Learning requires a multitude of steps until efficiency. The process of acquisition, consolidation and automatization depends on cortico-basal ganglia circuits. These circuits are topographically organized into parallel limbic, associative and sensorimotor loops coursing through the ventromedial (VMS), dorsomedial (DMS) and dorsolateral striatum (DLS), respectively; they dynamically interact and are recruited to different extents during initial, goal-directed and highly automatized learning stages, respectively. To date, a comprehensive expression profiling of the learning striatum is lacking. Such profiling will allow for the identification of relevant gene categories and key molecular drivers of striatal learning at different stages, potentially revealing regional differences.To address this, we combined an automated operant conditioning paradigm (high-throughput assessment of self-motivated learning) with cost-efficient, sensitive RNA-sequencing (affordable bulk sequencing of numerous samples) to compare expression profiles among three learning stages in three striatal regions per hemisphere in a total of 870 striatal biopsies. Matched yoked controls allowed to pinpoint learning-specific expression changes. This study also included the humanized Foxp2 mouse model, a model for enhanced striatal automatization and for language/speech evolution (Schreiweis et al., PNAS, 2014). Our study provides a unique resource to study molecular markers of striatal learning, and sheds light on gene categories that might have been implied in human speech/language evolution.