Of Rubber Hands and Auto-Association

This is, by no means, a newly-discovered phenomenon. On the contrary, it has been observed as a sort of dime-store parlor trick by countless neurology students for quite some time. It does, however, serve for me as the most recent addition to a growing body of examples for which predictive-associative memory very elegantly accounts for widely-varying neuro-psychological effects.

To elaborate, the high-level "map" of one's body in 3-dimensional space is a construction of associated sensory modalities: proprioceptive inputs, nociception, even vestibular sensation. Most importantly here, however, is that visual cues are also a major component. "Invariant representations," for those uninitiated in the term, are those sets of neurons (or respective columns) which, by means of excitatory feedback, have primed to fire together based on previously-experienced patterns. That is to say, having fired together before, they have since "wired" together.

More precisely, the invariant representation of the stimulus is the next-higher-level unit upon which the set converges: it casts the final vote as to whether the represented stimulus has been satisfied (it then, of course, functions as a component of the next-level invariant representation).

One notable feature of an invariant representation is that the entire set (or sequence, when discussing temporally-relevant representations) need not be directly activated to provoke a response from the group; feedback from above primes the expected pattern, which, if satisfied enough to activate the point of convergence, "speaks" for all of the inputs. In this way, a pattern may "auto-associate" to itself.

With enough of the components, the (very) high-level invariant representation for "my left hand" may be activated, even if I don't have one anymore. "I know my arm is somewhere over there, I feel pain, and, by God, I see a hand where I might expect to. That must be mine!"