COHORT THEORY
A model (Marlsen-Wilson, 1987) of the way in which words are retrieved from the lexicon in listening. A listener processes an utterance at a delay of around a fifth of a second behind the speaker; this is often not enough to provide full evidence for a particular word. Cohort Theory adopts the hypothesis that the listener retrieves a set of words (a cohort) which match the evidence of the signal so far. Thus, on hearing the phonetic string [k{p] they would retrieve CAP, CAPITAL, CAPRICORN, CAPTURE, CAPTAIN, CAPTIVE, etc. as word candidates. If the next sounds proved to be [t] and [I], the cohort would narrow to CAPTAIN, CAPTIVE and CAPTIVATE. Finally, the sound [n] would mark a uniqueness point, where only one word match, CAPTAIN, was possible. The uniqueness point of a word is not necessarily its last phoneme: for example, the word PSYCHOLOGY becomes unique at the /l/.
A major objection to the early version of Cohort Theory was that it was heavily dependent upon correct identification of word-initial phonemes. If a slip of the tongue led a speaker to produce ‘shigarette’ rather than cigarette, the appropriate cohort would not be selected. In addition, account needs to be taken of phonetic accommodation, which leads many words to diverge from their citation forms when they occur in connected speech. The model was therefore revised to include a principle of closeness of fit rather than exact match. An activation dimension was added: the cohort is now represented as a set of lexical items whose strength is boosted or weakened by incoming perceptual evidence until one of them achieves a match. Contextual evidence can play a role in narrowing down the cohort; but it is not taken into account until after about 150–200 milliseconds of the word; this is referred to as bottom-up priority.
A weakness of the model remains the fact that it assumes that a word is a discrete and easily identified unit. Many sequences that appear to constitute monosyllabic words may prove instead to be the initial syllables of polysyllabic ones, and vice versa. Using the ‘captain’ example, how is one to know if a match has been achieved after the word CAP or if one has to continue reducing the cohort? There may also be ambiguities of lexical segmentation where word candidates cross boundaries (for example, the word SISTER occurs within the sequence insist upon).
The ‘uniqueness’ concept has also been challenged. A study of the lexicon has suggested that over a third of words in normal speech are not unique by their offsets, while there is evidence that many monosyllabic words are not identified until well after their offsets.
Nevertheless, the notion of the cohort is implicitly or explicitly adopted in many accounts of auditory processing.
See also: Lexical access, Lexical segmentation, Speech perception, Uniqueness point
Further reading: Marslen-Wilson (1987)
