In their Keynote Article, Clahsen and Felser (CF) provide a detailed summary and comparison of grammatical processing in adult first language (L1) speakers, child L1 speakers, and second language (L2) speakers. CF conclude that child and adult L1 processing makes use of a continuous parsing mechanism, and that any differences found in processing can be explained by factors such as limited working memory capacity and incomplete lexical knowledge. The authors then suggest that the existing differences between L1 (both adult and child) and L2 processing provide evidence that parsing mechanisms are qualitatively different between these groups. They posit that this qualitative difference between L1 and L2 is due to L2 speakers having shallower and less detailed syntactic representations than L1 speakers. This commentary focuses on discussing this shallow structures account and considers what this means for L2 processing.
CF discuss three possible explanations for the differences between L1 and L2 processing. The three possibilities explored are whether the differences found can be accounted for by an incomplete target grammar, cognitive limitations, or interference from the L1. Data presented do not fully support any of these accounts, although it is generally accepted that all these factors do have an effect at some level of L2 processing. The authors then discuss the procedural/declarative and implicit/explicit memory distinction to explain L1 and L2 processing differences proposed by Ullman (2001) and Paradis (1994, 1997, 2004). To deal with the inadequacies seen in Ullman's and Paradis' proposal, CF then put forward their shallow structures account. This account is an alternative idea to those already proposed in the literature. It is proposed to account for the qualitative differences found between L1 and L2 sentence processing on the one hand, and the lack of such qualitative differences in the domain of morphology. Although this idea seems at first to provide a full explanation of the L1 and L2 differences as well as similarities, there do seem to be some issues that are inconsistent with both the hypothesis and the data. The questions that arise from these inconsistencies are: does the shallow structures account allow for real qualitative differences? and what other data are needed before we can determine precisely how L1 and L2 processing are different.
QUALITATIVE DIFFERENCES AND THE SHALLOW STRUCTURES ACCOUNT
The shallow structures proposal put forth by CF accounts for the finding that L1 and L2 processing differences in the domain of morphology were less dramatic than the differences in the syntactic domain. CF summarize previous L2 processing studies and compare these to findings from L1 processing, and are able to conclude that differences in morphological processing could be attributed to factors such as incomplete knowledge of the target language. Differences in syntactic processing, on the other hand, are claimed to be due to these shallower and less detailed representations (which are defined as representations without hierarchical syntactic organization). CF claim that the L2 speakers are able to acquire a nativelike lexical–semantic representation but their syntactic representations lack this hierarchical detail and their representations do not contain any abstract information. This shallow structures account does help to explain why L2 speakers' processing differs from that of native speakers. However, it then seems that the authors weaken their own interpretation by going on to claim that these same shallow representations also affect native-speaker processing under certain circumstances. This to me suggests that differences between L1 and L2 processing are considered by CF to be due purely to quantitative effects. If what the L2 speakers are doing is sometimes seen in L1 processing, then this seems to be alluding to the presence of some continuity between the L1 and L2 processing, similar to what was proposed to account for the differences between child and adult L1. If an L2 speaker could reach native-speaker level proficiency then they should also be able to add depth and detail to their representations. This contradicts what seemed to be CF's initial reason for proposing a shallow structures account, which is that L2 processing seemed to be qualitatively different from L1 processing at least at the level of sentence processing. How then can we resolve this inconsistency between the presence of qualitative differences and a proposed account that can also be used to explain some aspects of L1 processing?
WHAT OTHER DATA ARE STILL NEEDED?
The nature of any qualitative differences between L1 and L2 processing still needs to be addressed. CF do note that further research needs to be done, and they provide some direction for this. However, I would like to add that before any firm conclusions and hypotheses can be made concerning differences in the L1 and L2 we need to evaluate the effects of proficiency, transfer, and age of acquisition. It does not seem that any one of these can fully predict when native versus nonnative processing will occur.
Neuroimaging studies of early bilinguals (who are highly proficient in both their languages) still show differences in how they process each of their languages. Perani et al. (2003) looked at early (L2 learned at age 3) Spanish–Catalan bilinguals in a verbal fluency functional magnetic resonance imaging task. Functional comparisons indicated that there was less extensive areas of activation in the brain for the language acquired first. This finding indicates that even with a very early age of acquisition for the L2, functional differences can be found. This study also found that amount of exposure (even with very early L2 acquisition) to the L2 can affect the pattern of brain activation. The L1 Catalan participants in this study had less exposure to their L2 and showed a greater activation to Spanish than the Spanish participants showed to Catalan.
The effect of language transfer and interference also still needs to be investigated in more detail before its precise effects on L2 processing are understood. In a study looking at L2 grammatical gender and the effect of L1 it was found that having gender in the L1 does help the acquisition of an L2 grammatical gender system (Sabourin, 2003; Sabourin, Stowe, & de Haan, in press). On the other hand, an ERP study demonstrated that this effect of transfer was further complicated depending on whether the L1 and L2 grammatical gender systems were similar (Sabourin, 2003; Sabourin & Stowe, 2005). Just having grammatical gender in the L1 helped L2 learners acquire the necessary knowledge of the L2 system (in terms of scores on a behavioral task), but in terms of nativelike processing there was an added requirement not only of having an L1 gender system but that the system must be similar.
These findings, as well as those discussed by CF, suggest that L2 processing is very complex. Further research investigating many other types of linguistic structures and levels as well as looking at a greater number of types of L1s and L2s is still necessary before any firm conclusions can be made. In addition, a fuller understanding of how proficiency levels, age of acquisition, and transfer affects L2 acquisition and processing must be attained before we can begin to propose what it is that is different between L1 and L2 processing.
