Recurrent cortical networks encode natural sensory statistics via sequence filtering.

Recurrent neural networks can generate dynamics, but in sensory cortex it has been unclear if any dynamic processing is supported by the dense recurrent excitatory-excitatory network. Here we show a new role for recurrent connections in mouse visual cortex: they support powerful dynamical computations, but by filtering sequences of input instead of generating sequences. Using two-photon optogenetics, we measure neural responses to natural images and play them back, finding inputs are amplified when played back during the correct movie dynamic context- when the preceding sequence corresponds to natural vision. This sequence selectivity depends on a network mechanism: earlier input patterns produce responses in other local neurons, which interact with later input patterns. We confirm this mechanism by designing sequences of inputs that are amplified or suppressed by the network. These data suggest recurrent cortical connections perform predictive processing, encoding the statistics of the natural world in input-output transformations.

A unified model of the task-evoked pupil response.

The pupil dilates and reconstricts following task events. It is popular to model this task-evoked pupil response as a linear transformation of event-locked impulses, whose amplitudes are used as estimates of arousal. We show that this model is incorrect and propose an alternative model based on the physiological finding that a common neural input drives saccades and pupil size. The estimates of arousal from our model agreed with key predictions: Arousal scaled with task difficulty and behavioral performance but was invariant to small differences in trial duration. Moreover, the model offers a unified explanation for a wide range of phenomena: entrainment of pupil size and saccades to task timing, modulation of pupil response amplitude and noise with task difficulty, reaction time-dependent modulation of pupil response timing and amplitude, a constrictory pupil response time-locked to saccades, and task-dependent distortion of this saccade-locked pupil response.

Task-related hemodynamic responses in human early visual cortex are modulated by task difficulty and behavioral performance.

Early visual cortex exhibits widespread hemodynamic responses in the absence of visual stimulation, which are entrained to the timing of a task and not predicted by local spiking or local field potential. Such task-related responses (TRRs) covary with reward magnitude and physiological signatures of arousal. It is unknown, however, if TRRs change on a trial-to-trial basis according to behavioral performance and task difficulty. If so, this would suggest that TRRs reflect arousal on a trial-to-trial timescale and covary with critical task and behavioral variables. We measured functional magnetic resonance imaging blood-oxygen-level-dependent (fMRI-BOLD) responses in the early visual cortex of human observers performing an orientation discrimination task consisting of separate easy and hard runs of trials. Stimuli were presented in a small portion of one hemifield, but the fMRI response was measured in the ipsilateral hemisphere, far from the stimulus representation and focus of spatial attention. TRRs scaled in amplitude with task difficulty, behavioral accuracy, reaction time, and lapses across trials. These modulations were not explained by the influence of respiration, cardiac activity, or head movement on the fMRI signal. Similar modulations with task difficulty and behavior were observed in pupil size. These results suggest that TRRs reflect arousal and behavior on the timescale of individual trials.

Modeling other minds: Bayesian inference explains human choices in group decision-making.

To make decisions in a social context, humans have to predict the behavior of others, an ability that is thought to rely on having a model of other minds known as "theory of mind." Such a model becomes especially complex when the number of people one simultaneously interacts with is large and actions are anonymous. Here, we present results from a group decision-making task known as the volunteer's dilemma and demonstrate that a Bayesian model based on partially observable Markov decision processes outperforms existing models in quantitatively predicting human behavior and outcomes of group interactions. Our results suggest that in decision-making tasks involving large groups with anonymous members, humans use Bayesian inference to model the "mind of the group," making predictions of others' decisions while also simulating the effects of their own actions on the group's dynamics in the future.