Speechreading: illusion or window into pattern recognition

In the Logical Model of Perception (FLMP) perceivers are conceptualized as forming perceptual judgments by evaluating and integrating multiple ambiguous sources of information, in an optimal manner based on relative goodness of match. This model has been tested favorably against a variety of competing theories and modes. Recent extensions of the FLMP are described in thin article along with empirical applications and and progress in the study of speech perception by ear and eye is reviewed within this general theoretical framework. The model illuminates the differences that are observed across different languages in terms of information as opposed to information-processing.. Pattern recognition of bimodal speech is representative of pattern recognition in a variety of other domains, such as emotion perception, and there are in-dependent reasons why multimodal presentation of audible and visible speech is particularly conducive to accurate pattern recognition. We believe that the positive outcome of this research provides a framework for the development: of computer animated agents which can serve as language tutors and as conversational characters in other domains, easing the interaction of humans and machines.