In the typical advice-taking paradigm, a judge receives one or more pieces of advice and is then invited to make a final judgment based on the just presented advice. Under these conditions, people are sensitive to advice quality (e.g., Meshi et al., 2012). However, people often sample several pieces of advice from different advisors at different points in time. This research investigates the role of memory in delayed advice taking. Our prediction is that source memory moderates the influence of expertise on advice weighting. In Experiment 1 (N = 495), participants receive 48 pieces of advice for 24 different topics concerning the job as a doctor. Two pieces of advice about one topic are presented sequentially in one trial. In each trial, one piece of advice is provided by a doctor (i.e., an expert) and one is provided by a lawyer (i.e., a layperson). After advice presentation, participants provide their own judgment estimates and complete a source monitoring task in which they indicate the source of advice (doctor, lawyer, new). In a between-subjects design, participants either complete the judgment and source monitoring tasks immediately after advice presentation (after six trials in four blocks) or delayed (after having received all 48 pieces of advice). Combining multinomial processing tree modeling to measure source memory (Bayen et al., 1996) using a hierarchical latent-trait modeling approach and mixed effects regression weights of advice (Rebholz et al., 2023), we investigate the influence of item and source memory on expert and lay advice weighting.