The proposal by Clahsen and Felser (CF) has the potential of marking a turning point in second language (L2) acquisition research. Contrary to much L2 research to date, it suggests that some of the differences between native and (advanced) nonnative speakers may be at the level of grammatical processing, rather than grammatical representations. Accounting for L2 speakers' divergent behavior does therefore not necessarily involve positing “representational deficits”: L2 speakers can, and indeed do, attain target representations of the L2, but may compute incomplete (“shallow”) syntactic parses in comprehension. Such shallow processing is often accompanied by reliance (or overreliance) on lexical, semantic, and pragmatic information, which can lead to seemingly trouble-free comprehension in ordinary communication. It is only when speakers are faced with sentence ambiguities, which impose a greater than normal processing load, that the differences between adult L2 language acquirers and child first language (L1) acquirers become apparent: both child and adult language learners have difficulty in integrating structural and nonstructural information in on-line comprehension, but although children prioritize structural information, adult learners privilege nonstructural lexical–semantic information. CF's proposal opens up new perspectives on the nature of ultimate attainment in adult L2 acquisition; at the same time it raises some questions that, in my view, are of crucial importance for future research.

First, is shallow processing a strategy that L2 speakers employ only in particular circumstances, or do they have the ability for deep processing too? CF seem to believe that L2 speakers are restricted to shallow processing, that is, have no choice. It is possible that developmental data might show a decreasing reliance on shallow processing over time, and perhaps a discontinuous change marking the difference between advanced and nearly native L2 speakers. Second, is shallow processing more likely to occur in particular domains that favor this strategy because of their greater complexity? If computational overload is what induces L2 learners into shallow processing, this may be more visible with respect to constructions that are computationally more complex, and/or in on-line situations of language use that stretch L2 speakers processing abilities to their limits.

Third, does shallow processing take different forms? CF argue that it does not normally result in transfer of L1 parsing procedures, because these need a certain (minimum) amount of structural detail to be applied. However, it is not implausible to assume that shallow processing may at least occasionally result in the use of L1 syntactic representations if these are more “economical” than the corresponding target L2 ones.

Fourth, does shallow processing apply to production as well as comprehension? CF's arguments are based on behavioral and electrophysiological data from comprehension experiments, but to the extent that production involves syntactic analysis and encoding (Levelt, 1989), it may exhibit similar strategies and be amenable to similar explanations.

I now briefly illustrate the relevance of these questions to an area in which CF's proposal might shed new light. Recent research (e.g., Belletti, Bennati, & Sorace, 2005; Sorace, 2003; Tsimpli, Sorace, Heycock, & Filiaci, 2004) has shown that residual divergence and optionality in L2 nearly native grammars are largely restricted to the interface(s) between syntax and other cognitive domains. Furthermore, the same interfaces are vulnerable to instability/incompleteness in other populations of adult bilingual speakers, such as L1 speakers experiencing attrition from prolonged exposure to an L2. For example, L1 English nearly native speakers of L2 Italian have been found to overuse overt pronouns in contexts in which native Italian speakers would use a null pronoun, as in Example 1, where the overt pronoun is coreferential with a topic antecedent. The same speakers may also overuse preverbal subjects (often stressed) in contexts in which it would be more natural to use a postverbal subject because the determiner phrase (DP) is in narrow focus, as in Example 2.

1. Maria è andata via perchè lei era stanca. Maria is gone away because she was tired.
2. Chi è partito? Un mio amico è partito. Who has left? A friend of mine has left.

Not only the production but also the interpretation of overt pronouns and preverbal subjects is affected by optionality in these speakers. For example, they may interpret the overt pronoun lei in the biclausal forward anaphora sentence in Example 3 as coreferential with the subject of the matrix clause (la mamma); they may also interpret the preverbal subject in Example 4 as conveying new information (i.e., as referring to a different ship), instead of a partitive reading in which the sunk ship is one of the three already mentioned in the previous sentence (Pinto, 1997).

1. La mamma dà un bacio alla figlia mentre lei si mette il cappotto. The mother gives a kiss to the daughter while she puts on the coat.
2. Il governo ha mandato tre navi nel Golfo. E' affondata una portaerea. The government has sent three ships to the Gulf. Has sunk an aircraft carrier.

These speakers can use null subjects and postverbal subjects in a targetlike way: they have acquired a null subject grammar by setting the null subject parameter to the required positive value. However, this is clearly not sufficient to ensure consistently nativelike production and interpretation of pronominal subjects and subject placement in constructions at the syntax–pragmatics interface.

Do interface constructions pose residual problems at the representational or at the processing level? Consider the overuse of preverbal subjects in Example 2. One possibility (see Belletti, 2005; Belletti et al., 2005, for details) is that L2 speakers of Italian whose L1 is English fail to activate the verb phrase (VP) internal focus position required by focalization in Italian, as shown in a simplified form in Example 5.

1. [CP…[TP pro…è…partito…[TopP [FocP…un mio amico [TopP [VP…]]]]]

The result is the use of focus in situ, namely, an L1-based strategy that is more economical because it involves an “activated” DP internal focus position as the one overtly manifested in a sentence like “John himself came.”

It is worth noticing that L1 French speakers of L2 Italian often use clefting in the same context (Belletti & Leonini, 2004), which is an alternative way of activating the VP periphery (as shown in Example 6) and is widely available in French.

1. Ce…[Top [Foc [Top [VP être [sc Jean [CP qui est parti]]]]]]

Suppose, then, that this is a case of shallow processing: speakers do not always activate the VP periphery in producing narrow focus sentences, and instead of using postverbal subjects they resort to either an L1-based form of activation of the VP periphery (as in the case of French speakers), or an entirely different, but still L1-based, focalization strategy, such as DP internal focus in situ (as in the case of English speakers). Interpreting these phenomena in the light of CF's hypothesis allows us to identify their source in the persistence of an L1-based discourse “prominent” strategy, employed to compensate for the failure to compute the required L2 syntactic representation, despite the potential “grammatical” availability of the latter. In comprehension, shallow processing may also involve the optional lack of activation of the VP periphery, which is necessary to the reading of the postverbal subject as carrying focus on new information. The result is a partitive interpretation of an indefinite postverbal DP.

What about the different distribution of overt subject pronouns in nearly native Italian? The production and interpretation of overt subject pronouns in Example 1 and Example 3 may stem from shallow processing of the interface conditions governing the use of overt subjects (e.g., the obligatory presence of the feature “topic shift”; see Tsimpli et al., 2004) and consequent assimilation of strong Italian pronouns to the corresponding weak English pronouns, which, unlike Italian overt pronouns, can refer to topic antecedents. The strategy used in these circumstances would be different from the use of overt pronouns a default form to relieve processing overload due, for example, to insufficient knowledge of (or access to) agreement inflection (Bini, 1993; Liceras & Diaz, 1998; Sorace, 2005).

Both the phenomena just described involve optionality of shallow processing, that is, the L2 speakers' ability to perform full processing, at least at the nearly native level. Shallow processing, in this sense, would be a relief strategy that is available to all speakers but is relied on especially by bilingual speakers. Arguably, the coordination of syntax and pragmatics may be more likely to exceed bilingual speakers' capacity and to force them to resort to shallow computations. Neither access of pragmatics information nor access to syntactic knowledge is problematic, as CF point out. What remains inconsistent is the targetlike integration of this information within the syntactic representations that are computed in on-line production and comprehension. More behavioral and neurological data are needed to support this interpretation of near-native optionality; meanwhile, CF's contribution has paved the way to new explorations of these phenomena.