Crackles heard on auscultation can be represented graphically as a time-amplitude plot of the associated waveform. To assess the relative merits of several measures which might be considered for machine implementation in diagnostic instruments, we compared the reproducibility of those based on the initial voltage deflection which begins a crackle with those based on the largest deflection. The latter group showed less interobserver and less intraobserver variability when the same crackles were measured twice by each of two observers. Crackles from a teaching tape, categorized as fine and coarse, were used in this study. The ability of the various measures tested to distinguish between fine and coarse crackles on an individual basis was assessed and found to favor the measures based on the largest deflection. They showed an average of 9.96 percent incorrectly classified crackles, as opposed to 19.53 percent for the two measures based on the initial deflection.