Purpose: Loneliness is a recognised public-health concern that is traditionally regarded as a unidimensional construct. Theories of loneliness predict the existence of subtypes of loneliness. In this study, latent class analysis (LCA) was used to test for the presence of loneliness subtypes and to examine their association with multiple mental health variables. Methods: A nationally representative sample of US adults (N=1839) completed the De Jong Gierveld Loneliness Scale, along with self-report measures of childhood and adulthood trauma, psychological wellbeing, major depression, and generalized anxiety. Results: When treated as a unidimensional construct, 17.1% of US adults aged 18–70 were classified as lonely. However, the LCA results identified four loneliness classes which varied quantitatively and qualitatively: ‘low’ (52.8%), ‘social’ (8.2%), ‘emotional’ (26.6%), and ‘social and emotional’ (12.4%) loneliness. The ‘social and emotional’ class were characterised by the highest levels of psychological distress, followed by the ‘emotional’ class. The ‘social’ loneliness class had similar mental health scores as the ‘low’ loneliness class. Childhood and adulthood trauma were independently related to the most distressed loneliness classes. Conclusions: Current findings provide support for the presence of subtypes of loneliness and show that they have unique associations with mental health status. Recognition of these subtypes of loneliness revealed that the number of US adults aged 18–70 experiencing loneliness was twice as high as what was estimated when loneliness was conceptualized as a unidimensional construct. The perceived quality, not the quantity, of interpersonal connections was associated with poor mental health.