A modified version of A. Tversky's contrast model of similarity ("Features of Similarity," Psychological Review, 1977, 84, 327-352) is proposed as the basis of an account of metaphorical similarity statements. This approach, called the "salience imbalance" model (Ortony, A., "Beyond Literal Similarity," Psychological Review, 1979, 86, 161-180), predicts the radical asymmetries both in meaning and in judged similarity that are particularly evident in metaphorical comparisons. Some experiments testing the predictions of the salience imbalance model are reviewed. Of particular interest is the prediction that in metaphorical comparisons the salience of those properties that are shaped by both terms is judged to be greater with respect to the second term in the comparison than with respect to the first. Using the salience imbalance model as a starting point, an analysis of the pragmatic and knowledge representation needs of metaphor comprehension is presented, and the various components are considered from a developmental perspective. Results of some experiments (N not given) examining children's comprehension of literal and metaphorical lang are then presented to support the claim that the limitations on comprehension of metaphorical lang are not metaphor-specific limitations, but rather are due to the lack of appropriate world knowledge, the absence of sufficiently rich and differentiated representations of such world knowledge, or information processing overload. Finally, the implications of the proposed model are discussed in terms of the constraints that they impose on the kinds of processing mechanisms that are likely to be needed for a process model of metaphor comprehension. (LLBA, Modified AA, Accession Number 8704008)