Leah Roberts

University of York, UK

Comparing on-line sentence processing performance to off-line (metalinguistic) responses: What can they tell us about grammatical knowledge in the L2?

In the field of SLA, the use of on-line, reaction time measures taken during the reading of grammatical and ungrammatical sentences began as a supplement to traditional off-line judgments, in order to add validity to off-line measures of second language (L2) learners’ underlying grammatical competence (see Juffs & Rodriguez, 2013 for an overview). The underlying assumption has been that off-line data can uncover what readers know is and is not possible in a given language (i.e., ‘limits of grammaticality’), whereas the workings the parser (or sentence processor)—comprising a set of procedures for implementing that knowledge in real time— can be investigated using time-sensitive techniques. In other words, many SLA researchers engaging in psycholinguistic research assume that on- and off-line tasks tap into different types of knowledge, for instance implicit versus explicit/metalinguistic (or procedural versus declarative) knowledge respectively. Studies in which off-line/metalinguistic knowledge has shown to be native-like, but on-line parsing procedures or sensitivity to un/grammaticalities is different has been argued to support fundamental difference accounts of SLA (e.g., Shallow Structure Hypothesis, Clahsen & Felser, 2006), whereas where there has been a match between on- and off-line measures, particularly as a function of proficiency/exposure, learners are assumed to have access to native-like knowledge of the L2 (e.g., Hopp, 2010), aligning with the view that L1-L2 differences in ultimate attainment are quantitative rather than qualitative. A review of the L2 sentence processing literature shows that native-like performance is often revealed when learners need also to make meta-judgments during real-time comprehension but not when they are required to merely read for meaning (see, e.g., Havik, et al. 2008). Does this mean that L2 learners of sufficient proficiency ‘shallow process’ unless pushed to perform deep hierarchical processing by the task? In this talk I will discuss such findings in relation to ‘types of knowledge’ that researchers assume the different tasks are tapping into, and what this means for the development of grammatical knowledge in L2 acquisition.