Getting from Here to There: Limitations of Current Models of Morphological Processes in Word Recognition and Future Directions

Rueckl, J.

University of Connecticut and Haskins Laboratories

The development of explicit computational accounts of the role of morphology in word recognition has been slower than we might have hoped. I'll discuss three relevant sets of issues. First, what counts as a good theory? For example, how much intelligence are we willing to build in to our models? Second, what are the technical issues impeding progress on one class of computational models (connectionist models in particular). Third, what kinds of phenomena should the next generation of models address? I'll suggest several, including learning, cross-linguistic effects, and interaction of morphological factors with orthographic and phonological variables.