Rhythmic brain activity is thought to reflect, and to help organize, spike activity in populations of neurons during on-going behavior. We report that during learning, a major transition occurs in taskrelated oscillatory activity in the ventromedial striatum, a striatal region related to motivation-dependent learning. Early on as rats learned a T-maze task, bursts of 70-to 90-Hz high-γ activity were prominent during T-maze runs, but these gradually receded as bursts of 15-to 28-Hz β-band activity became pronounced. Populations of simultaneously recorded neurons synchronized their spike firing similarly during both the high-γ-band and β-band bursts. Thus, the structure of spike firing was reorganized during learning in relation to different rhythms. Spiking was concentrated around the troughs of the β-oscillations for fast-spiking interneurons and around the peaks for projection neurons, indicating alternating periods of firing at different frequencies as learning progressed. Spikefield synchrony was primarily local during high-γ-bursts but was widespread during β-bursts. The learning-related shift in the probability of high-γ and β-bursting thus could reflect a transition from a mainly focal rhythmic inhibition during early phases of learning to a more distributed mode of rhythmic inhibition as learning continues and behavior becomes habitual. These dynamics could underlie changing functions of the ventromedial striatum during habit formation. More generally, our findings suggest that coordinated changes in the spatiotemporal relationships of local field potential oscillations and spike activity could be hallmarks of the learning process.basal ganglia | beta rhythm | gamma rhythm | neural circuit reorganization | network dynamics R hythmic field-potential activities in particular frequency bands are known to characterize different on-going behavioral states: low frequency (α-band) rhythms distinguish sleep from waking, higher (β-band) frequencies distinguish movement from rest, and high (γ-band) frequencies distinguish attentive from inattentive states. These activities in different frequency bands can be coordinated by phase or amplitude coupling, and especially when synchronous, such oscillations are proposed to represent on-line processing modes that can facilitate interactions across brain regions and binding of neural ensembles, and that can provide a temporal structure for neural activity related to memory processing and the creation or maintenance of brain states (1-4).It is well documented that spike activity can be organized relative to particular rhythms detected in local field potentials (LFPs), and can even be dependent on such rhythms (5, 6), yet little is known about how such oscillatory activity changes across learning, a dynamic state in which spike activity is reorganized. We explored the possibility that changes in the oscillatory structure of a network might reorganize local circuit activity as a behavior is learned and stamped in as habitual. We took as our starting point that learning-relate...