Objectives: Explorations of relationships between loneliness and depression have focused on loneliness as a uni-dimensional construct. We hypothesised that reciprocal relationships may exist between depressive symptomatology and social and emotional subtypes of loneliness. Methods: Using data from 373 adults aged over 50, who participated in an observational cohort study, we employed a cross-lagged approach within a Structural Equation Modelling framework, to investigate reciprocal links between depressive symptomatology, and social and emotional loneliness, across two waves of data collection, two years apart (controlling for age, sex, education, comorbidities, social network index, and perceived stress). Results: Both depressive symptomatology and loneliness decreased slightly between waves. Autoregressive effects were strong for all three variables of interest. Cross-lagged pathways were evident, such that depressive symptomatology at baseline predicted both emotional (b ¼ 0.26, p < 0.05) and social (b ¼ 0.17, p < 0.05) loneliness at follow-up. Neither emotional (b ¼ 0.07, p > 0.05) nor social (b ¼ 0.05, p > 0.05) loneliness at baseline predicted depressive symptomatology at follow-up. Conclusions: Results challenge existing understanding of the associations between loneliness and depression. Further investigation of emotional and social loneliness in individuals with depressive disorders is warranted. Findings are discussed in relation to mechanisms that may explain the relationships observed, and possible implications.