Collaborative dialogue and L2 development

Although collaborative dialogue has been suggested by sociocultural SLA

researchers as the appropriate unit of analysis for the investigation of learner interaction, it serves to explain the learning of knowledge in general rather than L2 alone. As a result, in SLA research, collaborative dialogue was often examined in relation to LREs, which

Swain (2001b) described as “an instance of collaborative dialogue” (p. 286) and more specifically “any part of a dialogue where students talk about the language they are

producing, question their language use, or other- or self-correct their language production” (ibid). Leeser (2004) elaborated on the nature of LREs by specifying what they entailed, “(a) question the meaning of a linguistic item; (b) question the correctness of the

spelling/pronunciation of a word; (c) question the correctness of a grammatical form; or (d) implicitly or explicitly correct their own or another’s usage of a word, form or structure” (p. 56). Among these many types of LREs, explicit discussions of word

meaning or vocabulary-focused LREs and grammar issues or form-focused LREs seemed to be the major categories of LREs. For example, Swain and Lapkin’s (2002)

investigation of the talk of a pair of French immersion students collaborating on a jigsaw task suggested that vocabulary-focused and form-focused LREs accounted for eighty percent of the overall LREs they produced. Malmqvist (2005) likewise found that the LREs generated by Swedish learners of German in completing dictogloss tasks were mainly related to vocabulary (58%) and grammar (42%). Findings of similar studies (e.g., Kim and McDonough, 2008, 2011; Leeser, 2004; Lowen, 2003, 2004; Williams, 1999, 2001) confirmed the high prevalence of vocabulary-focused and form-focused LREs in collaborative dialogue, indicating learners’ concern for their lexical choices and

grammatical accuracy during their dyadic interaction.

As previously mentioned, collaborative dialogue is viewed as the site in which L2 learning occurs and the basis for the internalization of co-constructed linguistic

knowledge. Therefore, in studies concerning LREs, tailor-made posttests appeared to be the most commonly adopted approach for the measurement of SLA. Tailor-made

posttests in general involved a detailed analysis of learners’ collaborative dialogue, especially the vocabulary and form-focused LREs that emerged in their pair work, along with the creation of test questions targeted at assessing their retention of the lexical items and grammatical forms discussed in these LREs. Given that LREs varied across dyads, tailor-made posttests were usually pair-specific and consisted of discrete rather than integrative item types in order to “measure the learning of the exact aspect of language about which students has metatalked” (Swain, 1998, p. 76). A glimpse of learners’ performance on the tailor-made posttests indicated their success in retaining the L2 features they had collaborated on. In particular Swain (1998) found that the correct solutions that learners reached in their LREs on forming feminine adjectives from masculine ones in French tended to be the accurate answers they provided to the tailor- made dyad-specific posttest questions. Williams (2001) subsequently revealed that learners achieved between 40% to 94% accuracy on the posttest items that were created based on their successfully resolved LREs. Other more recent studies (McDonough & Sunitham, 2009; Kim, 2008; Tocalli-Beller & Swain, 2007; Watanabe & Swain, 2007; Zeng & Takatsuka, 2009) in a similar vein suggested that learners were able to convert the L2 lexical and grammatical knowledge they correctly co-constructed into their accurate performance on the tailor-made posttests. Despite Williams’ (2001) caution that learners’ accurate answers to the test items were distant from the integration of the linguistic structures discussed in the LREs into their interlanguage, the strong link between correctly resolved LREs and the satisfactory scores that learners gained in the posttests to some extent indicated the positive impact that collaborative dialogue has on L2 development.

There are, however, caveats regarding the effects of collaborative dialogue on SLA. First of all, learners’ tendency to “stick with the knowledge they had constructed collaboratively” (Swain, 1998, p. 79) can sometimes be detrimental to their L2 learning. Swain (1998) and Williams (2001), for example, found the close connection between the dyads’ incorrectly resolved LREs and their incorrect answers to the items in the tailor- made posttests. Swain et al. (2002) also pinpointed the various negative effects of peer collaboration on L2 learning from her review of the extant studies on collaborative dialogue. In addition, the examination of the relationship between collaborative dialogue and SLA thus far seemed to be mostly concerned with the quantitative measures (for example, posttest scores) of individual learners’ immediate retention of their LREs, whereas the microgenetic qualitative analysis of L2 learning that is central to sociocultural SLA posits that L2 development ought to be investigated in relation to learners’ correct use of the lexical items and grammatical forms in their collaborative dialogue for communication. In other words, the goal of the microgenetic analysis is to “discern internalization of L2 knowledge by learners as their interactions unfold

utterance-by-utterance” (Ellis & Barkhuizen, 2005, p. 236). As Swain et al. (2002) aptly pointed out, “More studies which investigate the transfer of knowledge to new contexts and longer effects are called for” (p. 181).

In addition to these two limitations, it also must be emphasized that in sharp contrast to the ubiquity of tailor-made posttests, pretests that assess learners’ prior knowledge of the L2 features were for the most part absent in the literature. For the time being, only the studies conducted by Swain and her colleagues (e.g., Swain, 1998, 2001b; Swain & Lapkin, 2001, 2002; Tocalli-Beller & Swain, 2007; Watanabe & Swain, 2007)

incorporated the results of pretests as the baseline for determining the growth of learners’ L2 knowledge that derived from their collaborative dialogue. This exclusion of pretests appears to be driven by three reasons. The first one is grounded in theoretical

considerations. According to Ellis (2000), since sociocultural SLA claimed that L2 learning occurs in interaction rather than as a result of it, it would be pointless to use pretests and posttests to separate the SLA process; instead, collaborative interaction or L2 learning should be examined “in its totality in order to show the emergence of learning” (p. 272).

The second one relates to practical issues. As Loewen (2005) noted, the

unpredictability of the linguistic items that emerged from learners’ collaborative work on communicative tasks made it next to impossible to create pretests to assess their prior knowledge. The third one is Williams’ (2001) claim that the presence of LREs per se was indicative of learners’ shortage of prior knowledge. According to her, if the learners in her study had been familiar with the LREs, they would not have “(1) requested

information about the word or form, (b) entered into a negotiation sequence surrounding it, or (3) produced an utterance containing a non-target-like version of it, prompting feedback from the teacher or another learner” (p. 335). Although these arguments seemed plausible, they ignored the very important fact that to reach the correct solutions to their collaborative dialogue, one member of the dyads should have at least some previous knowledge about the vocabulary or form-focused LREs. Without pretests assessing the degree to which learners were knowledgeable about the linguistic items in their LREs, it would be unwarranted to draw conclusions about the role of collaborative dialogue in L2 learning solely by virtue of learners’ performance on the posttests. This deficiency may

account for the discrepancies in the benefits that more proficient learners received from collaborative dialogue: while some studies (e.g., Leeser, 2004) suggested that higher proficiency learners did not benefit much from their collaboration with the low proficiency learners; others (e.g., Williams, 2001) maintained that higher proficiency learners were more likely to be the beneficiaries of the LREs they produced with their lower proficiency interlocutors. Apparently, due to the lack of pretests in these studies, it was hard to determine “if correct test scores represent the incorporation of new linguistic knowledge into the learner’s interlanguage system or if they represent a consolidation of previously existing knowledge” (Loewen, 2005, p. 382).