Background <p>Cognitive theories suggest that anxiety symptoms are associated with increased recall of threatening information, but previous evidence has been inconsistent. We examined whether recall of socially rewarding or threatening information was associated with concurrent and subsequent generalised anxiety disorder (GAD) symptoms.</p> Methods <p>We used data from a cohort study (<i>N</i> = 530, 68% female) and the baseline of a randomised controlled trial (<i>N</i> = 653, 58% female). All participants had a history of depressive symptoms. Both studies included a computerised task assessing incidental word recall and measured GAD symptoms using the Generalised Anxiety Disorder Assessment (GAD-7). We tested concurrent associations in both samples and lagged associations in the cohort, which measured GAD scores at four time-points (two weeks apart) and recall at the first three time-points. We used multilevel linear (cohort) and linear (RCT) regression models, before and after adjusting for confounders.</p> Results <p>In the cohort, there was strong evidence that better recall of socially rewarding information was associated with lower GAD scores concurrently (coef=-0.18; 95% CI=-0.31–0.04). There was no evidence for an association with recall of socially threatening words (coef=-0.04, 95% CI=-0.20-0.12). Longitudinally, there was no evidence that recall of socially rewarding or threatening information was associated with subsequent GAD scores. In the RCT, there was evidence that better recall of socially rewarding information was associated with lower concurrent GAD scores (coef=-0.32; 95% CI=-0.56–0.08).</p> Conclusions <p>GAD may be characterised by difficulty in recalling socially rewarding information but not memory for socially threatening information. Our findings indicate that recalling less socially rewarding information may be a marker of current GAD symptoms, but not a risk factor longitudinally.</p>