In passerine songbirds, the lateral portion of the magnocellular nucleus of the anterior neostriatum (LMAN) plays a vital role in song learning, possibly by encoding sensory information and providing sensory feedback to the vocal motor system. Consistent with this, LMAN neurons are auditory, and, as learning progresses, they evolve from a broadly tuned initial state to a state of strong preference for the bird's own song and acute sensitivity to the temporal order of this song. Moreover, normal synaptic activity in LMAN is required during sensory learning for accurate tutor song copying to occur (). To explore cellular and synaptic properties of LMAN that may contribute to this crucial stage of song acquisition, we developed an acute slice preparation of LMAN from zebra finches in the early stages of sensory learning (18-25 days posthatch). We used this preparation to examine intrinsic neuronal properties of LMAN neurons at this stage and to identify two independent excitatory inputs to these neurons and compare each input's pharmacology and short-term synaptic plasticity. LMAN neurons had immature passive membrane properties, well-developed spiking behavior, and received excitatory input from two sources: afferents from the medial portion of the dorsolateral thalamus (DLM), and recurrent axon collaterals from LMAN itself ("intrinsic" input). These two inputs differed in both their pharmacology and temporal properties. Both inputs were glutamatergic, but LMAN responses to intrinsic inputs exhibited a larger N-methyl--aspartate component than responses to DLM inputs. Both inputs elicited temporal summation in response to pairs of stimuli delivered at short intervals, but -2-amino-5-phosphonovalerate (APV) significantly reduced the temporal summation only of the responses to intrinsic inputs. Moreover, responses to DLM inputs showed consistent paired-pulse depression, whereas the responses to intrinsic inputs did not. The differences between these two inputs suggest that intrinsic circuitry plays an important role in transforming DLM input patterns into the appropriate LMAN output patterns, as has been suggested for mammalian thalamocortical networks. Moreover, in LMAN, such interactions may contribute to the profound temporal and spectral selectivity that these neurons will acquire during learning.