Cognitive theories of depression assert that negative self-referent cognition has a causal role in the development and maintenance of depression symptoms, but few studies have examined temporal associations between these constructs using intensive, longitudinal sampling strategies. How these processes unfold over time has not been adequately studied; thus, examining associations across different timescales is also needed. In three samples of undergraduate students, we examined associations between change in self-referent processing measured with a smartphone-based self-referent encoding task (SRET) and depression symptoms measured with the Patient Health Questionnaire across 5 daily assessments (Sample 1, N = 247, 1,194 measurements, 79% adherence), 7 daily assessments (Sample 2, N = 313, 1,784 measurements, 81% adherence), and 7 weekly assessments (Sample 3; N = 155, 833 measurements, 81% adherence). Random intercept cross-lagged panel models indicated large cross-lagged effects in two of the three samples (Samples 1 and 3 but not Sample 2), such that more negative self-referent thinking (i.e., endorsements of negative adjectives on the SRET) than usual was significantly associated with a subsequent increase in depression symptoms at the next time lag. Notably, the reverse association was not observed in any sample. Change in depression from usual was not associated with increases in negative self-referent processing at the next time point. These findings suggest that change in negative self-referent processing may be causally linked to future increases in depression on a day-to-day and week-to-week basis, although confidence in this conclusion is tempered somewhat by a lack of replication in Sample 2.