Abstract: Adaptations that trade false positives for speed are ubiquitous in evolutionary design. Many, like blinking, are inflexible. Other types of adaptive response appear to be influenced by an individual’s early learning environment. Emotion may be one such case. This paper is part of a larger project that uses machine learning and dynamical systems theory to explain how humans form mature emotional dispositions or habits that persist across the lifespan and are resistant to change (habitual anger, chronic anxiety, persistent depression, etc.). It proposes that a phenomenon known as ‘circular causality’ may play a useful explanatory role. Circular causality is used to cast the relationship between an emotion and its component parts (appraisal, feeling, expression, arousal) in a perhaps unconventional way: An emotion is, in an important sense, identical to its parts. Nonetheless, there is also a sense in which the emotion causes changes in these same parts. This account may help us resolve a puzzle about top-down causation. It may also be key to explaining emotional development across the lifespan.