ERIC Number: ED108130
Record Type: RIE
Publication Date: 1975-Apr-2
Reference Count: 0
Formal Models of Word Recognition. Final Report.
Travers, Jeffrey R.
Existing mathematical models of word recognition are reviewed and a new theory is proposed in this research. The new theory integrates earlier proposals within a single framework, sacrificing none of the predictive power of the earlier proposals, but offering a gain in theoretical economy. The theory holds that word recognition is accomplished by filtering visual feature information from the printed word through a hierarchy of letter, letter-cluster, and word detectors. The detectors are Bayesian decision devices which estimate the likelihood of the presence of their target configurations by combining information from lower detectors with a priori knowledge about the structure of words in English. In addition, several empirical studies on issues related to the theory were conducted. Two of these studies demonstrated that skilled readers draw visual information from all the letters in a word at once, rather than from one letter at a time; and that statistical co-occurance of letter sequences affects the perceptibility of those sequences, independent of their pronounceability. A third study, on whether covert pronunciation of words is necessary to apprehend their meaning, proved inconclusive. (Author/WR)
Publication Type: Reports - Research
Education Level: N/A
Sponsor: National Inst. of Education (DHEW), Washington, DC. Office of Research Grants.
Authoring Institution: Swarthmore Coll., PA.