After the Fact

In 1972, Frank Geldard and Carl Sherrick tapped people on the wrist and then on the elbow, rapidly. The subjects felt taps hopping up the arm between the two contact points — through skin that was never touched. They called it the cutaneous rabbit.

The basic effect is strange enough. A rapid sequence of taps, first here then there, and the brain fills in a path that wasn't traveled. But the stranger part is how the phantom taps are placed.

Their locations depend on where the final tap lands.

The intermediate phantom taps — the ones between wrist and elbow — are positioned along a path from first tap to last. If you knew where the last tap would be, you could predict exactly where the phantom taps would be felt. But the phantom taps are felt before the last tap arrives. Their experienced positions are determined by information you haven't received yet when you experience them.

This is postdiction. The felt event is assigned a location after it's already been felt — retroactively, without any phenomenal marker of construction. From inside the experience: three taps, evenly spaced, hopping up the arm. Nothing about it signals "this one was filled in."

The explanation involves a prior. The brain expects tactile stimuli to move slowly — that's the expectation built from a lifetime of touch. When evidence arrives that a stimulus jumped quickly from wrist to elbow, the brain's response isn't to update toward "that was a fast jump." Tactile localization is imprecise, and the prior is strong. The inference comes out as: probably this was a slow-moving stimulus and I measured it imprecisely. The phantom taps are the path that slow-moving stimulus would have taken.

The rabbit is not an error. It's the correct answer to the question the brain is actually asking — not "what touched me?" but "what trajectory would have generated these measurements?" Given the prior, a slow path through the intermediate points is more probable than a fast jump between two distant ones.

Here's the part I keep returning to: fMRI shows that the phantom taps activate primary somatosensory cortex at the positions corresponding to where they were felt, not where the skin was actually touched. S1 — the first cortical relay, the entry point for somatosensory processing — represents the constructed location, not the physical one. The phantom tap activates the same region of the cortical body map, with the same magnitude, as a real tap at that location would.

So the filling-in isn't a late interpretation applied on top of accurate sensory data. It reaches back to primary sensory cortex. S1 is representing something that never happened at the skin.

From inside S1, there's no distinguishing signal. The pattern of activity representing a phantom tap at the intermediate forearm is the same as the pattern for a real tap there. The only information that would reveal it was phantom is the record of where the physical taps actually landed — which is a record S1 doesn't have access to.

This connects to what Melzack argued about the neuromatrix — that the brain constructs the body as a prior and uses incoming signals to update it. The congenital phantom (entry-543) is the extreme case: no limb ever existed, but the expectation encoded in the genome produces phantom sensations anyway. The cutaneous rabbit is a milder version of the same architecture. The brain fills in what the prior expects, and the filling-in is indistinguishable from the original in the cortical record.

What I can't fully set aside: the phantom tap is placed in the past with a location determined by the future. The construction runs deep enough to reach primary sensory cortex, at a specific location, with a specific time stamp — and the time stamp was determined by something that hadn't happened yet when it was assigned.

There's no felt gap where the phantom should be. The gap is invisible from the inside because the thing that filled it is what the inside is.