25 26 Sensory information is encoded by populations of cortical neurons. Yet, it is unknown how this 27 information is used for even simple perceptual choices such as discriminating orientation. To 28 determine the computation underlying this perceptual choice, we took advantage of the robust 29 adaptation in the mouse visual system. We find that adaptation increases animals' thresholds 30 for orientation discrimination. This was unexpected since optimal computations that take 31 advantage of all available sensory information predict that the shift in tuning and increase in 32 signal-to-noise ratio in the adapted condition should improve discrimination. Instead, we find 33 that the effects of adaptation on behavior can be explained by the appropriate reliance of the 34 perceptual choice circuits on target preferring neurons, but the failure to discount neurons that 35 prefer the distractor. This suggests that to solve this task the circuit has adopted a suboptimal 36 strategy that discards important task-related information to implement a feed-forward visual 37 computation. 38 Graf et al., 2011). This orientation identification strategy is attractive in that, once it is learned, 56 the same computation can be used to generalize across multiple tasks (i.e. detection and 57 discrimination). However, this computation likely requires complex circuits (with knowledge of 58 the full tuning curve of each neuron) and learning rules to act upon the combined output of a 59 diversely tuned population (Deneve et al., 1999). 60 61 Instead, when faced with perceptual choices, human and animal subjects often implement task-62 specific strategies that require less complex circuits and learning rules (Zhang et al., 2010; 63 Fulvio et al., 2014;Yu et al., 2017; Djurdjevic et al., 2018). Such a task-specific computation 64 4 might directly evaluate the identity or level of activity within distinct neuronal ensembles, 65 agnostic to the tuning of neurons in those ensembles, as in a linear classifier. By removing the 66 need for an absolute stimulus estimate, the circuits that compute perceptual choice may operate 67 faster and be more amenable to simple cellular associative learning rules. 68
69Our goal is to understand how these computational approaches are realized by biological 70 circuits, and how sensory information is integrated by these circuits to make perceptual 71 decisions. Stimulus-specific adaptation is a useful tool for evaluating how sensory information is 72 used to guide perceptual choice since it has predictable effects on neuronal activity and sensory 73 encoding (Müller et al., 1999; Dragoi et al., 2000). By sparsifying and increasing the signal-to-74 noise of neuronal population responses, stimulus-specific adaptation is expected to increase the 75 information about the presented orientation (Ulanovsky et al., 2003;Wark et al., 2007). In 76 addition, if the stimulus orientation is estimated, adaptation to the distractor is also expected to 77 improve orientation discrimination thresholds by decreasing the con...