Sunday 28 November 2010

Loose Ends? Commentary on Sorace

[This commentary on a position paper by Antonella Sorace has now appeared in Linguistic Approaches to Bilingualism 1: 35-38]

Writing a commentary that makes an original contribution without being self-serving or carping—or simply rehearsing what everyone else will have written, in poorer prose—is never easy. In the present case, however, the task is rendered particularly difficult by the fact that Sorace has largely done the work for us. Rather than being presented with an article containing radical or overbearing claims, as one typically finds in position papers, we are here asked to comment on what is essentially a self-assessment exercise: a moderate, well-informed and nicely argued discussion of the main claims that would have featured in the position paper, had it existed. But it doesn’t, and in offering in its place her own review, in which all of the obvious criticisms of the IH are addressed and skillfully deflected, Sorace has effectively ‘headed us off at the pass.’ That said, the proposal that emerges from Sorace’s treatment is not completely immune to further criticism: in the limited space remaining, I shall make a couple of specific points that bear on a more general— persisting—concern about the overall proposal, namely, the problem of circularity.

In order to illustrate the IH, Sorace reviews a series of studies of L2 acquisition and/or L1 attrition by various colleagues and co-workers: in each instance, a near minimal opposition is presented between two superficially similar phenomena: on the one hand, a grammatical construction or property that is easily acquired by second language learners (or is unaffected by attrition), and on the other one that presents some measure of difficulty for near-native L2 learners (and L1 attriters), reflected in unstable or errorful performance vis-à-vis monolingual controls. In every instance, the latter construction is identified as the one involving an interface of some kind, whereas the former calls on ‘narrow syntax’ (either exclusively, or to a larger degree). Though Sorace is generally careful to avoid any explicit claim, the clear implication in each case is that it is the ‘interface properties’ of the construction in question that are responsible for the learners’ difficulty. But it is also true in most cases that it is the relative difficulty of the construction that provides the diagnostic of interface involvement: again by implication, if monolinguals and near-native learners showed complete convergence in performance on a given construction, then interface properties would not be to the fore. The danger of circularity of reasoning should be obvious: in the absence of some principled and independent criteria for defining what an interface is—let alone for distinguishing external vs. internal interfaces—the term ‘interface’ runs the risk of becoming simply a non-explanatory formal label for ‘learner difficulty.’ Sorace is clearly aware of this problem, as most of the article—section 3, in particular—is devoted to trying to unpack the notion of interface. Yet in spite of a sterling attempt, it is hard to escape the conclusion that, ultimately, the tail is wagging the dog: that is to say, the phenomenological or behavioural effects are being used to determine the architecture of the system.

Aside from the general worry produced by this form of argumentation, working from constructional evidence to theory brings with it some more specific theoretical and methodological difficulties. One such problem is that of cross-linguistic generalization: what to do if the constructional properties associated with core or interface in one grammar are aligned slightly differently—or even in a reverse fashion—in another. A case in point arises in Sorace’s treatment of contrasts between Topicalisation vs. Focus constructions in Italian and Greek. In discussing the apparently contradictory results of Tsimpli and Sorace (2006), whose Greek learners showed no difficulty with Focus constructions, but instability with Topicalisation vs. those of Belletti (2007), whose Italian learners showed difficulty with narrow focus constructions, Sorace suggests that the contrary results may indicate that Greek and Italian Focus constructions are not in fact formally identical (‘[implying] the need for a more fine-grained differentiation among interface conditions’.] If this is the case, though, then one or other set of contrasts must be abandoned as diagnostic of a formal interface: either way around, it seems, the problem of circularity raises its head.

Matters become still more complex with respect to the Topic/Focus distinction when we look a little further afield, to ‘Topic-prominent’ languages such as Vietnamese, in which null arguments are licensed and identified not by verbal agreement, but exactly by (overt and non-overt) Topic arguments. As the following examples from Cao Xuân Hạo (1992:145) clearly demonstrate, pro subjects in Vietnamese are exclusively licensed by preceding topics, rather than by subjects: the (a) examples make sense because the ‘hanging topic’ (1a), and topicalized shifted object in (2a) provide appropriate antecedents for the null arguments in the following clause; by contrast, the topicalized subjects in the (b) examples are inappropriate as antecedents, giving rise to interpretive anomalies:

(1)  a.              Xã      bêni       ruộng       tốt    nên proi  rất giàu.
                        village side     rice-field  good should     very rich
                        ‘As for the neighbouring village [Topic], its rice-fields [Subject] are good (fertile), and therefore it [pro subject = the village] is very rich.’

       b.       [*Ruộng của    xã         bên]i tốt nên [pro]i rất giàu.
                  rice-field poss village side good should very rich
                 #’The neighbouring village’s rice-fields [Subject/Topic] are good (fertile), and therefore they [pro subject = the rice-fields] are very rich.’

(2)     a.     Ông nàyi  tôi quen   nhưng proi không   phải    là     bạn  tôi.
                  man dem  I  know   but               neg  right  cop friend    I
                  ‘As for this man [Topic], I [Subject] know (him), but he [pro subject = the man] is not my friend.’

          b'.      *Tôi quen ông này nhưng [pro] không phải là bạn tôi.
                    I know man dem    but               neg right cop friend I
                    #‘I [Topic = Subject] know this man, but I [ pro subject = I] am not my friend.’

These examples, however, seem to force the conclusion that Topicalisation in Vietnamese (and other null-topic languages) is functionally equivalent to null subject licensing in Italian, which Sorace definitely assumes to be a core syntactic process, certainly, one in which ‘errors involving misuse and misunderstanding of null subject pronouns are not attested: both native and near-native speakers of Italian have a clear and determinate preference for the subject of the matrix clause as the antecedent of the null subject pronoun.’ This begs the question of how to treat East Asian Topicalisation: as equivalent to Italian Topicalisation constructions (high interface involvement) or, as equivalent to Italian null-subject licensing (low interface involvement)? The predictions for near-native learners are entirely unclear: what we certainly can’t do is to let their behaviour decide.

In short, though the IH is better looking than ever before following Sorace’s rigourous self-analysis, there is still something unsettling about it: while it may be that more data will help to decide the issues, in the absence of some more principled criteria, scepticism will remain.(Cao 1992)

Reference

Cao, X. H. (1992). "Some preliminaries to the syntactic analysis of the Vietnamese sentence." Mon-Khmer Studies 20: 137-152.