A core aspect of human cognition involves overcoming the constraints of the present environment by mentally simulating another time, place, or perspective. Although these self-generated processes confer many benefits, they can come at an important cost, and this cost is greater for some individuals than for others. Here we explore the possibility that the costs and benefits of self-generated thought depend, in part, upon its phenomenological content. To test these hypotheses, we first developed a novel thought sampling paradigm in which a large sample of young adults recalled several recurring thoughts and rated each thought on multiple content variables (i.e., valence, specificity, self-relevance, etc.). Next, we examined multi-level relationships among these content variables and used a hierarchical clustering approach to partition self-generated thought into distinct dimensions. Finally, we investigated whether these content dimensions predicted individual differences in the costs and benefits of the experience, assessed with questionnaires measuring emotional health and wellbeing. Individuals who characterized their thoughts as more negative and more personally significant scored higher on constructs associated with Depression and Trait Negative Affect, whereas those who characterized their thoughts as less specific scored higher on constructs linked to Rumination. In contrast, individuals who characterized their thoughts as more positive, less personally significant, and more specific scored higher on constructs linked to improved wellbeing (Mindfulness). Collectively, these findings suggest that the content of people’s inner thoughts can (1) be productively examined, (2) be distilled into several major dimensions, and (3) account for a large portion of variability in their functional outcomes.