What can behavior tell us about neural dynamics?

For decades, systems neuroscience treated behavior as an endpoint: a simplified output, such as a button press, lick, or saccade, used to isolate neural computations. Richer behavioral measurements have expanded this view by revealing neural dynamics hidden by simple outputs. Yet behavior is not a direct readout of brain activity, because similar behavioral outputs can arise from different neural mechanisms. Interpreting neural dynamics is therefore an inverse problem with no straightforward solution. Behavior can constrain this problem by exposing sensorimotor loops, narrowing plausible explanations for population activity, and formalizing neural-behavioral relationships in joint models. Together, these approaches position behavior as a principled constraint on interpreting neural dynamics and latent cognitive, affective, and computational states.