Neurophysiological investigation of the lateral prefrontal cortex during the task of binocular flash suppression

Abstract

Multistable visual phenomena, wherein unchanging sensory input elicits in an observer, perceptual fluctuations, have been instrumental in unravelling the neural correlates of conscious perception. Such paradigms, when combined with single unit recordings in macaques trained to report their perception, have allowed neurophysiologists to elucidate, if the cells in various regions of the brain are correlated with subjective experience or respond to the invariant retinal input. Results obtained from such an approach has so far revealed that the proportion of feature selective cells which fire in concordance with perception, increase as one progresses in the ventral visual pathway, with this fraction being up to 90% in the temporal lobe. The next station in the ventral stream of vision is the lateral prefrontal cortex (LPFC), which has reciprocal connectivity with the inferotemporal cortex and displays responses which are selective for complex visual stimuli. However, it’s not clear if this feature selective neural activity is just the result of sensory input or is related to subjective perception. Utilizing the task of binocular flash suppression (BFS), a psychophysical paradigm capable of dissociating perception from the retinal message, we probed the neural responses in the LPFC. The results revealed a robust perceptual modulation of both the spiking activity as well as high frequency gamma oscillations in this region of the brain. Even though single unit activity is robustly modulated according to perceptual content, a measure of effective functional connectivity between pairs of neurons, such as correlated variability could be revealing of interactions among neuronal populations during visual ambiguity. We therefore computed the spike count correlations across pairs of simultaneously recorded neurons during subjective visual perception. Interestingly, such interneuronal correlations among single units which preferred the same stimulus were close to zero during incongruent visual input, thus reflecting a modulation of the correlation structure during visual perception. Simulations with biophysically realistic networks suggested that the source of decorrelation was an active suppression of input fluctuations. This suggests that such a decorrelated state might be critical for representation of conscious content during visual conflict. These results together provide credence to the ‘frontal lobe hypothesis’ proposed by Crick and Koch, which suggested that the planning stages of the brain must have explicit access to the conscious visual percept so as to direct motor output. Such access is essential, if the LPFC needs to carry out one of its major function which is of cognitive control. Interestingly, when a control related signal, namely the modulation pattern of the beta band oscillations in the LPFC was analyzed, its modulation pattern was unchanged not only across monocular and incongruent visual stimulation but also during perceptual dominance and suppression. This suggests that a signal which is related to control processes is unaffected by local conscious or unconscious neural processing. Lastly, we observed an enormous diversity among the patterns of single unit activity recorded in the LPFC and the neurons which displayed visual preference were just a minority. In order to elucidate, if there were any other patterns of activity which were related to the task, we clustered the neuronal responses using a non-negative matrix factorization (NNMF) method. This revealed five sequential dominant response patterns (or components) whose peaks were temporally distributed across various phases of the trial. A majority of the units with firing profiles similar to the patterns obtained, maintained their responses across monocular or incongruent stimulation suggesting that visual conflict did not affect their spiking modulation. Interestingly, an assessment of the effective functional connectivity across the pairs of neurons belonging to different temporally distributed components revealed that such correlated variability was maximum among units which were temporally coincident. However, we observed successive decorrelation as the pairs of units were chosen from temporally separated populations. This suggests a computational principle mediating a representation of sequential patterns of activity in the LPFC. Together, the results presented in this thesis suggest a role for the LPFC in representation of conscious content. At the same time, we find that such a role of this region is coexistent with other major functions typically attributed to this area, such as cognitive control or temporal encoding of task events through sequential neural activity.