Not all grammatical features are robustly transmitted during the emergence of creoles

The paper engages with the Creole Debate concerning whether creoles form a typological class and whether they are structurally simpler than non-creole languages. Prior proposals include McWhorter’s Creole Prototype and the claim that creoles are the world’s simplest grammars, versus views that creoles are not linguistically exceptional and reflect sociohistorical ecologies (e.g., DeGraff, Mufwene). Aboh (2015) proposes creoles as mixed grammars via competition/selection from contact languages. Blasi et al. (2017) argue via statistical modeling that grammars are robustly transmitted during creole emergence, questioning pidgin stages and creole-specific innovations. This paper challenges both sides: (i) certain core grammatical components (notably bound morphology and tones) tend to be reduced during creolization, so transmission is not robust across all domains; (ii) creoles are not necessarily simplest, as they can retain or develop complex distinctions in syntax, semantics, and phonology. The paper outlines a cognitive framework of SLA processes underlying creolization, critiques Blasi et al.’s feature selection, and illustrates complex aspects of creole grammars.

Key positions in the debate include: McWhorter (1998, 2001) posits a Creole Prototype (minimal inflection, minimal tones, semantic transparency) and argues creoles are the world’s simplest grammars. Opposing views (DeGraff 2003, 2005; Mufwene 1997) deny creole exceptionalism and attribute perceived simplicity to sociocultural histories, warning against biases. Aboh (2015) models creoles as mixed grammars via competition/selection among features from European and African sources. Blasi et al. (2017) use APICS and WALS features with machine learning to claim grammars are robustly transmitted in creole emergence. Work connecting SLA and creolization (Schumann 1978; Andersen 1980, 1983; Siegel 2003, 2006; Plag 2008–2009; Sessarego 2013) emphasizes adult SLA processes. Good (2012, 2015) distinguishes paradigmatic vs syntagmatic complexity and argues morphology/tones are less likely to transmit. Additional studies highlight simplification/regularization tendencies in adult learning (Slabakova 2008, 2009; Saldana et al. 2018) and feature geometry/interface effects (Harley & Ritter 2002; Sorace 2011; Sessarego 2012, in press; Rao & Sessarego 2016).

The study is theoretical/analytical rather than experimental. It advances a cognitive framework for creolization grounded in SLA: (1) Borrowing under recipient language agentivity affects primarily lexical items; (2) Imposition under source language agentivity affects structure (syntax, phonology, semantics), whereby L1 patterns are imposed on L2 outputs; (3) Grammatical reduction during SLA disproportionately targets bound morphology and tones, the acquisition bottleneck, leading to reduced paradigms and generalization of defaults. Following Good (2012), the paper distinguishes paradigmatic complexity (multiple forms within a category, hard to transmit) from syntagmatic complexity (combinatorial patterns, easier to transmit). The paper critiques Blasi et al. (2017)’s feature selection from APICS/WALS: of 92 features, 30 syntax, 27 semantics, 12 phonology, 7 lexicon, 15 morphology, 1 tones—skewed toward syntax/semantics (≈62%), with morphology+tones only ≈17%, thus omitting domains central to the debate. It argues this bias drives their conclusion about robust transmission and misclassification of highly inflected or tonal non-creoles as creoles in their model. Comparative illustrations are provided: Media Lengua (a mixed language created by proficient bilinguals) vs Palenquero (a creole) to show preservation vs loss of rich morphology; and Sranan Tongo case studies to demonstrate complex semantic, syntactic, and phonological distinctions in creoles via substrate imposition. No new quantitative modeling is performed; rather, existing datasets and examples from the literature are analyzed to motivate the framework.

- Transmission is not robust across all grammatical domains during creole emergence: bound morphology and tonal features tend to be reduced or not transmitted due to SLA bottlenecks and paradigmatic complexity. - Creoles are not the simplest grammars: through imposition, they can maintain or develop complex distinctions in semantics (e.g., copula systems distinguishing predicate types), syntax (e.g., serial verb constructions), and phonology (e.g., distinct nasal phonemes), as shown for Sranan Tongo compared to English. - Mixed languages (e.g., Media Lengua) arise under different cognitive conditions (proficient bilingualism), allowing robust preservation of morphosyntax, unlike creoles formed via adult SLA by speakers dominant in substrate languages. - Blasi et al. (2017)’s conclusion of robust grammatical transmission is likely an artifact of feature selection biased toward syntax/semantics and away from morphology/tones. Feature distribution in their dataset: 92 features total, with 30 syntax, 27 semantics, 12 phonology, 7 lexicon, 15 morphology, and 1 tone; syntax+semantics ≈ 62% vs morphology+tones ≈ 17%. - Machine-learning classifications based on such skewed feature sets can mislabel highly morphological or tonal non-creoles (e.g., Mapudungun, Quechua, Yoruba, Malagasy, Arabic, Greek) as creoles, undermining the robustness claim. - Understanding creole formation requires modeling borrowing, imposition, and grammatical reduction, rather than treating all features equally or relying on random/"unbiased" selection.

The findings address the core question of whether creoles are typologically distinct and/or simpler by showing that creole formation is governed by specific SLA-driven processes with differential effects on grammatical domains. This explains why lexicons tend to derive from European lexifiers (borrowing) while many structural traits reflect African substrates (imposition), and why paradigmatically complex domains (morphology, tones) are frequently reduced. Consequently, claims of robust transmission and extreme simplicity are oversimplifications. The critique of feature selection underscores that statistical detection of a "creole profile" depends on which domains are represented, and that excluding morphology/tones biases outcomes. The study argues for integrating cognitive mechanisms with sociohistorical ecologies and, where possible, diachronic data to reconstruct creole trajectories, rather than inferring pathways solely from synchronic typological profiles. This reframing shifts the debate from counting features to explaining learnability and transfer under realistic contact conditions, with implications for typology, SLA, and contact linguistics.

The paper proposes a cognitive-process-based account of creolization highlighting borrowing, imposition, and grammatical reduction as distinct mechanisms shaping creole grammars. It argues that not all grammatical features are robustly transmitted—morphology and tones are often reduced—while creoles can exhibit substantial complexity in syntax, semantics, and phonology via substrate imposition, countering the claim that they are the world’s simplest grammars. The critique of Blasi et al. (2017) shows their results reflect a skewed feature selection that largely omits morphology and tone, inflating perceived robust transmission. The paper recommends moving beyond random feature aggregation to analyses grounded in cognitive constraints and contact ecologies, and calls for integrating historical/diachronic evidence to trace diverse creole developmental paths.

The study is theoretical and does not provide new empirical modeling or comprehensive quantitative reanalysis; it critiques existing feature selections and illustrates arguments with selected examples. It notes that synchronic features alone may not reveal evolutionary trajectories; robust conclusions require diachronic linguistic evidence and rich sociohistorical records, which are often scarce. The paper does not claim that all creoles underwent a pidgin stage and acknowledges variability across cases. Future work should incorporate morphology and tone in balanced datasets and re-run simulations accordingly.