The role of inhibition and excitation in visual attention

Abstract

The natural world contains more information than our brain has the capacity to process. We make use of selective attention to filter this information ‐ enhance the relevant and suppress the distracting. While the behavioral improvements due to attention are easy to quantify experimentally, the neural circuits that give rise to these improvements remain unclear. In the visual system, modulation of neural activity due to attention is ubiquitous in measurements ranging from scalp electroencephalography to extracellular recordings from single neurons. A number of recent models of attention have proposed that inhibition may play a key role in modulating the computational abilities of neural networks at the locus of attention. While the fundamental importance of inhibition in the nervous system has been known since Sherrington's work at the turn of the 20th century, the mechanisms by which inhibition and excitation interact in service of neural computation are not well understood. Computations in cortex take place in networks of noisy neurons, and modulation of that noise likely plays an important role in attention. Recent models of cortical networks, employing balanced excitation and inhibition, are able to control the level of noise in the system. One prediction from such a model is that the attended state may be mediated be the activity of inhibitory neurons which serve to both modulate the noise in the network while simultaneously increasing response gain. While such a model is easy to implement, it is difficult to test in awake, behaving animals because extracellular recordings are blind to the cell class of the recorded neuron. We developed a novel method of probabilistically classifying neurons based on multiple aspects of their waveform shape that correspond to known differences between cell classes. Using this method, we were able to achieve a high degree of certainty of the classification of 95% of recorded V4 neurons in two rhesus macaque monkeys performing a demanding spatial attention task. We found that inhibitory neurons had stronger effects of attention than excitatory neurons in three key measures: (1) firing rate modulation; (2) decorrelation of their pairwise spiking response; and (3) the match of their response profile to the task's temporal demands. These results suggest that inhibitory neurons are indeed the recipients of a shared signal during the attentive state, and that this signal leads to a decrease in correlation in the rest of the population. Our results thus provide a network‐level account for the role of inhibition in attention.Support or Funding InformationNIH Grants EY023456, EY018894, EY022928, EY008098 and DA023428