Heritage speakers and their grammars have become a prime objective for formal investigations into the human language capacity. In their keynote article, Polinsky and Scontras (Polinsky & Scontras, Reference Polinsky and Scontras2019) take these investigations further by asking what it would take to develop a model of the nature of heritage language grammars. By answering that question, we also gain deeper insights into the nature of human linguistic competence more generally and how this may change across the lifespan. Crucially, Polinsky and Scontras take on the hard problem of creating the foundations for a predictive theory of stable and vulnerable properties (set against an appropriate baseline; see Benmamoun, Montrul & Polinsky, Reference Benmamoun, Montrul and Polinsky2013, Polinsky, Reference Polinsky2018, and Polinsky & Scontras, Reference Polinsky and Scontras2019 section 1.3 for extensive discussion of this issue) in heritage language grammars. Predictive models would entail that we have a far better understanding of the various factors that shape heritage grammars, which in turn will inform outcomes of language contact situations more generally (see also Aboh, Reference Aboh2015). Their contribution is extremely timely and immensely valuable as the field is moving forward in trying to better model “language competence under unbalanced bilingualism” (Polinsky & Scontras, Reference Polinsky and Scontras2019, p. 10).

If we have sufficient information about the variables that determine heritage grammars, their interrelationship, and their weighting, it should be possible to predict a speaker's grammar. Although they don't discuss the typological or weighting issue, Polinsky and Scontras present two factors that trigger divergence. They also present three possible outcomes of such divergence. Here I will focus on the third, shrinking, since, as it stands, this outcome seems too general.

Scontras, Fuchs and Polinsky (Reference Scontras, Polinsky and Fuchs2018) find that in English-dominant heritage speakers of Spanish, the gender and number heads in the nominal domain have been fused. Such a process of fusion (or loss of one of the heads, in this case the gender head, cf. Scontras et al., Reference Scontras, Polinsky and Fuchs2018, p. 20) is subject to an adjacency requirement on functional projections. That is, setting aside details, (1) turns into (2).

1. (1) [_DP … [_NumP{xy} … [_GenP{xy} … [_nP … ]]]]

2. (2) [_DP … [_Num-GenP{xy} … [_nP … ]]]

Polinsky and Scontras (Reference Polinsky and Scontras2019) argue that this shows that heritage speakers often shrink their mental syntactic representations. They do not consider the possibility that fusion takes place in the morphological component (a possibility in theories like Distributed Morphology). Regardless of this latter possibility, shrinking is arguably not a necessity, as we will see next.

Riksem (Reference Riksem2017) investigates changes in the grammar of heritage speakers of American Norwegian, the heritage language stemming from Norwegians who moved to the US generally in the 19th century (see Haugen, Reference Haugen1953). Her paper is concerned with nominal morphology and compares data from the speakers in Haugen (Reference Haugen1953) with a subset of the speakers in the Corpus of American Nordic Speech (CANS; Johannessen, Reference Johannessen and Megyesi2015). Riksem (Reference Riksem2017) finds two main patterns: (i) omission of functional suffixes, both in plural and/or definite noun phrases, (ii) an increased usage of functional exponents from English. The first pattern is consistent with Polinsky and Scontras (Reference Polinsky and Scontras2019) whereas the latter does not really fit any of the three proposed possible outcomes.

Riksem (Reference Riksem2017) discusses two ways in which these patterns can be analyzed: (i) the syntactic structure is intact, and the changes are due to a change in the morphophonological exponents, (ii), the syntactic structure itself may have changed. The former analysis relies on a model within second language acquisition called the Missing Surface Inflection Hypothesis (MSIH), proposed by Lardiere (Reference Lardiere and Archibald2000) and Prévost and White (Reference Prévost and White2000). This model holds that there is no one-to-one relationship between overt morphological exponents and the underlying syntactic heads. The learner may not have acquired the relevant exponents, or the matching conditions between the syntactic structure and the exponents are not met. A fundamental claim is that a learner would rather omit a form than produce the wrong form (Lardiere, Reference Lardiere and Archibald2000), which again aligns with Polinsky and Scontras. However, as Riksem discusses (Reference Riksem2017, p. 21), the MSIH does not make clear predictions concerning where and how inflection may go missing, making it possible for avoidance to explain any instance where the syntax and the morphophonology do not align according to a given baseline. Riksem (Reference Riksem2017) favors the second hypothesis whereby the syntactic structure itself is the culprit for the diachronic changes. Notably, this does not appear to be shrinking of syntactic structures; rather a case of restructuring (cf. Hopp & Putnam, Reference Hopp and Putnam2015 on this notion). Such restructuring would be subject to a range of factors, including the inherent properties of the languages that are in contact in the mind/brain of the bilingual speaker. Arguably, then, restructuring (of the kind which is not shrinking) of grammatical representations is another potential outcome which may need to be added to the present typology.