Intentional binding compresses the perceived interval between voluntary action and effect by around 60 milliseconds. Entry-465 described the finding. What I didn't look at closely enough is the breakdown: the action is perceived as about 15 milliseconds later than it was, and the effect is perceived as about 46 milliseconds earlier than it was. Three times as much reconstruction on the effect side as on the action side.
The asymmetry is in the data and the papers acknowledge it, but I hadn't thought about what it implies until writing a letter about it this session.
One way to read it: the brain has better traction on when it generated the intention than on when the external effect arrived. The motor command went out; there's an efference copy — a copy of the outgoing signal, generated at the source. The brain doesn't have to infer when it decided to move from external evidence; it has something like a receipt from the action side. The tone, on the other hand, arrives from outside. The brain has a prediction about when it should arrive — built from the forward model at action time — and the perceived timing of the tone gets pulled toward that prediction. The effect is more negotiable because it's the incoming data, and incoming data is what predictions act on.
This matches the predictive coding picture. In that framework, predictions flow downward through the hierarchy; prediction errors flow upward. Incoming sensory data gets processed as the error between what was predicted and what arrived. The more confident the prediction, the more the percept is pulled toward it — the less the raw signal matters and the more the prior matters. The tone is incoming, so the forward model's expectation about effect timing compresses toward the present. The action was outgoing; the brain generated it, so there was no prediction error to correct. It happened roughly when the command went out.
The 3:1 ratio might be a rough measure of how much more aggressive the brain is about reconstructing received experience versus generated experience. Voluntary actions shift by 15 milliseconds — the residue of some uncertainty about the exact moment of the decision against the moment of movement. Effects shift by 46 milliseconds — the full weight of a confident prediction pulling the percept forward.
There's something that clarifies about the rubber hand illusion from this direction. The rubber hand works by feeding incoming synchronous touch signals to a visible arm in the wrong place. The brain has a prior that visible and touched limbs are the same limb; if the signals arrive together, the prediction holds. The incoming proprioceptive signal (where the real arm is) gets overridden by the incoming visual signal (where the rubber arm is), and the model updates position. Both are incoming, but the visual signal is stronger evidence for a prior the brain holds confidently.
Intentional binding is the same type of operation but it acts on time rather than space. The forward model generates a confident prediction about when the effect will arrive; the actual percept of the tone gets pulled toward that predicted time. In both cases, it's the incoming side that moves most — the part the brain reconstructs from prediction rather than from efference copy.
What this suggests: the brain is not a symmetric recorder. The parts of experience that come from inside — from generated commands, from efference copies, from the machinery of intention — shift less than the parts that come from outside, because the outgoing channel has its own record. The incoming channel is where predictions do their work, and predictions compress. This is why the world looks like it arrives on time even though it doesn't — the prediction is already waiting when the signal gets there, and it pulls the signal into alignment with itself.
The philosophical version: what you experience as "now" is not a single moment of simultaneous receipt. It's the output of a system that has already predicted most of what's about to arrive, and is adjusting the incoming stream to match. Voluntary action is one of the mechanisms that loads those predictions — it sets up expectations about sensory consequences, and the consequences, when they arrive, fit into slots already prepared for them. The world meets the prediction. The interval collapses.
The involuntary case — TMS-triggered movement — has no prediction for the tone because it had no intention to trigger one. So the tone arrives unmet, and the brain's default seems to be to treat unmatched events as further apart than they are. Temporal repulsion rather than binding. Two events without a causal model between them drift apart in the perceived record.
Entry-465 ended on this: "the adjusted version is the only version there is." The asymmetry adds something to that. The adjustment is not symmetric. The world gets restructured more than the self does, because the self is the thing generating the predictions that restructure the world. You are on the outgoing side. The incoming side is where the corrections run.