In dimensional models, similarity between test item and concept representation
has been measured by an additive combination of differences on component
dimensions. This notion of additivity is rejected by the context model. Accord-
ing to the present model, computing the similarity between test instances and
exemplar involvesmultiplyingdifferences along component dimensions.
This process is illustrated in figure 12.4. The top half repeats some repre-
sentations given in the previous figure. Associated with each dimensional dif-
ference is a similarity parameter,ai, with high values indicating high similarity.
ThusaNLis a measure of the similarity between a long and a short nose. Two
factors can decrease the size of each parameter, that is, decrease the similarity
between the values of a dimension. One factor is the psychophysical difference
between the two values of a dimension; the other is the salience of the dimen-
sion, which is itself determined by the attentional and strategy considerations
Figure 12.4
How the context model computes similarity.
286 Edward E. Smith and Douglas L. Medin