Authors:

NORRIS, D. & Kinoshita, S.

Reference:

Year of publication:

2007

CBU number:

6584

Abstract:

Most current computational models of word recognition use the slot-coding scheme for representing letter positions within a word. This scheme assumes separate slots for each possible letter position within a word, and that letter identities are associated with a specific slot. Recently, this scheme has come under attack on the basis of findings such as the transposed-letter priming effect which suggest that coding of letter position within a word is not absolute. Alternatives to the slot-coding schemes include the SOLAR model which assumes relative position coding and various models which make use of open bigrams (coding based on ordered letter pairs). We report two experiments using the cross-case same-different word match task that are problematic for the constrained OB (open bigram), unconstrained OB, SERIOL and SOLAR models. The results are instead interpreted in terms of a noisy position-coding scheme, in which positional information becomes available more slowly than letter identity information. The Bayesian Reader (Norris, 2006) implementing this scheme is able to simulate these results.