How lexical ambiguity distributes activation to semantic neighbors: Some possible consequences within a computational framework

The role which the diversity of a word’s contexts plays in lexical access is currently the object of research. Vector-space models such as Latent Semantic Analysis (LSA) are useful to examine this role. Having an objective, discrete model of lexical representation allows us to objectify parameters in order to define contextual focalization in a more measurable way. In the first part of our study, we investigate whether certain empirical data on ambiguity can be modeled by means of an exclusively symbolic single representation model such as LSA and an excitatory-inhibitory mechanism such as the Construction-­Integration framework. Our observations support the idea that some ambiguity effects could be explained by the contextual distribution using such a model. In the second part, we put abstract and concrete words to the test. Our LSA model (exclusively symbolic) and the excitatory-inhibitory mechanism can also explain the penalty paid by abstract words as they activate other words through semantic similarity and the advantage of concrete words in naming and semantic judgments, though it does not account for the advantage of concrete words in lexical decision tasks. The results of this second part are then discussed within the framework of the embodied/symbolic view of the language process.