Joan Bybee

In the early and mid-70's, Bybee proposed that the connection between the abstract phonological representation of a word and the actual forms experienced by language users was a more direct one than previously postulated. Her theory of Natural Generative Phonology, elaborated upon and expanded the work of Theo Vennemann, proposing less abstract mental representations of sound structure while arguing for greater proximity between phonetic and phonological forms.

Although belonging to a formalist tradition, Bybee's early work already contained elements that challenged the performance/competence model that underlay all Generative assumptions. Natural Generative Phonology proposed that the mental representation of language results from speakers’ exposure to actual language in use.[3] The proposal that the structure of language derives from actual communication rather than from abstract rules wired in the brain represented a major departure from the mainstream linguistics, an idea Bybee pursued in all her subsequent work.

In 1985, Bybee published her influential volume Morphology: A study of the Relation between Meaning and Form, in which she uncovered semantic regularities across 50 genetically and geographically diverse languages. These meaning similarities manifest themselves in recurring cross-linguistic patterns in morphological systems with respect to tense, aspect and mood. This work runs counter to Chomskyan generative theory, which describes grammar as an independent module of the brain that works in an abstract manner completely detached from semantic considerations.

Alongside linguists Dan Slobin and Carol Moder, Bybee's work helped popularize the concept of mental schemas (or schemata) to explain grammatical structure, especially in terms of connections between morphological forms within a paradigm. Bybee defines schemas as "an emergent generalization over words having similar patterns of semantic and phonological connections".[4] For instance, the English irregular verbs snuck, struck, strung, spun and hung are connected through a schema that builds on similarities between these verbs and across the lexicon: the meaning of past tense, the vowel [ʌ], the final nasal and/or (sequence of) velar consonants, as well as the initial fricative consonant /s/ or /h/.

Connections between individual forms and schemas exist in a network (see below) whose links can be strengthened, weakened and at times also severed or created. According to Bybee, the force that binds the links in a network is actual language usage.

Informed by studies child language development, morphological change and psycholinguistic experimentation, Bybee proposed in the late 1980s and early 1990s a model to account for the cognitive representation of morphologically complex words: the Network Model. Words entered in the lexicon have varying degrees of lexical strength, due primarily to their token frequency. Words with high lexical strength are easy to access, serve as the bases of morphological relations and exhibit an autonomy that makes them resistant to change and prone to semantic independence.[5]

Diachronic studies figure prominently in Bybee's body of work. Specifically, her work has explored the ways in which grammar emerges through language use via grammaticalization. Grammaticalization describes the concept that individual words or constructions may come to express abstract grammatical meaning (e.g. future tense) as users increasingly pair frequent words with a given meaning.