14. January 2020

The Anticipation of Events in Time

The anticipation of events in time

The reaction time to a stochastic event is proportional to the reciprocal of its probability (figure by Georgios Michalareas).

In music, sports, and other domains of every-day life, the brain needs to know when events occur in order to respond quickly. But how do humans anticipate such events in time? Probably the most popular hypothesis to-date has been that the brain estimates the Hazard Rate of events. Researchers from the Max Planck Institute for Empirical Aesthetics contradict this hypothesis and demostrate that the brain employs a simpler and more stable model of its environment based on the reciprocal of event probability.

How does the brain learn when an event is more probable to happen? It has been a long-standing question how the brain represents probabilities across time. The main candidate mechanism to date has been the computation of Hazard Rate, i.e., the probability that an event is about to happen given that it has not happened already. In their article in Nature Communications, Matthias Grabenhorst, Georgios Michalareas and other reasearchers demonstrate with behavioral experiments that the brain uses a much simpler computation instead—it just estimates the reciprocal of probability. This is a fundamental finding, shedding light onto a canonical principle of how the brain models probabilities. The close relationship between the reciprocal probability and Shanon information content (also termed Surprisal) hints that the brain actually represents probabilities in terms of information.

“Probability itself is the fundamental parameter used by the brain”, says lead author Matthias Grabenhorst.

A second important finding in this work concerns the uncertainty in estimation of elapsed time. Previous research has demonstrated that the longer the elapsed time, the higher the uncertainty in its estimation by the brain. Grabenhorst and his colleauges present evidence that this principle of monotonically increasing uncertainty with elapsed time does not always hold. It is actually the probability distribution of events across time which determines when uncertainty is lowest or highest.

Finally, the authors show that the above findings hold in three different modalities: vision, audition, and somatosensation. This commonality hints to either a central mechanism used by all three modalities or by a canonical peripheral mechanism employed in multiple sensory areas of the brain.


Original Publication: 
Grabenhorst, M., Michalareas, G., Maloney, L. T., & Poeppel, D. (2019). The anticipation of events in time. Nature Communications, 10(1). doi:10.1038/s41467-019-13849-0

Matthias Grabenhorst
Georgios Michalareas