Sinusoidal amplitude modulation (SAM) is a key feature of complex sounds. While psychophysical studies have characterized SAM perception, and neurophysiological studies in anesthetized animals report a transformation from the cochlear nucleus' (CN; brainstem) temporal code to the inferior colliculus' (IC; midbrain's) rate code, none have used awake animals or nonhuman primates to compare CN and IC's coding strategies to modulation-frequency perception. To address this, we recorded single-unit responses and compared derived neurometric measures in the CN and IC to psychometric measures of modulation frequency (MF) discrimination in macaques. IC and CN neurons often exhibited tuned responses to SAM in rate and spike-timing measures of modulation coding. Neurometric thresholds spanned a large range (2-200 Hz DMF). The lowest 40% of IC thresholds were less than or equal to psychometric thresholds, regardless of which code was used, while CN thresholds were greater than psychometric thresholds. Discrimination at 10-20 Hz could be explained by indiscriminately pooling 30 units in either structure, while discrimination at higher MFs was best explained by more selective pooling. This suggests that pooled CN activity was sufficient for AM discrimination. Psychometric and neurometric thresholds decreased as stimulus duration increased, but IC and CN thresholds were higher and more variable than behavior at short durations. This slower subcortical temporal integration compared to behavior was consistent with a drift diffusion model which reproduced individual differences in performance and can constrain future neurophysiological studies of temporal integration. These measures provide an account of AM perception at the neurophysiological, computational, and behavioral levels.