Objective: Access to social relationships has been linked with better cognitive performance. We examined whether social resources interact with education to predict cognitive outcomes, which could indicate that social resources fulfill a compensatory role in promoting cognitive reserve. Method: We applied multilevel growth models to 6-wave, 13-year longitudinal data from the Berlin Aging Study (aged 70–103 years at first occasion; M = 84.9 years, 50% women) and have taken into account key individual difference factors, including sociodemographic variables, medically diagnosed comorbidities, and depressive symptoms. To account for possible reverse causality, analyses were conducted on a subset of the BASE participants without dementia (n = 368), and in follow-up analyses with the full sample (n = 516) using wave-specific longitudinal assessments of probable dementia status as a covariate. Results: Larger networks were associated with better performance on tests of perceptual speed and verbal fluency, but did not interact with education, providing little support for a compensatory reserve hypothesis. An interaction of education with emotional loneliness emerged in the prediction of perceptual speed, suggesting that the educational divide in speed was minimal among people who reported lower levels of loneliness. Discussion: We discuss our results in the context of differential implications of social resources for cognition and consider possible mechanisms underlying our findings.