Mehl and colleagues challenge the assumption that autobiographically meaningful music is primarily a vehicle of positive emotion. Thirty younger and thirty older adults listened to self-selected music in two conditions: memory-evoking music and familiar music that did not evoke autobiographical memories. Affect was measured before listening and at three subsequent time points. Positive affect increased after both kinds of music, but negative affect increased only after memory-evoking music, indicating that autobiographical musical cues are especially capable of producing mixed emotional states. The experiment’s critical operation is the use of familiar non-memory-evoking music as a control, allowing autobiographical content to be distinguished from familiarity itself. The paper therefore supplies an empirical mechanism for understanding why personally significant music cannot be equated with pleasurable or desirable listening. Its wider importance lies at the intersection of affect, ageing, memory and music-based intervention: a cue’s mnemonic potency can carry emotional costs as well as benefits, making autobiographical salience an ambivalent rather than uniformly rewarding property.