We next recorded the activity of single neurons at multiple stages of the auditory pathway while birds heard the songs that they had learned during behavioral training, the chorus alone, and the auditory scenes used in behavioral testing. From each bird, we recorded single unit responses in the auditory midbrain

(MLd, homolog of mammalian inferior colliculus, n = 100), the primary auditory cortex (Field L, thalamorecipient and immediately adjacent regions, n = 99), and a higher-level auditory cortical region (NCM, n = 170; Figure 2A) that receives synaptic input from the primary auditory cortex (Table S1). Most primary auditory cortex (AC) neurons were recorded in the subregion L3, which provides the majority of input to the higher-level cortical region AZD2281 NCM (Figure S2). All electrophysiology experiments were performed with awake, restrained animals. Action potential widths of higher-level AC neurons formed a continuous distribution with two clear Talazoparib order peaks (p = 0.0001, Hartigan’s dip test), suggesting two largely independent populations (Figure 2B).

Higher-level AC neurons with narrow action potentials (0.1–0.4 ms) were classified as narrow spiking (NS, n = 35; 0.254 ± 0.047 ms, mean ± SD), while neurons with broad action potentials (>0.4 ms) were classified as broad spiking (BS, n = 135; 0.547 ± 0.102 ms, mean ± SD). BS and NS neurons were also largely segregated based on song-driven firing rate, with 90% of BS neurons firing fewer than 3.5 spikes/s and 89% of NS neurons firing greater than 3.5 spikes/s. In contrast to this bimodal distribution, widths of midbrain and primary AC action potentials formed unimodal distributions with peaks in the NS range and tails extending into the BS range that included only a small fraction of neurons (7%, midbrain; 11%, primary AC). None of the BS-like midbrain neurons had driven firing rates less than 3.5 spikes/s, and only 2% of primary AC neurons fired fewer than 3.5 spikes/s. These analyses suggest that the higher-level AC contains a largely distinct population of neurons with very broad action potentials and from low firing rates.

Although individual neurons in each brain area responded to song playback with increased firing rates relative to spontaneous firing (mean z-scores of 4.17, 4.40, 3.31, and 1.36 in midbrain, primary AC, higher-level AC NS, and BS populations, respectively), individual BS neurons in the higher-level AC fired fewer spikes, produced more precise spike trains, and were highly selective for individual songs. Song-driven firing rates of BS neurons were significantly lower than those of neurons in the midbrain, primary AC, or NS neurons in the higher-level AC (2.4 ± 2.7, 39.8 ± 25.4, 32.4 ± 20.1, and 19.0 ± 11.7 Hz, respectively; Figure 2D). Despite the low firing rates of BS neurons, the spikes that individual BS neurons produced were highly reliable across multiple presentations of the same stimulus.