The Persistence of Involuntary Memory: Analyzing Phenomenology, Links to Mental Health, and Content

Abstract

In daily life, memories of one’s personal past are often retrieved involuntarily (i.e., unintentionally and effortlessly). Termed involuntary autobiographical memories (IAMs), recent evidence suggests that these are often recurrent (i.e., the same event is remembered repetitively), though controversy surrounds their basic nature. Some research suggests that they are mostly positive or benign, whereas others suggest that they directly contribute to mental health disorders. Here, we show that while recurrent IAMs are common and frequent in general populations, they consistently predict symptoms of mental health disorders. In Study 1, we characterized recurrent IAMs in a large-scale survey of undergraduates. Most participants had experienced recurrent IAMs within the past year (52%), most of which were self-rated as negative in valence (52%). Experiencing negative recurrent IAMs predicted significantly more symptoms of depression, posttraumatic stress, social anxiety, and general anxiety. In Studies 2a and 2b, we examined whether age and trait emotion regulation might modulate recurrent IAMs, because older adults are well-known to have enhanced emotion regulation compared to younger adults. Results indicated that age (Study 2a) reversed the valence distribution: younger adults’ recurrent IAMs were mostly negative, whereas older adults’ were mostly positive. Further, trait emotion regulation (Study 2b) also modulated valence in a sample of younger adults: high emotion regulators were significantly less likely to report negative recurrent IAMs. Regardless of age or trait emotion regulation, experiencing negative recurrent IAMs again predicted greater symptoms of mental health disorders. In Study 3, we asked how analyzing content (e.g., written descriptions of recurrent IAMs) might expand our understanding of these memories, beyond self-reported valence ratings. We developed the first adaptation of computational methods (e.g., machine learning) to understand autobiographical memory content, enabling us to discover content categories (“topics”) in recurrent IAMs. We found that participants experienced recurrent IAMs about a variety of events, ranging from the mundane to the extreme. In Study 4, we extended this computational approach to measure how content might predict mental health above and beyond self-reported valence ratings. Results indicated that elevated symptoms of each disorder were uniquely related to recurrent IAMs about specific topics. Our results suggest that it is imprecise to say that negative recurrent IAMs are related to increased symptoms – our current work pinpoints which specific topics in recurrent IAMs predict mental health. This dissertation provides insight into the nature of recurrent IAMs in large samples of general populations. Importantly, this dissertation distinguishes how these memories and their relationships to mental health are modulated by individual differences. Finally, this dissertation provides a novel framework and methodology (e.g., computational text analysis) for analyzing autobiographical memory content in concert with phenomenology, opening avenues for research to be conducted at an unprecedented scope and scale.