A Cohort Model of Visual Word Recognition

A model of word recognition is proposed which assumes that when a word is encountered, the first available orthographic code activates all lexical entries that are positionally consistent with that information (i.e., the word's cohort). As subsequently encoded orthographic information becomes available, activation is withdrawn from lexical entries with which it is inconsistent, and the cohort is resolved when only a single candidate remains. The model is intended to characterize: (1) The initial encoding of the graphic information; (2) the recording of that information into an abstract orthographic form; (3) the manner in which the appropriate lexical entries are then activated; (4) the process whereby subsequent orthographic encoding resolves those activated entries to a single candidate; and finally (5) the way in which subjects use information for making decisions as it emerges during this processing. The model includes two critical points. The first is that letter identity information is encoded in the form of a complex orthographic unit termed a wickelgraph. A wickelgraph includes an encoding of the target letter's identity information as its core, plus, as a context, the identity of its two immediately adjacent letters. The set of possible wickelgraphs is assumed to instantiate the rules of orthography. The second point is that once a cohort is established, the perceiver can sample its status at any point during resolution and base a decision on the outcome of that sample. Some tasks (e.g., naming) may allow for a decision based on the initial status of the cohort, while others (e.g., lexical decision) can require its complete resolution. Six experiments are reported that illustrate some of these cohort effects in lexical access.