The Tale of Two Lexicons: Decomposing Complexity across a Distributed Lexicon

Abstract

The notion of complexity is evasive and often left to intuition, yet it is often invoked when studying heritage language grammars. In this article, we propose a first pass at decomposing the notion of complexity into smaller components in a formal grammatical model. In particular, we argue that a distributed model of the lexicon (i.e., one that assumes that principles that generate both words and phrases are one and the same) allows us to identify three components: syntactic features, the hierarchical organization of features, and the mapping between syntactic features and their exponents. Based on grammatical gender in different language pairs, in particular the heritage language American Norwegian (AmNo), we illustrate how this distributed model can account for developments in heritage language grammars whereby the grammatical gender system is considered to have become less complex. More generally, the article demonstrates that a distributed architecture is better suited empirically and theoretically as a heuristic to understand complexity effects in heritage grammars and beyond.

(In)frequently asked questions: On types of frequency and their role(s) in heritage language variability

“…In contrast, less frequent HL forms would be more likely to favor the emergence of grammatical innovations (Backus, 2020) or morphosyntactic variability (Poplack et al, 2013;Perez-Cortes, 2022a). These hypotheses are compatible with recent theoretical proposals regarding the nature of the lexicon, especially those that advocate for an exoskeletal approach to morphology (Embick, 2015;Lohndal and Putnam, 2021). In particular, the adoption of a distributed view of lexical items (as the result of abstract morphosyntactic (synsem) features being mapped onto specific (morpho) phonological exponents) provides us with a systematic way to model and predict how frequency in the input could either reinforce such mappings, or allow for a disassociation between them, generating a wide range of outcomes that could have consequences at the level of production as well as representation (Perez-Cortes et al, 2019).…”

When covert modality sneaks into your grammar: wh-infinitives in American Norwegian

“…The proposal that wh -items can lexicalize covert modality and undergo Aʹ-movement supports a version of representation economy (Scontras, Polinsky, and Fuchs, 2018) that avoids ‘silent elements’ (Laleko and Polinsky, 2017). At the same time, the occurrence of complex spans such as these are not as simplex as one-to-one feature-exponent mappings, which also seem to be preferred in heritage language (morpho)syntax (Lohndal and Putnam, 2021). The existence of spans such as these in heritage language syntax certainly raise interesting questions concerning how best to understand exactly what ‘representational economy’ and complexity in these systems mean.…”

“…The existence of spans such as these in heritage language syntax certainly raise interesting questions concerning how best to understand exactly what ‘representational economy’ and complexity in these systems mean. As recently proposed by Lohndal and Putnam (to appear), the tendency to minimize morphosyntactic complexity in agglutinating heritage languages (e.g. Hungarian and Turkish) via ‘expanded structures’ such as span could be viewed as a benefit since it reducing the number of mappings between spans and exponents.…”

“…According to this line of reasoning, the size and (feature) content of syntactic objects (i.e. spans) would be ‘bigger’ in certain situations where the trade off might result in an ease of communication or parsing (Lohndal and Putnam, to appear). 23 Second, these findings further confirm that non-finite clauses (Putnam and Søfteland, 2021), the (morpho)syntactic expression of modality (Putnam et al, 2019), and Aʹ-dependencies more generally (Cuza, 2013, 2015; Cuza and Frank, 2010, 2011; Cuza et al, 2016; Hopp et al, 2019; Montrul, Foote, and Perpiñan, 2008) in heritage language grammars can be affected over their course of development.…”

Feature-exponence mapping in language contact

“…The keynote states that instead of feature reassembly 'complexity' in the sense of Lohndal & Putnam (2021) is relevant: Feature bundles that are more complex take longer to acquire, whereas a feature bundle consisting of one feature is predicted to be simpler. This is also because the mapping from feature bundles to exponents tends to be more complex when more features are part of the bundle.…”