Transcutaneous oxygen tension measurements (PtcO2) were obtained for subjects in two groups: peripheral vascular patients (N = 15) and disease-free controls (N = 9). Readings were taken in each of three locations, arm, knee and foot, in supine and erect positions, at fixed and random times, on each of 3 consecutive days. The dependability (the proportion of true variance in total variance) of PtcO2 measurements was evaluated using a generalizability model. The preliminary generalizability study analysis indicated that day-to-day variation of PtcO2 level was larger than within-day variation. Therefore, in order to decrease the measurement error variance one would preferably increase the number of days for measurements (nd) rather than increase the number of measurements within each day (nt). A decision study analysis was also performed which resulted in estimates of error variance and two interdependent dependability measures: dependability coefficients (DCs) and signal to noise ratios (S/Ns). PtcO2 dependability values were generally different for the two groups. Cases had high DCs and S/Ns (DC greater than or equal to 0.9, S/N greater than or equal to 9) in all location-position combinations except for arm measurements. On the other hand, controls had low and unacceptable DCs and S/Ns (DC less than 0.8, S/N less than 4) in all location-position combinations. Cases and controls had generally lower dependability values when PtcO2 ratios were analyzed. The only two ratio-position categories for which dependability values were in the acceptable range (DC greater than or equal to 0.8, S/N greater than or equal to 4) were foot/arm erect and foot/arm supine for the cases.(ABSTRACT TRUNCATED AT 250 WORDS)