Akrami et al. 2018

Posterior parietal cortex represents sensory history and mediates its effects on behaviour

Athena Akrami, Charles Kopec, Mathew Diamond, Carlos Brody, Nature, 2018

Athena Akrami

It’s been known for decades that prior stimulus history affects perception and memory. In spite of many psychophysical, theoretical, and neurophysiological (using human subjects and fMRI or EEG) little is known about what neural mechanisms allow prior information about the statistics of stimuli to affect perception.

A behavioral task which displays effects consistent with the use of priors in Bayesian inference is the sequential comparison of two graded stimuli separated by a delay period of a few seconds. This task requires subjects to maintain an analog value in memory, allowing quantitative assessments of how that memory is stored. We refer to this task as a “parametric working memory” (PWM) task. It has long been known that human PWM subjects show “contraction bias”, in which the estimated magnitude of memorized stimuli is systematically biased by prior stimulus statistics (Hollingworth 1910, Ashourian et al 2011). We have developed an auditory version of this task for rats, adapted from a previous tactile task (Fassihi and Akrami et. al 2014). Using a regression model, we quantify how performance on each trial depends on the stimulus and trial history, and we find that rats show a contraction bias effect, due to the recent history of sensory stimuli, remarkably similar to humans.

We carried out the first local brain inactivations during PWM, using optogenetics to bilaterally silence posterior parietal cortex (PPC), and found that, surprisingly, this led to improved performance in the task. Quantitative analysis suggested that performance improvement could be due to a reduction in the biases caused by the history of stimulus statistics. Furthermore, recordings from PPC show that neurons carry significant amount of information about the previous trials’ sensory stimuli, even while carrying no information during the delay period about the current trial’s first stimulus. These results suggest a sensory history-dependent firing rates in PPC during PWM may underlie the century-long psychophysics phenomenon known as contraction bias. Our data identifies the PPC as a critical node in conveying information about priors in the Bayesian sense.

Here’s Nature News and Views by Prof. Laura Busse on Athena’s paper: Working memory freed from the past.  

And here’s Neuron Spotlight by Prof. Miguel Maravall et al: Cortical Lifelogging: The Posterior Parietal Cortex as Sensory History Buffer