PHONETICS: The TRACE model is an interactive model of auditory word recognition, meaning that the model allows an effect of sentence context at any moment during word recognition. In the connectionist TRACE model, word elements are represented by nodes in a network. These word nodes may be positively affected by excitatory connections with both lower (feature, phonemic) and higher (sentential) levels of representation. In this way, the likelihood of recognising a contextually appropriate word is influenced. Each cell in the network has a variable activation level and a threshold determining at which level of activation the cell starts to influence other cells: this influence may be either positive or negative (i.e. inhibitory connections). Selection of a word is defined in terms of competition between activated word nodes. The word node which in the end dominates all others will be recognised. The flow of information through the network is relatively slow. Therefore, effects of sentence context may only emerge later during processing, for instance during selection. Due to this delayed context effect, the model assigns somewhat more importance to the acoustic information than the logogen model.
|LIT.||McClelland, J.L. & Elman, J.L. (1986)|