Modelling German Word Stress

Standard linguistic and psycholinguistic approaches to stress assignment argue that the position of word is determined on basis abstract information such as syllable weight number syllables in word. In present study, we contrasted this approach with a perspective based learning analogies according which speakers learn associate basic form cues position. To do so, use simple two-layer neural network trained an error-driven mechanism predict German morphologically complex words. We find networks forms outperformed represent information. Moreover, most standard assign from right left. tested proposal found words, yielded better results than left, supporting approach. By contrast, words left right. discuss implications our for theories by taking into account cues, assigning direction, representation mental lexicon.