(a) (Left) When the internal model is tuned to 0 ms as in natural situations, the probability distribution for the occurrence of touch on the left index finger (approximated as a normal distribution (Bays et al., 2005) based on the uncertainty in the predicted time of the touch arising from noise in motor and sensory systems) peaks at 0 ms after the movement of the right index finger. Touch presented at 0 ms shows the strongest attenuation, while touch at 100 ms is less attenuated because it is less likely to have been self-generated. (Middle) When exposed to systematic delays of 100 ms between the finger movement and the touch, the model parameter is gradually updated to 100 ms, which can be viewed as a simple incremental shift in the probability distribution by 100 ms. Before the exposure, there is an error associated with the touch predicted at 0 ms and presented at 100 ms but no error related to the naturally presented touch at 0 ms. During the learning period, this pattern gradually reverses: a prediction error for the touch presented at 0 ms appears and grows over exposure time, while the prediction error for the touch at 100 ms decays and reaches a minimum. (Right) After prolonged exposure, the touch at 0 ms has low probability, produces a large prediction error and will not be attenuated, whereas the touch at 100 ms has high probability, produces no prediction error and will be attenuated. (b) Participants were instructed to use their right index finger to tap a sensor (active tap) that delivered a tap on their left index finger (test tap). (c) In the exposure trials, participants simply tapped the sensor with their right index finger (RF) and received the tap on the left index finger (LF) with a 0 ms or a 100 ms exposure delay (intrinsic delay of the force setup ≅ 36 ms). In the response trials, participants received a second tap on their left index finger (comparison tap) and were required to indicate which tap was stronger: the test or the comparison tap. The test tap could be presented with a test delay of either 0 ms or 100 ms. (d) Psychophysical data from a representative participant demonstrate how the somatosensory attenuation phenomenon is quantified. The horizontal gray dashed line indicates the 50% point of psychometric functions, and the vertical gray dashed line indicates the true intensity of the test tap (2 N).