The article by Clahsen and Felser (CF) on grammatical processing in language learning is a timely and much-needed synthesis of research on this topic. It correctly identifies both morphological processing and syntactic processing as key areas that require attention. This commentary raises two issues: the relationship between the grammar and the parser, and the need for a transition theory in adult second language (L2) learning.

CF advance a viewpoint, one long been held by Clahsen and colleagues, that adult L2 acquisition and representation is fundamentally different from first language (L1) acquisition and adult native speaker competence (Bley-Vroman, 1989; Clahsen & Muysken, 1986). The article proposes further evidence that this is the case: first, CF point to evidence that young children focus primarily on structure during processing, to the exclusion of plausibility constraints; second, children are less able to revise an analysis, and resort to bottom-up processing in contexts of high processing demand. In contrast, adult learners make primary use of lexical information to resolve ambiguities, but seem unable to make use of structural gaps (Marinis, Roberts, Felser, & Clahsen, 2005).

The question that arises is whether the data discussed by CF constitute evidence that L2 processing is just different, or fundamentally different. In this context, a clarification about the relationship between the grammar and parser is necessary. CF are implicit about their assumptions, and it would be helpful if their characterization of this relationship were clearer. It is accepted that syntactic processing involves both bottom-up and top-down information. However, the concern is what system drives the bottom-up portion. If one adopts the somewhat controversial position (in some quarters anyway) that the relationship between the grammar and parser is one of local application of global principles (Pritchett, 1992; Weinberg, 1999), then a claim that L2 parsing is fundamentally different becomes problematic. If we assume that a formal grammar is one that describes operations on a set of symbols, then one has to decide which set of symbols one will pick to describe processing behavior. If the choice is a Principles and Parameters grammar, or its current Minimalist incarnation (Chomsky, 1995), then one is dealing with parser that is involved in feature checking. CF's use of gaps and empty categories puts them squarely in this camp. (Note that this is not necessarily true of a theory that would be based on head-driven phrase structure grammar, for example, because it is a theory that does not posit empty categories; Borsley, 1998). Under current minimalist assumptions, feature checking is required even for lexical integration, which CF concede that L2 learners do well. Given this assumption, L2 learners do not face difficulties with the formal feature checking mechanism; rather, one must assume that a specific subset of features are affected, for example, [±wh] and perhaps other functional features.

In this context, CF propose the “shallow processing” hypothesis. This hypothesis is that thematic and pragmatic information are used in processing but not structural gaps involving [±wh] features. However, their characterization of this shallow processing will need some further formalization before it can be empirically tested. Specifically, why is it that some uninterpretable features that need to be checked can be checked successfully, for example, Case and argument features, but others cannot. Moreover, in CF's sentence, which is used to illustrate shallow processing, an object relative clause that modifies a Subject is presented. In this context, we have two verbs next to one another. As Juffs (2005) suggests, this problem may cause L2 learners problems, as it seems to for adult native speakers (e.g., King & Just, 1991). It is unclear whether object relatives that modify an object cause similar problems: the prediction is that a sentence such as “the tiger chased the lion that the hunter tracked _______ in the forest” should cause the same difficulty as “the lion that the hunter tracked _______ chased the tiger in the forest.” If shallow processing is correct, will L2 learners find both these structures difficult because the mechanism is “impaired,” or will the objects that are modified be easier because the features of the object and the relative pronoun are the same?

The second issue that CF's review of the literature raises, indeed, the first words of their abstract zeroes in on this issue, is the theoretical claim that processing is a part of a transition or acquisition theory (Fodor, 1998; Gregg, 1996). The data from children are suggestive in that children appear to pay attention to structural cues to the exclusion of pragmatics, a behavior that suggests that parsing failure might drive changes in the grammar as suggested by Fodor (1998) and by Johnson and Newport's (1989) “less is more” hypothesis. However, it is not clear from the literature that CF review that parsing is, in fact, as deeply involved in acquisition as theory proposes. Crucially, no empirical studies are reviewed that specifically test the hypothesis that that parsing breakdown is, in fact, a trigger for restructuring of the grammar in children. This is not something that CF are responsible for, but given current formal approaches to language learning, one might have expected some experimental evidence to show how the focus on structure among children leads to restructuring of the grammar. In a sense, this is another gap in the research, and not a critique of CF's review. However, in L2 acquisition, there does seem to be some evidence that targeted processing can assist in acquisition (VanPatten, 1996, 2002; but cf. DeKeyse, Salaberry, Robinson, & Harrington, 2002).

On a more practical note, CF's characterization of L2 learners' processing has a pedagogical implication. If it is true that the L2 learners' deficit is that they process only “shallowly,” then the data CF present are further evidence that teaching techniques that emphasize communicative competence (Canale & Swain, 1980; Paulston, 1974; i.e., approaches that encourage integration of grammatical, pragmatic, sociolinguistic, and strategic competence), to the exclusion of attention to form may not be appropriate for successful L2 acquisition. This is to say that it may not be surprising that learners have not been able to home in on some narrow structural properties of their second language for processing, if instruction encourages use of pragmatics for successful comprehension. If shallow processing is available in the L1, and can be transferred to the L2, then a least effort principle may permit them to process the L2 without the necessary structural analysis. Methods that require processing and comprehension, drawing attention to processing errors (e.g., VanPatten, 1996) will be helpful in this regard.

In their conclusion, CF note that their conclusions are preliminary. It remains true that there are very few studies that use psycholinguistic techniques to investigate L2 processing. This article should be seminal in promoting further research.