We clarified here an old statistical debate in neurophysiological data analysis: should hypothetical data be generated by randomizing the data around the original measurement (blue) or by randomizing the data in pre-determined intervals (red)? The question seems inocuous, but we showed that the wrong randomization can lead to dramatic conclusions regarding the assessed hypothesis: we can 'hallucinate' a temporal structure when there is none.
More generally, the ultra-precise temporal recording of hundred of cells' activity in the behaving animal raises stimulating statistical problems. How to do any statistics when one cannot reproduce the same exact condition twice (non-repeatability), and when one cannot observe much faster than the timescale of the system's state (non-stationarity).
We considered here the basis of shape perception by touch from a contact mechanics perspective. To better grasp the problem, you can start by doing a simple experiment: stroke the teeth of a comb back and forth with a pencil, your fingertip laying on the teeth (Hayward, 2008). You should feel a raised object moving on the finger, when, in fact, the skin is stretched laterally. We highlighted an intriguing similarity between the first stages of shape processing in touch and vision. The equations of
an elastic medium in contact with a rigid object is
all you need to exhibit this link.