Brains transform sensory information into decisions and decisions into behaviors, which ultimately determine fitness. Behavior can be broken down into a set of discrete chunks of movement, called actions. The basal ganglia (BG) and in particular the input nucleus of the BG, the striatum, is critical for the proper sequencing and selection of actions. At a cellular level, the striatum is comprised of spiny projection neurons (SPNs) that constitute the direct pathway (dSPNs) and indirect pathway (iSPNs). Under the center-surround model of action selection, dSPNs are thought to facilitate the expression of an action while iSPNs are thought to inhibit the expression of other actions. However, it is not clear how each pathway contributes to action selection due to methodological constraints in acquiring an objectively quantitative description of behavior. Our lab has recently developed a pipeline, known as MoSeq, that acquires high-resolution behavioral data and uses an unsupervised algorithm to model stereotyped pose dynamics (actions or “syllables”). Here I propose to combine this state-of- the-art behavioral acquisition and detection technology with both cellular-resolution imaging and optogenetic perturbation to study the population dynamics underlying action selection in the striatum. I hypothesize that SPNs exhibit syllable-specific tuning, where dSPNs are tightly tuned to facilitate the expression of related syllables, while iSPNs are more broadly tuned to suppress the simultaneous expression of other syllables. I will dissect these two processes by recording and manipulating each SPN class during specific syllable expression. In aim 1, I will perform cellular-resolution recordings of the direct or indirect pathway using genetically encoded calcium indicators and miniaturized microendoscopy in the striatum. I will examine the differential roles of the direct and indirect pathways in the context of behavioral tuning. My preliminary data suggest that dSPNs are more sparsely tuned than iSPNs. In aim 2, I will functionally test the center-surround model via direct and indirect pathway inhibition. I will use the inhibitory anion-conducting rhodopsin, ACR2. Using a system capable of detecting syllable expression in real-time, I will perturb each pathway triggered upon the expression of specific syllables to compare the same selection context across many trials. In summary, the experiments proposed here will contribute to a mechanistic understanding of how the BG performs action selection on a moment-to-moment timescale. This proposal is the first to test the predictions made by the center- surround model and will advance our understanding of how the BG encodes actions.