Most of the readings on this topic seem to consider the concept of a “Jennifer Aniston” cell to be extraordinary. Perhaps that’s why the grandmother cell concept was quickly rejected in the 1960s when first proposed by Lettvin. But it’s not really a single neuron doing the decision-making; rather, that cell is simply providing the output value of a neural net comprised of many neurons. These other cells process the visual image and funnel their results into one neuron, which then determines whether or not to fire based on the neurotransmitters it has received as inputs. It makes sense that, for a face which we see often, such as a grandmother or Jennifer Aniston (for Friends junkies), the brain would develop an efficient method of recognizing that image.

Less-technical references are available at New Scientist, and Science Week.