It is now standard in the psychological literature to assume that the functional architecture for the system involved in spelling a word from memory uses two routes, a phonological route and a lexically based route. We describe a modular connectionist model based on this dual route architecture. Both routes in the model, tested in isolation, are able to simulate important aspects of the relevant psychological data. Some progress has been made towards combining the two routes into a single system. In attempting a coherent connectionist account, however, we are forced to address from first principles the difficult problem of the synchronisation and integration of information from each route into an output which combines the capabilities of both. We believe that the interactions between cognitive modules may be more difficult to model than the modules themselves, and that connectionist approaches, by forcing these interactions to be addressed at a basic level, may help to focus attention on difficult problems of psychological modelling which might otherwise not be addressed.
Interactions between knowledge sources in a dual-route connectionist model of spelling / Glasspool, Dw; Houghton, G; Shallice, Timothy. - (1995), pp. 209-226.
Interactions between knowledge sources in a dual-route connectionist model of spelling
Shallice, Timothy
1995-01-01
Abstract
It is now standard in the psychological literature to assume that the functional architecture for the system involved in spelling a word from memory uses two routes, a phonological route and a lexically based route. We describe a modular connectionist model based on this dual route architecture. Both routes in the model, tested in isolation, are able to simulate important aspects of the relevant psychological data. Some progress has been made towards combining the two routes into a single system. In attempting a coherent connectionist account, however, we are forced to address from first principles the difficult problem of the synchronisation and integration of information from each route into an output which combines the capabilities of both. We believe that the interactions between cognitive modules may be more difficult to model than the modules themselves, and that connectionist approaches, by forcing these interactions to be addressed at a basic level, may help to focus attention on difficult problems of psychological modelling which might otherwise not be addressed.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.