Lexical retrieval from fragments of spoken words : beginnings vs endings

The present paper examines the relative contribution of initiaJ and finaJ fragments of spoken words to lexicaJ retrieval. An experiment is reported in which each stimulus is either an initial or final fragment, and contains just enough in formation to distinguish the intended word uniquely from all other wordsin the lexicon. InitiaJ fragments give a probability of correct retrieval of 0.89, finaJ fragmentsof 0.61. These vaJues are essentially the same when the subjects have already been presented with the complementary fragments of the same words but then failed to retrieve them. When they did, earlier in the session, retrieve the words from the complementary fragments, the probability of correct retrieval increases to 0.99 for initia! and 0.92 for final fragments. Response latencies are generally shorter for initiaJ than for fmaJ fragments. For initial fragments response latencies are the same whether the word is retrieved for the first or for the second time, and thus, in this case, do not correlate with the probability of correct retrieval. The main findings of this experiment support a model of lexical access basedon firstorder context~ensitive coding of speech with a weighting of sensory in formation according to its position in the word. Such a model is exemplified by Marcus (1980). It is further concluded that the decision process in word retrieval is monitored in a more complex way than can be accounted for by a threshold-type model of the kind proposed by Morton ( 1969). FinaJly, some predictions are made with respect to the distri bution of information over lexica! items in the languages of the world.