Predictive Coding of Novel versus Familiar Stimuli in the Primary Visual Cortex

Abstract

To explore theories of predictive coding, we presented mice with repeated sequences of images with novel images sparsely substituted. Under these conditions, mice could be rapidly trained to lick in response to a novel image, demonstrating a high level of performance on the first day of testing. Using 2-photon calcium imaging to record from layer 2/3 neurons in the primary visual cortex, we found that novel images evoked excess activity in the majority of neurons. When a new stimulus sequence was repeatedly presented, a majority of neurons had similarly elevated activity for the first few presentations, which then decayed to almost zero activity. The decay time of these transient responses was not fixed, but instead scaled with the length of the stimulus sequence. However, at the same time, we also found a small fraction of the neurons within the population (~2%) that continued to respond strongly and periodically to the repeated stimulus. Decoding analysis demonstrated that both the transient and sustained responses encoded information about stimulus identity. We conclude that the layer 2/3 population uses a two-channel predictive code: a dense transient code for novel stimuli and a sparse sustained code for familiar stimuli. These results extend and unify existing theories about the nature of predictive neural codes.

Higher performers upregulate brain signal variability in response to more feature-rich visual input

“…At least since the 1950s, various researchers have suggested that early visual regions actively work to reduce redundancy in stimulus content (e.g., via "redundancy filtering" (Van Hateren 1992, 1993). In accord with such theories, the visual cortex often exhibits rapid ensemble-level suppression to more common (or expected) stimuli or stimulus trains, and greater ensemblelevel response to more unexpected (or complex) stimuli (Hamm and Yuste 2016;Homann et al 2017;Vinken et al 2017). This has important implications for BOLD variability.…”

“…Specifically, we and others have postulated that modulation of brain signal variability could reflect differences in stimulus input (Knill and Pouget 2004;Ma et al 2006;Beck et al 2008;Garrett, Samanez-Larkin, et al 2013;Orban et al 2016). For example, early visual regions may be actively suppressed in response to more common stimuli, yet may exhibit a more dynamic response to more differentiated stimuli (Hamm and Yuste 2016;Homann et al 2017;Vinken et al 2017). However, evidence for this effect is lacking in humans, as are the associated behavioral consequences.…”

Adaptation supports short-term memory in a visual change detection task

“…In the experiments, neurons showed varying levels of adaptation (some neurons did not adapt) while in 191 the models with persistent activity, the majority of units showed sustained responses. Interestingly, in the 192 context of predictive coding, a two-channel code has been proposed, with one channel showing transient 193 responses that exhibit a similar adaptation profile in neurons across repeated stimuli, and the other channel 194 showing sustained responses in a smaller subset of the population across all image presentations [14]. It 195 remains to be seen whether these types of models can be reconciled with our current findings.…”

A Long Timescale Stimulus History Effect in the Primary Visual Cortex

“…For the purpose of imaging, we implanted a chronic cranial window above the visual areas of each mouse. We followed the same surgical procedure as outlined in Homann et al [26], adapted from Dombeck et al [27]. Prior to surgery, we administered Dexamethasone at 2 mg/kg to prevent brain swelling.…”