BACKGROUND: Previous longitudinal studies have linked multimorbidity to loneliness (feeling alienated) and social isolation (having reduced social contact). However, the nature of these associations over time is unclear. OBJECTIVE: To examine bidirectional associations of multimorbidity with loneliness and social isolation over a 14-year follow-up in a nationally representative cohort of adults aged ≥ 50 years. METHODS: This retrospective cohort study used seven waves of data (collected between 2004/2005 and 2018/2019) from adults in the English Longitudinal Study of Ageing. Multimorbidity was defined as the presence of ≥2 long-term conditions. Loneliness was measured using the 3-item University of California Los Angeles (UCLA) scale. Social isolation was derived based on cohabitation status, frequency of contact with children, relatives, and friends, and social organisation membership. We used Cox proportional hazards models adjusted for social isolation or loneliness, demographic and health behaviour variables. RESULTS: The cohort consisted of 6031 adults with baseline and follow-up data on loneliness, social isolation, multimorbidity, and other covariates. Loneliness was associated with increased risk of incident multimorbidity [aHR (95 % CI): 1.38 (1.15-1.65)], whereas social isolation was not [aHR (95 % CI): 0.97 (0.81-1.16)]. Multimorbidity was associated with increased risk of incident loneliness [aHR (95 % CI): 1.55 (1.30-1.84)], but not significantly associated with subsequent risk of incident social isolation [aHR (95 % CI): 1.09 (0.92-1.28)]. CONCLUSIONS: An independent bidirectional association exists between loneliness and multimorbidity. Interventions targeting loneliness may prevent or delay multimorbidity and also improve wellbeing for people with multimorbidity.