How does the brain derive meaning from social vocalizations and drive behavior in response? Vocalizations are temporally complex, requiring integration across multiple time-scales for appropriate contextualization. We hypothesize that corticofugal projections modulate temporal integration in feed-forward auditory circuits to facilitate this. We tested this using an ethological mouse behavior that depends on left auditory cortex. Mothers find and retrieve isolated pups into the nest when pups emit ultrasonic vocalizations (USVs). Virgin females don’t usually retrieve pups, but learn after several days of co-housing with a mom and litter. We chemogenetically silenced neurons projecting to inferior colliculus (corticocollicular) or tail of striatum (corticostriatal) during pup retrieval in experts, and found that corticocollicular, but not corticostriatal neurons were key for successful retrieval. We compared encoding of USVs over days of co-housing using 2-photon Ca2+ imaging in awake virgins. Corticocollicular neurons showed sustained increases in activity during blocks of USV presentation compared to pure tones, while activity was similar for USVs and tones in corticostriatal neurons (N=8 CC mice, N=7 CS). This was corroborated with in vivo whole-cell recordings from optotagged projection neurons (n=8 CC, n=5 CS neurons). In both groups, we observed delayed responses to USVs, which were larger in magnitude on expert days (N=4 mice per group for tracked neurons). To examine how cortical projections modulate activity across auditory areas with precision for spike timing, we performed simultaneous silicon probe recordings from auditory cortex, inferior colliculus, and thalamus (N=6 mice). A subpopulation of neurons in each area didn’t respond to single call syllables but instead integrated over longer sequences. We hypothesize that cortical feedback facilitates temporal integration over multiple syllables, which could enable categorization based on temporal features. To test this we optogenetically silenced corticofugal projections (N=1 mouse), which altered the timing of call-evoked spikes in some IC neurons.