<p>Measurement of pupil size changes is a promising method to investigate information processing. Our study contributes to this stream of research (1) by investigating whether pupillary changes during performing a demanding working memory (n-back) task are associated with performance, and (2) by testing whether these pupil responses demonstrate temporal stability over a period of multiple weeks. We conducted two experiments using two different variants of an n-back task differing in task durations and found that both phasic and tonic pupil size measures are characterized by good test–retest reliability. Furthermore, phasic, task-evoked pupillary responses were positively correlated with task performance, whereas no evidence was found for a similar link in the case of tonic pupil size measures, possibly due to design characteristics. Importantly, the dissociation of validity and reliability regarding tonic pupil size measures suggests that pupil responses might be temporally stable also in such cases when the pupil response itself is not predictive of task performance. Our results imply that pupillometry can be an appropriate tool to assess individual differences in cognition or underlying neural functioning, but one must ensure that the between-sessions correlations among participants are caused by individual differences in information processing, and not by other methodological factors (e.g. participant-specific, systematic measurement noise).</p>