Models of relevant cue integration in name retrieval

Semantic features have different levels of importance in indexing a target concept. The article proposes that semantic relevance, an algorithmically derived measure based on concept descriptions, may efficiently capture the relative importance of different semantic features. Three models of how semantic features are integrated in terms of relevance during nameretrieval are presented. The models have been contrasted with empirical results from a naming-to-description task administered to three different groups of participants: young people, healthy elderly and semantically impaired Alzheimer patients. Predictions of the empirical results made by the models are used to provide a measure of identifiability or the extent to which the models can be distinguished from one another. In three studies we show that an additive-type rule is consistently superior to multiplicative rule and winner-take-all rule in predicting naming accuracy in a naming-to-description task. Finally, we investigated the implications of the integration rules for degraded semantic knowledge.