Investigations of how the brain processes language have been concerned almost entirely with localizing linguistic functions in the cortex, ignoring the more interesting questions of what is going on in those locations and how linguistic information is
represented. The reason is that questions of localization are easier to study, through aphasiology and brain imaging. But there is abundant evidence, both linguistic and neurological, that reveals answers to the more basic questions. The evidence strongly supports a connectionist view of linguistic information rather than one in which the brain stores symbols as such. Symbolic accounts are incompatible with neuroanatomical evidence and require impossible assumptions about brain function. In contrast, relational network theory, a version of connectionism, provides an account of the operation of the brain that is consistent with numerous details of cortical anatomy and function, while also being consistent with quantitative estimates of cortical capacity.