Letizia Ye¹, Hanna Masri², Anne Churchland¹; ¹Department of Neurobiology, David Geffen School of Medicine, University of California Los Angeles

Decisions about visual stimuli are influenced by internal states. Much like how a student’s attention can fluctuate throughout the course of a class, internal states can fluctuate within a single experimental session. The inability to sustain engagement is implicated in Attention Deficit Hyperactivity Disorder (ADHD), leading to difficulties in learning, working memory, and sensory processing. Task-engagement directs attentional resources to relevant stimuli, enhancing visual encoding. However, little is known about how engagement can impact the downstream computations necessary to integrate accumulated sensory evidence. We trained wild-type (WT) mice (n=10) and dopamine transporter heterozygous knockout (DAT+/-) mice (n=7), a model of ADHD, on a freely-moving evidence accumulation task. Mice were presented with a sequence of Poisson-distributed stimuli and judged whether the rate was above or below a category boundary of 12Hz. We used maximum a posteriori (MAP) to fit a combined hidden Markov Model and generalized linear model with Bernoulli emissions (GLM-HMM) and characterize internal states. DAT +/- showed lower overall task-performance compared to WT mice. Interestingly, DAT +/- mice in the engaged state outperformed WT mice in the engaged state. DAT +/- mice spent more time in disengaged states compared to WT mice, and had a lower probability of remaining in the engaged state. We then computed a psychophysical kernel that captures the weight of sensory information on the animal’s upcoming decision. WT mice exhibit distinct and consistent state-dependent differences in the strategy of evidence accumulation. Engaged decision-makers strongly prioritized early sensory evidence, while sensory weight was reduced across all time bins during disengagement. By contrast, DAT +/- mice weighed sensory information more uniformly over time, and displayed greater heterogeneity across states. These results demonstrate that sensory integration strategies differ with engagement, and highlight how increased fluctuations in engagement may underly behavioral and cognitive differences in ADHD.