A sound at a low level is heard as much softer after having decreased continuously from higher levels than if presented after a period of silence at that same low level. Canévet [Acustica61, 256–264 (1986)] demonstrated this phenomenon for a tone that (1) decreased from 65 to 20 dB in 180 s; he also presented a tone that (2) increased from 20 dB, or (3) was presented as pairs of bursts at various levels in random order. Below about 40 dB, loudness changed most rapidly in the decreasing condition so that, at 20 dB, the tone was judged ten times softer than in conditions (2) and (3). In the present experiments, magnitude estimation was used to examine the possible role of judgmental biases and adaptation in this rapid loudness decline, which we calldecruitment. Results show that decruitment did not come about because subjects made many successive loudness judgments; loudness declined as much when a tone was judged only twice, at the beginning and end of its 180‐s decrease. In contrast, interrupting the decreasing tone so that it was heard only at 70 dB and 160 s later at 30 dB greatly diminshed the decruitment. Similarly, pairs of 500‐ms tone bursts presented at successively lower levels instead of continously decreasing did not show decruitment, suggesting that sequential biases are irrelevant. The likely cause of decruitment is sensory adaptation.