The cognition of dialect variation: evaluating the relationship to bilingualism

1 Introduction

Dialect variation is well-documented in the world, but there has been limited theorizing regarding the cognitive mechanisms that enable a person to speak multiple dialects (see review in Heggdal Lønes et al., 2023). Bilingual cognition has received comparatively more research, with many models put forward to explain how speakers navigate the multifaceted demands of using multiple languages, such as switching between languages and avoiding production of words from the nontarget language (e.g., Green and Wei, 2014).

Bilingualism and bidialectalism are intuitively similar phenomena, where multiple linguistic systems are represented in a given person's mind, so it could make sense to extend models of bilingual cognition to bidialectal cognition. Focusing on language production, we ask whether bidialectalism and bilingualism should be theorized as primarily distinct or equivalent phenomena, considering how broad questions about the relationship between bidialectalism and bilingualism reveal gaps for future research¹.

2 Typological preliminaries

The traditional distinction between “languages” and “dialects” hinges on mutual intelligibility: two linguistic systems are different dialects if mutually intelligible and different languages if not (Wolfram and Schilling, 2015). For clarity, we will use the terms dialect/bidialectalism in situations of clear mutual intelligibility, language/bilingualism in situations of mutual unintelligibility, and codes/bicodalism as general terms.

The intelligibility-based distinction between languages and dialects proves less clear upon closer inspection. Labels are applied inconsistently, with mutually intelligible codes sometimes labeled as “languages” when crossing national boundaries (e.g., Norwegian/Swedish, Hindi/Urdu) or mutually unintelligible codes labeled as “dialects” when unity is politically favored (e.g., Chinese “dialects”). Of greater theoretical significance, mutual intelligibility can be borderline, reflecting the continuous nature of levels of relatedness between different codes. For example, at the lexical level, cognates are found across language pairs, with 85% lexical overlap serving as an approximate threshold for mutual intelligibility or “dialects” (Kirk et al., 2018).

At the same time, the intelligibility-based definition captures a meaningful difference that can exist between pairs of codes. For example, an American and a New Zealander would typically communicate using their own varieties of English. Conversely, if an L1 American English speaker and an L1 Dutch speaker wish to communicate, one person must also speak the other's L1. To what extent, then, can bilingual and bidialectal cognition be explained by appealing to the same cognitive mechanisms?

3 Languages vs. dialects: Difference Threshold and Linguistic Continuum Hypotheses

While “languages” and “dialects” exist on a typological continuum of code similarity, on the level of individual speakers, it remains plausible that a qualitative difference exists between bilingualism and bidialectalism. The field will be better positioned to understand the relationship between bilingual and bidialectal cognition with guiding hypotheses to resolve this apparent tension. We propose two hypotheses, inspired by discussions regarding the similarity of bidialectalism to bilingualism (e.g., Kirk et al., 2018; Melinger, 2018, 2021, 2026).

3.1 Qualitative difference: the Difference Threshold Hypothesis

One possibility is that the distinction between “languages” and “dialects” is psychologically real, meaning that there are qualitative differences between speaking more than one language vs. more than one dialect. We call this the Difference Threshold Hypothesis. Under the Difference Threshold Hypothesis, there is some level of similarity at which two slightly different codes (“dialects”) function cognitively as one variable system. Sufficiently different codes (“languages”) have distinct representations.

Usage-based evidence appears to support this claim. “Bidialectalism” is often defined by different relative usage rates of particular linguistic features of each dialect, rather than categorical distinctions between different forms. A short passage from a given “bidialectal” speaker will often feature multiple variants associated with each dialect, without a clear boundary between dialects (Labov, 1972; see discussion in Muysken, 2000). For this reason, Labov (1998) described African American English and Standard American English as “co-dependent systems” in the minds of speakers, and Hazen (2001) concluded that there was no evidence for bidialectal speakers who maintain two distinct systems. This differs from bilingualism, where the common assumption is that each language is represented as a distinct system (but see Section 3.2.1).

There is also experimental evidence for a distinction between bidialectal and bilingual lexical processing: (Melinger 2018, 2021, 2026) instructed participants to name pictures in one dialect while task-irrelevant words were presented in another dialect. She observed a translation equivalent interference effect for dialect pairs; this is qualitatively different from language pairs, where translation equivalent priming is typically observed (see review in Hall, 2011). Heggdal Lønes et al. (2023) and Melinger (2026) explain this difference in terms of differing lexical organization, where languages are distinguished by language tags, while dialect differences are represented at the conceptual level.

Why would such a qualitative difference between languages and dialects emerge? One possibility is that additional cognitive resources are needed to inhibit material with a high degree of overlap (see discussion in Oschwald et al., 2018), rendering bidialectal control more resource-intensive than bilingual control. In addition, due to mutual intelligibility of dialects but not of languages (in most cases), a Dialect A-dominant speaker can be understood by a listener who only speaks Dialect B, meaning the speaker has limited communicative incentive to produce Dialect B. In other words, the cognitive demands for dialect separation are elevated compared to language separation, while the communicative demands to maintain separate systems are reduced.

Traditional terminology might create some confusion in evaluating the Difference Threshold Hypothesis. First, given the imprecision in colloquial usage of dialect/language terminology, if such a threshold exists, it is likely that some pairs that have typically been called “dialects” would fall on the “language” side of the threshold and vice versa. Second, linguistic similarity is probably a more appropriate construct than “mutual intelligibility” (see Melinger, 2026 for discussion of Neapolitan and Standard Italian, for example). To a bicodal speaker, both codes are intelligible.

3.2 Quantitative difference: the Linguistic Continuum Hypothesis

While the evidence presented above suggests that meaningful differences can exist in the cognition of certain dialect pairs vs. language pairs, other evidence suggests that the descriptive pattern of a gradient distinction between languages and dialects is also reflected in the minds of bicodal speakers. Under the Linguistic Continuum Hypothesis, “bilingual” and “bidialectal” speakers both have multiple, interactive codes represented in their minds, with a range of overlap and interaction between the two.

3.2.1 Separate “dialects,” integrated “languages”

Contrary to the Difference Threshold Hypothesis, there are apparent exceptions to both the characterization of dialects as more integrated and languages as more separate.

For “dialect” contrasts, in cases of diglossia, such as Swiss/High German in Switzerland, the two codes are kept separate depending on context (Vorwerg et al., 2019). Similar to switch costs in naming lexical items when switching between languages, switch costs have been partially found for bidialectal Swiss/Standard German speakers (Vorwerg et al., 2019) as well as German/Öcher Platt and Dundonian Scots/English speakers (Kirk et al., 2018). Both studies infer similar mechanisms for bilingual and bidialectal cognition based on this evidence.

More importantly, there is reason to doubt the idea that language pairs are consistently separate. Much bilingualism research has demonstrated that the two languages mechanistically interact in the mind (for review, see Van Hell, 2023), broadly lending credence to this idea. At the lexical level, both languages are activated when processing words in only one language (for review, see Van Hell and Tanner, 2012). At the phonological level, non-selective phonological priming has been observed between languages, across different experimental paradigms, suggesting shared representations between languages (e.g., Zhang et al., 2025). Interaction between phonological systems is also evident in naturalistic code-switches: using bilingual English-Spanish corpus data, Fricke et al. (2016) found cross-language influences on voice onset time (a consonant contrast that differs between the languages) prior to a lexical code-switch.

At the morphosyntactic level, evidence for shared syntax between languages has accumulated through syntactic priming paradigms (e.g., Hartsuiker et al., 2004, 2016; Jackson, 2018; Van Gompel and Arai, 2018). While syntactic priming has been observed between typologically different languages (e.g., Korean and English; Shin and Christianson, 2009), the degree of cross-language overlap affects the magnitude of syntactic priming (Van Gompel and Arai, 2018). This suggests that a continuum of shared morphosyntax exists across language pairs.

Researchers have typically interpreted these findings as reflecting the interaction between two distinct language systems, with models including nodes for each language. However, language nodes are absent in some computational models of bilingualism (e.g., SOMBIB, Li and Farkas, 2002) that propose dynamic, emergent lexical organizations. Moreover, there are cases of dense code-switching where multiple language switches occur within a clause. Muysken (2000) terms this congruent lexicalization, and in extreme cases, “There is one grammatical system into which alternative elements from several varieties are inserted” (p. 135)—even for unrelated languages such as Sranan and Dutch.

Cognitive models of code-switching postulate that dense code-switching involves minimal levels of language separation and inhibition (e.g., Green and Wei, 2014). This suggests that some cases of bilingualism align with the continuum of codal usage that has frequently been described for dialect pairs. Indeed, proponents of translanguaging have argued that bilinguals' “languages” are not separate (Otheguy, 2025).

3.2.2 Cognition of variation in context

Why do speakers' codes vary in their degree of integration? A promising family of explanations relates to the demands of the social context. Understanding the complexity of language in social contexts has been the defining aim of sociolinguistics, and studies have reliably found that speakers adjust the linguistic variants they use depending on their interlocutor, the situation, and the topic at hand (e.g., Rickford and McNair-Knox, 1994). While there has been less work on the linguistic cognition of dialect differences, preliminary evidence suggests that responses to linguistic patterns depend on listeners' backgrounds and social properties of the speaker—at intersections of the lexical and phonological (Casasanto, 2010) and the morphosyntactic (Sandstedt et al., 2025; Maher et al., in press) levels.

A growing body of research on bilingualism focuses on interactional contexts and examines how bilinguals adapt cognitive control and language processing to situational factors. A guiding model in this area is the Adaptive Control model (Green and Abutalebi, 2013) proposing that language control processes adapt to the demands imposed by the interactional context (see also the Systems Framework approach proposed by Titone and Tiv, 2023). This has been corroborated experimentally. For example, Kaan et al. (2020) observed attenuated neural signatures of switch costs in the presence of another bilingual, and Beatty-Martínez et al. (2021) found that community code-switching practices modulate attention to code-switched speech.

This is a promising area of future research where language and dialect pairs have shared dimensions of variability, such as switching/mixing norms and social prestige, and interactions among these dimensions might resolve conflicting evidence for the hypotheses we have proposed. Crucially, these dimensions of variability could offer ways to elaborate the Linguistic Continuum Hypothesis, accounting for apparent dialect/language distinctions. Bilingualism and bidialectalism might differ based on common correlations among switching norms, prestige, and relatedness, rather than levels of difference between codes.

4 Discussion

The relationship between “dialect” and “language” differences has long been an area of curiosity, and we hope that the presented hypotheses provide a path to integrating research on dialect variation and bilingualism. Considerable research remains to empirically distinguish between the Difference Threshold Hypothesis and Linguistic Continuum Hypothesis. Since “bilingualism” and “bidialectalism” are both heterogeneous categories themselves, each individual study places a new data point across multiple axes of code relatedness, separation, social dynamics, and task demands.

We propose three future directions to fill current gaps in the literature. First, current models of bicodalism are constrained by a focus on the lexical level. This is especially problematic in understanding bidialectal cognition, since the majority of lexical items are shared. Future work should give greater consideration to the phonological and morphosyntactic levels. Second, the phenomenon of congruent lexicalization (Muysken, 2000) should be elaborated as a cognitive theory; to determine whether bilingualism and bidialectalism are qualitatively different phenomena, we need theories of the cases that are most similar. Finally, more work is needed outside of a European context, especially for closely-related codes. While there has been some research on Chinese “dialects,” for example (e.g., Chen and Zhou, 2022), these are often not situations of mutual intelligibility, making them especially interesting cases on the border between bilingualism and bidialectalism.

Language scientists are now positioned to more comprehensively consider the different ways that the mind implements linguistic variability, combining the rich accounts from sociolinguistics on dialect variation in context with advances in insights on the cognition of bilingualism. We hope that the hypotheses and proposed paths to combine research on bilingualism and bidialectalism will enhance inquiry in this area.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Matthew W. Crocker, Saarland University, Germany

Reviewed by: Mehrgol Tiv, McGill University, Canada

Footnotes

We focus on cross-code cognition at different linguistic levels and do not discuss the larger literature on how bilingual/bidialectal experiences may impact executive functions (see, e.g., Antoniou and Spanoudis, 2025; Oschwald et al., 2018).