Cortical layers tend to be more of a way of dividing a specific low-level task up into sub-functions of information processing. For example, they do different networking/interconnect jobs in different layers and as one would expect different areas of the neocortex have varying layer counts (ranging from 1 to 6).

A more comprehensive computational unit of the cortex is actually the neocortical column which is a "horizontal" patch of neocortex. This is a functional unit that is probably very close to a microprocessor in analogy in that it contains a set of variables, some processing rules, input and output wires, and the capability of executing functions. A whole patch of cortical columns forms a sort of processing center, and interestingly enough, those tend to be specialized for certain kinds of information processing tasks. That's why in an fMRI scan whole areas of the cortex tend to light up together, and it's often the same areas geographically even if you scan a lot of people.

So if we're talking about a theory of localized neocortical involvement, my first impulse would be to check out those specialized areas for (non-)activity - something that led to pretty good empirical results before in related studies - as opposed to using the layers for that hypothesis.