The attenuation of neural responses to repeated stimuli (response attenuation) is often attributed to either adaptation or predictive suppression. While the two mechanisms differ dramatically in their theoretical underpinnings and computational complexity, they have been difficult to separate experimentally, because they make rather similar prediction in many commonly used paradigms. To that aim we developed a new experimental paradigm that leverages the time-course of response attenuation to arbitrate between the two theories. If response attenuation is mediated by adaptation, we expect the strongest attenuation for stimuli presented with short delays to previous sounds, and a monotonic recovery of responses for increasingly longer delays. If, however, response attenuation is mediated by predictive suppression, we expect the strongest attenuation for stimuli presented at the most likely delay, and weaker attenuation for stimuli presented either at shorter or longer delays. To quantify the temporal specificity of response attenuation, we studied auditory evoked EEG and multi-unit responses of macaque monkeys to pure tone pips presented either in 1) highly regular contexts with predominantly predictable timing and identity, or 2) in random contexts with mostly unpredictable timing and identity. In line with the adaptation theory, we found that neural responses were most strongly attenuated if tones were presented at short delays rather than the most likely delay. Furthermore, attenuation was not modulated by the degree of confidence in the upcoming delay or tone identity. In summary, the data strongly support the notion that response attenuation in monkey primary auditory cortex is mediated by adaptation, not predictive suppression. It is possible that monkeys either do not form the necessary temporal predictions, or that they are not fed back all the way to primary auditory cortex.