DescriptionDeviants are stimuli that violate the ongoing sequence or distribution of sensory events; they are potentially salient and the sensory system constantly monitors them. However, detecting a deviant is not a trivial process and requires to compare the current stimulus with prior memories or predictions. Traditionally, deviance detection has been studied in both humans and animals by presenting pure tones in a paradigm in which a rare tone stimulus (the oddball) occurs at random in a repeated sequence of a different tone (the standard). The current study seeks to more fully investigate the process and neural substrate of auditory deviance detection by using several paradigms, including both an extension of the traditional oddball approach that uses complex sounds as stimuli and also a new context paradigm. Multi-unit auditory responses to these stimuli were recorded from the auditory forebrain of awake male zebra finches. Results show that an oddball effect (larger responses to a sound when it is deviant than when it is common) can be elicited with complex stimuli like zebra finch calls (as well as with tones), and that the effect magnitude increases as common and deviant stimuli become more different acoustically. These results are consistent with a simple form of stimulus-specific adaptation that generalizes to similar sounds. However, the order in which blocks of stimuli were presented changed the size of the oddball effect, suggesting a role for a memory of stimulus patterns that persists over longer durations and across many intervening stimuli. In the new context paradigm, where deviance could be defined mathematically, the neural response to a given stimulus depended on the larger context in which it was presented, again suggesting a perceptual learning effect. These experiments advances the study of deviance detection by using neural data to identify and separate the longer term effects of stimulus familiarity and pattern from the immediate effects of presentation order and relative frequency that are studied in simple oddball paradigm.