Montrul's study is an important contribution to a recently emerged research approach to the study of bilingualism and languages in contact, characterized by its sound theoretical basis and its reliance on data from different – and traditionally non-integrated – domains of language development: bilingual first language acquisition (Müller and Hulk, 2001; Paradis and Navarro, 2003; Serratrice, 2004; Serratrice, Sorace and Paoli, in press), adult second language acquisition (Filiaci, 2003; Sorace, 2003), and native language attrition (Gurel, 2002; Tsimpli, Sorace, Heycock and Filiaci, in press). The generalization that is emerging from this approach is that interfaces between syntax and other cognitive systems (i.e. discourse pragmatics, lexical-semantics) exhibit more developmental instability than narrow syntax. For L1 attrition, which is the specific focus of the paper, this means that aspects of grammar at the syntax–discourse interface are more vulnerable to attrition than purely syntactic aspects. The identification of restrictions on the domain of occurrence of attrition is consistent with much previous descriptive research on this topic (e.g. Seliger and Vago, 1991). More recently, the same conclusion has been reached by a study on individual language attrition by Tsimpli et al. (in press), who investigated knowledge of the referential pronominal system in Greek and Italian in very advanced speakers of English. In this paper, Montrul tests the generalization on second-generation speakers of Spanish – or “heritage speakers” – a bilingual group that presents different characteristics from the adult L2 speakers investigated in Tsimpli et al.'s study. In addition to referential subjects, she also focuses on a different interface area of grammar – direct objects – that had not been investigated before. In these respects, Montrul's study is a welcome development. In other respects, however, the data are less than convincing and do not allow a straightforward interpretation. My commentary focuses on three fundamental questions raised not only by this study, but also by this type of research in general. The main focus will be the expression of referential subjects since this aspect of grammar has been investigated in other studies and therefore offers the possibility of direct comparison among results.