Cognition as a Window into Neuronal Population Space

Abstract

Understanding how cognitive processes affect the responses of sensory neurons may clarify the relationship between neuronal population activity and behavior. However, tools for analyzing neuronal activity have not kept up with technological advances in recording from large neuronal populations. Here, we describe prevalent hypotheses of how cognitive processes affect sensory neurons, driven largely by a model based on the activity of single neurons or pools of neurons as the units of computation. We then use simple simulations to expand this model to a new conceptual framework that focuses on subspaces of population activity as the relevant units of computation, uses comparisons between brain areas or to behavior to guide analyses of these subspaces, and suggests that population activity is optimized to decode the large variety of stimuli and tasks that animals encounter in natural behavior. This framework provides new ways of understanding the ever-growing quantity of recorded population activity data.

Low rank mechanisms underlying flexible visual representations

“…Our results are consistent with only two possibilities: 1) that the modulations in visual cortex associated with these processes do not affect perception, or 2) that readout is suboptimal in a particular way, such that the first mode (which is associated with the changes in shared variability) plays a larger role than it would in an optimal readout scheme. While the first possibility has been proposed for attention (in favor of an alternative model in which attention is mediated through changes in the visual information that is communicated to downstream areas involved in decision-making; (89)(90)(91)(92), it seems implausible for a low level process like contrast. Contrast is associated with changes from the earliest stages of visual processing, which are then communicated to visual cortex.…”

Simultaneous multi-area recordings suggest that attention improves performance by reshaping stimulus representations

“…This suggests that the 152 effects of attention on the stimulus information that can be decoded from small neuronal 153 populations like the ones we recorded are likely to be even more minimal for larger populations, 154 making it seem unlikely that attention-related improvements in information coding account for 155 the robust improvements in behavioral performance that we observed. 23,30 . However, we can infer their stability 200 by measuring the stimulus information gleaned by 201 each decoder using weights from the opposite attention condition from which they were 202 calculated (see Methods).…”

“…The idea that changing correlated variability better aligns sensory responses to a fixed readout is 357 also consistent with our recent observation that in the change detection task, monkeys' choices 358 are well-aligned with the axis in population space that explains the most correlated noise 21 . One 359 exciting possibility is that the correlated variability axis represents the fixed readout dimension, 360 perhaps because it is well-positioned to decode the motion direction of the broad set of stimuli 361 that animals encounter outside the limited environment of most laboratory tasks 23 . If so, 362 reducing noise correlations and increasing firing rate gains would improve the stimulus 363 information projected along that readout axis (following the intuitions in 43 ).…”

“…To assess the stability of the weights for each decoder in the two attention conditions, we 569 assessed the stimulus information gleaned by each decoder using the sensory responses from one 570 attention conditions and the weights calculated from the other (Figures 3 and 4). Because the 571 responses of visual neurons are non-unique and because our behavioral response is binary, the 572 weights found with our linear decoding methods are non-unique 23,30 . It is therefore not 573 informative to make direct comparisons of the weights across conditions.…”

Internally generated population activity in cortical networks hinders information transmission

“…Recording from neuronal populations while simultaneously monitoring an animal’s behavior during a structured task offers a glimpse into how neuronal activity supports computation. Correlations between spike counts from pairs of neurons in response to repeated stimulus, often referred to as noise correlations, are modulated by a variety of cognitive factors that are known to affect task performance 4 . For example, noise correlations decrease with animal arousal 12,13 or task engagement 14 .…”