants were a language background questionnaire, an English proficiency C-test, and a short exit survey.

The L1/L2 results from the EPT and the SPRT provide clear-cut evidence that resumption eases both production and comprehension of RCs under processing strain; moreover, strong facilitation effects for resumption held even when consistent acceptors of English resumptive RCs were removed from analysis. For the adult L1 processing literature, our findings contest widespread claims (e.g., Ferreria & Swets, 2005; Heestand et al., 2011; Morgan et al., 2020; Polinsky et al., 2013) that resumption helps only with RC production. For the L2 literature, our finding that resumption is in many cases a purely processing-based phenomenon casts doubt on a traditional assumption (e.g., Algady, 2013; Eckman, 2004; Hyltenstam, 1984; Marefat & Abdollahnejad, 2014; Pavesi, 1986), namely that L2ers' performance patterns always reflect L2 representations licensed by their implicit mental grammar.

While a sizable portion of the L2 participants consistently accepted English resumptive RCs, it also emerged that L2ers with higher English proficiency scores were more likely to reject resumptive RCs, at least in environments where English allows only gapped RCs. This result indicates that L2ers can become sensitive to the ungrammaticality of resumption in English RCs, and, moreover, that this sensitivity develops fastest where processing is easiest (i.e., short- and long-distance RCs).

We hope this project will inspire other L2 researchers to leverage psycholinguistic methods to explore issues at the intersection of implicit linguistic knowledge and real-time language processing. The present findings could also be of use in second/foreign English-language instruction because they provide insights into why resumptive RCs sometimes crop up in L2ers' speech/writing. Specifically, L2 English resumption is often a processing artifact rather than a sign of nontarget RC representations. In working towards a deeper understanding of what gives rise to conspicuous features of nonnative spoken/written output, we hope our research will help reduce the all-too-frequent stigma associated with them. Research of the kind embodied in our project could in time lead to improved teaching techniques as well as to advancements in areas such as machine translation and human-computer interaction.

Finally, by collecting data over the internet, we were able to sample from a larger and more diverse participant pool than in a traditional laboratory setting. In this way, web-based experimentation can increase the likelihood that research data accurately represent the populations of interest. Because L2 acquisition research can influence pedagogical practices, language policy, and popular opinion toward nonnative speakers, web-based experimentation has potential benefits not only for researchers but also for society at large.

Last Modified: 05/27/2024
Modified by: Bonnie D Schwartz