Sparseness and Entropy in Semantic Change: Precedents from Early Vision

What is the difference between description and explanation in linguistics? We take explanation to be the reduction of description to independently motivated general principles. For language, a promising source of independently motivated general principles can be found in neuroscience, and especially computational neuroscience. As an example, we consider the notion of historical change in word meaning, which in cognitive semantics is often described in terms of conceptual metaphor, metonymy and synecdoche. Why does language rely on these three mental abilities and not some others? We argue that it is because the brain depends on the computational principles of sparseness and entropy to generate efficient mental representations. These principles have been found at work in primary and secondary visual cortex, so that a proper understanding of semantic change takes us through the physiology of early visual cognition.