I don't think it is a matter of a neuron's changing its "behavioral
function"; rather, changes of connections among neurons. Doubt a
single neuron has a "behavioral function" (although of course there was
the famous hyppothetical "grandmother neuron"!)(i.e. the endpoint of
converging and increasingly specific visual stream inputs en route to
the infratemporal region).
Well, it seems clear that single cortical neurons can't mediate specific functions due to their broad sensitivity profiles. This is probably true of "face neurons" also. Specific sensory percepts or movements are most likely mediated by patterns of activity in populations of neurons in distributed networks, with unique spatiotemporal patterns for each percept or movement. If so, a change in the function of a cortical area would imply a change in the meaning of specific patterns of activity in specific sets of neurons -- a pattern signalling finger 1 might begin signalling finger 2 or the hand, etc. Furthermore, such a change in meaning would probably have to occur in all areas of the interconnected distributed network, something hinted at by Merzenich. The problem is -- how could the brain know that the meaning of a particular spatiotemporal pattern has changed, especially if such changes are constantly occurring with ever-changing inputs? GSD