Size of GSR made by 300 Ss to 20 repetitions of a visual stimulus is presented as mean magnitude and mean amplitude, illustrating that the amplitude method (averaging only those responses which are greater than zero) is susceptible to distortion resulting from a systematic elimination of Ss who initially make small responses. From trial to trial, the amplitude function comes more and more to be due to the behavior of Ss who initially make large responses. Even though the latter Ss’ responses actually reduce across trials, the amplitude function rises. Magnitude (including zeroes) docs not suffer from this distortion. It is suggested that the definition of a zero response is the source of the problem.