Using data on the pattern of occurrence of words embedded in the polysyllabic words of the English vocabulary, argues that recognition and segmentation of continuous speech appear to be based on competition between lexical hypotheses. It is concluded that lexical competition is an essential component of models of continuous speech recognition. (53…

This paper presents a discussion of the constraints imposed on lexicalisation during production by language-specific patterns, such as whether words exist in a language to describe a given event and whether language-specific syntactic and phonological information correlates with semantic properties. First, we introduce in broad strokes relevant…

Describes a connectionist model, using a simple recurrent network trained on a phoneme prediction task, that accounts for the child's ability to identify word boundaries. The model shows that aspects of linguistic structure that are not overtly marked in the input can be derived by efficiently combining multiple probabilistic cues. (Author/MSE)

Proposes an alternative account of the child's learning of nouns and adjectives that relies on properties of the semantic categories to be learned and of the word-learning task itself. In five experiments, a simple connectionist network was trained to label input objects in particular contexts; the network learned categories resembling nouns…

Proposes that an important role for connectionist research in language acquisition is analyzing what linguistic information is present in the child's input. Recent connectionist and statistical work analyzing the properties of real language corpora suggest a priori objections against the utility of distributional information are misguided. This…