People’s reports of their thoughts, feelings, and behaviors are used in many fields of biomedical and social science. When these states have been studied over time, researchers have often observed an unpredicted and puzzling decrease with repeated assessment. When noted, this pattern has been called an “attenuation effect,” suggesting that the effect is due to bias in later reports. However, the pattern could also be consistent with an initial elevation bias. We present systematic, experimental investigations of this effect in four field studies (study 1: n = 870; study 2: n = 246; study 3: n = 870; study 4: n = 141). Findings show clear support for an initial elevation bias rather than a later decline. This bias is larger for reports of internal states than for behaviors and for negative mental states and physical symptoms than for positive states. We encourage increased awareness and investigation of this initial elevation bias in all research using subjective reports.