Recently, we have shown that pupil dilation during a recognition memory task can serve as an index of memory retrieval difficulties in autism. At the time of publication, we were unaware of specific data-analysis methods that can be used to shed further light on the origins of such memory related pupil dilation. Specifically, by distinguishing “tonic” from “phasic” changes in pupil dilation and considering their temporal progression, it is possible to draw inferences about the functional integrity of a locus coeruleus—norepinephrine system (LC-NE) that is known to play a key role in regulating memory encoding and retrieval processes. We therefore apply these analyses to our previously published eye-tracking data of adults with ASD (N = 24) and neurotypical development (TD, N = 30) during the recognition memory task. In this re-analysis, we related pupil dilation during encoding and retrieval to recognition accuracy in a per-trial analysis of linear mixed models. In ASD, we replicated attenuated recognition accuracy, which was accompanied by attenuated pupil dilation during encoding and retrieval. Group differences in pupil dilation during retrieval occurred late during the trial (after 1.75 s) and indicated an altered top-down processing like attenuated attribution of semantic salience in response to previously encoded stimuli. In addition, only in the ASD group were higher pupil dilation during encoding and lower pupil dilation during retrieval associated with decreased recognition accuracy. This supports altered modulation of memory encoding and retrieval in ASD, with LC-NE phasic activity as promising underlying mechanism. Lay Summary: We investigated the changes of pupil size during memory testing in autism spectrum disorder. Adults with ASD remembered fewer items correctly than neurotypical individuals (TD). This reduced memory was related to increased pupillary responses at study and decreased pupil dilation at test only for adults with ASD indicating a different modulation of memory by the locus coeruleus.