Cognitive approaches to uniformity and variability in morphology

1 The uniformity-variability symbiosis

All languages achieve a balance between uniform patterns and localized variations, and this dynamic is especially pronounced in morphological systems. In numerous grammatical frameworks, this balance is often viewed as a dichotomy: broad, general patterns adhere to central principles, while deviations are relegated as exceptions, anomalies, or minor divergences, not central to the linguistic structure. This approach conflicts with the common occurrence of irregular forms and the historical evidence where deviations from uniformity not only persist but expand (Maiden 2011a, 2011b). As a consequence, system-internal principles and constraints that apply to core patterns, as in the case of the one form – one meaning principle (Anttila 1977) or the paradigm economy principle (Carstairs 1983), are intrinsically limited in scope. In the best case, they encapsulate common or salient patterns or generalizations. In the worst case, they express typological biases or theoretical idealizations. But in virtually all cases, they overfit a subclass of patterns in ways that cannot be extended to a description of the system as a whole.

A recognition of the limitations of system-internal principles is implicit in the identification of external factors as the locus of explanation in the domain of morphology. The factors that impact language learning and usage, including general learning biases and processing constraints, appear to play a significant role in shaping grammatical structures, often exerting opposite influences (Blevins et al. 2017; Marzi et al. 2019). Learning generally favours uniform patterns and transparent co-relations, which facilitate generalisation and extrapolation (cf., Divjak et al. 2023; Ellis 2006; Romain et al. 2022), whereas usage-driven processing constraints are most acute for general patterns that define a dense space of potential alternatives, conferring advantages on form-meaning units that are unique, with more pronounced schematicity, entrenchment, and conventionality (cf., Booij 2017; Langacker 2019). The influence of these competing factors will be expected to vary across languages, modulated by the inertia and attraction associated with established patterns, including the transparency of the inherited word stock, the complexity of the inflectional system, and the size and semantic consistency of families of derivational forms.

Language’s dynamic nature entails trade-offs and varying equilibrium states. For instance, systems often fluctuate between facilitating learning and enhancing processing efficiency, seldom maximizing both. These trade-offs mirror broader communication system dynamics, where a perfect balance between encoding and decoding efforts, or between different communicative tasks, is unattainable. Each grammar embodies a distinct solution to these trade-offs, integrating linguistic traits with, among others, psychological, social, and historical factors. The interplay of different pressures crafts a complex array of solutions within any absolute limitations set by speech capacity – though this concept remains ambiguous. For example, as demonstrated by the uncertainty balancing reported by Filipović Đurđević and Milin (2019), there appears to be no fixed bound on complexity along either the paradigmatic or syntagmatic axis in isolation, but, instead, constraints on spikes in complexity along both axes simultaneously.

2 General goals

The choice of topics covered in this special issue is guided by three interrelated objectives. The first involves contributing to the development of a typology of splits: divergent patterns and irregularities that can exist concurrently within a single linguistic system. The second aims to delineate the scope of external factors influencing language: those pressures outside the language that shape these internal splits. The third objective is to synthesize empirical observations and theoretical insights to pinpoint which language usage factors might predispose certain structural splits over others. The inherently varied yet analysable nature of morphological systems makes them a perfect case study for addressing the set objectives, and the synthesis is intended to initiate a broader discussion of how external demands – cognitive, communicative, and others – may influence the internal organization of language.

Each of the articles within this special issue brings a unique perspective to the study of the ways that regular and irregular units coexist within linguistic systems. These diverse perspectives share several core aims with usage-based linguistics. First and foremost, they focus on analysing entire linguistic systems as they are used in real-world contexts rather than relying solely on selectively chosen anecdotal examples. They also strive to understand the underlying functions and mechanisms that sustain variation within these systems and acknowledge that elements within a language are not inherently “correct” or “wrong” in any absolute sense but make distinctive contributions as part of a broader system’s complexity and dynamics. Through these lenses, the contributions explore the intricate balance between uniformity and variability, offering insights into the nature of language structure and variation.

3 Descriptive goals

In most languages described as exhibiting inflection class morphology, inflectional paradigms and classes are typically characterized by a high degree of regularity. However, it is also common for these generally uniform systems to contain anomalies, such as gaps or unexpected forms, that deviate from the regular patterns. For these exceptional elements to preserve their place within a language, they generally need to possess unique, or at least distinctive, patterns of occurrence and co-occurrence, which resist or counterbalance the levelling forces exerted by more commonly occurring types. Despite widespread acknowledgement of these “anomalous” phenomena, available descriptions tend to be sparse, while our understanding remains incomplete, largely because it has been widely, if uncritically, assumed that deviations from a norm have no unifying properties that would be amenable to detailed analysis. This assumption has recently been challenged by research such as that by Divjak et al. (2021), which illustrates how less probable allomorphs in Polish remain learnable within their linguistic ecosystem, thereby maintaining their presence and purpose (pp. 64–66).

Few languages are characterized primarily by exceptional elements, yet the Papuan language Yele is often cited as fitting this description (cf., Henderson 1995; Levinson 2022). The conventional tendency to deem exceptions as peripheral, outside a language’s core grammar and communicative essence, reinforces the view that exceptionality must have limits. However, this preference for uniformity has limited thorough explorations into the breadth of anomalies, the conditions under which they prosper, and their possible prominence within linguistic systems. This special issue aims to bridge these gaps by examining the interplay between regularity and exceptionality across different linguistic systems. Contributions on Czech (Bermel et al.), Croatian (Hržica et al.), Estonian (Aigro and Vihman), and French (Copot and Bonami), explore the intricacies of diverse inflectional systems, particularly in areas where these systems diverge from expected norms of uniqueness and consistency. Through comprehensive analyses of these languages, the contributions provide a broader understanding of linguistic systems as multifaceted and adaptable, comprising diverse, often mismatched elements that nonetheless align along a spectrum of schematicity or conventionality (see Langacker 2019). The overarching aim is to lay the groundwork for a typology that views languages as complex and adaptive systems (cf., Beckner et al. 2009; Ellis 2016), tackling the task of scrutinising the myriad forms of irregularities and standard patterns that exist concurrently.

4 Theoretical goals

The exploration of the division of communicative labour between regular and exceptional has been approached from various theoretical angles. Usage-based models offer one fruitful point of departure for these discussions, building on the aforementioned notions of schematicity, entrenchment, and conventionality (in Langacker 2009, 2019). Then, the frequency effect, recognized as one of the most significant predictors of language behaviour, also exhibits contrasts that can serve as a unique proxy for variability or exceptionality (see Divjak 2019 for an in-depth discussion). Application of measures from information theory, including word surprisal (Hale 2003) and paradigm entropy (Ackerman and Malouf 2013), quantify how the distribution of individual word units diverge from expected linguistic patterns, offering a measure of exceptionality from a macro perspective. The use of relative entropy in the studies by Milin et al. (2009) extends this framework, addressing unpredictability within nominal paradigms and classes. This methodological approach is further refined and broadened by Filipović Đurđević and Milin (2019), which integrate the syntagmatic analysis in Hale (2003) with the paradigmatic analyses in Ackerman and Malouf (2013) and Milin et al. (2009), providing a comprehensive model of the interaction between these dimensions and enhancing our grasp of linguistic unpredictability on a global scale.

In their work “Paradigmatic predictability in derivational morphology”, Copot and Bonami investigate the predictability of newly formed lexemes derived from common bases, challenging the conventional “rooted tree” model of derivational morphology. Their empirical study with French speakers and novel word formations advocates for a paradigmatic perspective on derivation. This viewpoint suggests that words are interlinked through broad derivational schemas without needing direct ties to a singular base word, advocating for a word-and-paradigm morphology (Blevins 2016) that resonates with cognitive and related linguistic frameworks (cf., Booij 2010, 2017; Booij and Audring 2017; Langacker 1987, 2009, 2019).

5 Explanatory goals

The final objective of this special issue is to leverage empirical descriptions and theoretical insights to propose functional explanations for observed patterns of language adaptation, particularly the tension between uniformity and variability. The functional and communicative effects of different combinations of uniform patterns and exceptional items, and their implications for the resilience and adaptability of a language is a central concern of usage-based linguistic accounts in general, and cognitive linguistics in particular. It has long been recognized in cognitive psychology that human learning mechanisms, which are capable of statistical approximation and error correction, adapt to diverse inputs (as outlined by, for example, Rescorla 1988; Widrow and Lehr 1990), and to language inputs more specifically (see, for example, Arnon and Ramscar 2012; Baayen et al. 2011; Chuang and Baayen 2021; Divjak et al. 2024; Ellis 2006; Milin et al. 2023; Pirrelli et al. 2020; Ramscar et al. 2013). These processes, which maintain a balance between statistical regularities for generalization and idiosyncrasies necessitating specific processing and response, are essential for language development and usage.

The symbiotic co-evolution of language users and systems reflects this adaptive dynamic, where exceptionality, especially in morphology, is likely to be finely tuned for and by usage – i.e., learning and processing. In fact, a balance between regular structures and exceptional forms not only facilitates generalization but may also be essential for efficient linguistic performance and adaptation. A hypothetical language that is “perfect” from the standpoint of an abstract model of language design might be prone to overfitting during the learning process, with the consequence that language users would be intolerant of deviations, incapable of assimilating new linguistic patterns, and hampered by a rigid and ineffective communication system. Traditional linguistic analyses, particularly those rooted in formal grammars, have focused predominantly on uniform patterns, often at the expense of understanding the more probabilistic and variable nature of language, where certainty is not absolute but falls within a spectrum of predictability. Cognitive Linguistics offers tools for embracing the inherent systematicity present in language, even when full predictability is not achievable, allowing learners to dynamically adapt their current state of language knowledge over time.

This volume’s contributions utilize cognitive linguistic methodologies to explore these phenomena further. Aigro and Vihman in “What drives speakers’ preferences? Probing parallel form usage” investigate the impact of frequency and structural patterns in morphological choice, utilizing corpus data to demonstrate the negligible role of traditionally cited lexical-semantic features in these decisions. Meanwhile, Hržica, Košutar, Bošnjak Botica, and Milin in “The role of entrenchment and schematization in the acquisition of rich verbal morphology” take a close look into Croatian child language acquisition through parental reports, uncovering the nuanced role of type frequency over token frequency in morphological development, suggesting a different timeline for the impact of entrenchment in language learning. Bermel, Knittl, Alldrick, and Nikolaev, in “Ideal and real paradigms: Confronting evidence from grammar and corpora”, examine the disconnect between normative grammatical recommendations and actual language use, particularly in the context of defective and overabundant paradigm cells. Their findings challenge conventional understandings of entrenchment and pre-emption, advocating for expanded definitions to better explain linguistic data. These three studies collectively underscore the complexity of linguistic adaptation and the need for a multifaceted approach to understanding language as a dynamic, user-oriented system.

We hope that this special issue offers meaningful insights and plausible answers to the queries posited in this introduction, at the intersection of morphology and cognitive linguistics. Our goal is to initiate a productive dialogue that deepens our understanding of linguistic structures and processes, fostering a collaborative exploration between natural theoretical allies.

Petar Milin, Neil Bermel, James P. Blevins.

Data availability statement

Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.