A Single-mechanism Dual-route Model of German Verb Inflection

We present a constructivist neural network model of German past participle verb inflection. The model builds its architecture in response to the learning task in a way consistent with neurobiological and psychological evidence. In contrast to most single-route connectionist models with a fixed, homogeneous architecture our model (1) reaches adult performance on an extensive corpus of German verbs, (2) shows U-shaped learning curves – even at the level of individual verbs – within a static learning environment, and (3) captures verb type specific dissociations with respect to neurological impairments. In contrast to dual-route symbolic theories, the model’s emergent notion of inflectional classes is based on distributional factors in the learning environment, thus obviating the need for in-built assumptions such as specific processing mechanisms based on grammatical class. By focusing on the German participle we demonstrate that the performance of the model does not depend on the existence of a dominant ‘default’ class. Taking seriously the constructivist, experience dependent nature of brain development, we suggest that such a single-mechanism dual-route model presents a step forward in the long standing, yet unresolved, past tense debate.