Diurnal variation in stature is one potentially significant source of error in the evaluation of short-term growth. In order to assess the pattern of diurnal variation in height during a 12-h period, the standing height of two healthy prepubertal male siblings, 7 and 11 years of age, was measured by stretched and unstretched techniques five times a day at 3-h intervals over 8 consecutive days. The major loss of height occurred during the first 3 h after rising when the mean stretched height decreased by 0.94 cm (unstretched 1.03 cm). The maximal height loss was achieved by 1500 hours when the mean stretched height had decreased by 1.44 cm (unstretched 1.41 cm) from that recorded in the morning. Over the next 6 h there was a significant (p < 0.001) increment in both stretched and unstretched heights (+0.38 and + 0.42 cm, respectively). A stretch technique did not reduce the diurnal loss. In the second study daily height was measured in the same two siblings over 365 days: (1) to compare the reliability of measurements between stretched and unstretched techniques; and (2) to assess the effect of measurement techniques on non-linear regression analysis and time series analysis used in the evaluation of short-term growth. The differences between the original measurements and the estimates of 'true height' derived from regression curves were used as a measure of accuracy. A smaller SD of these differences and a significantly smaller number of outliers (outside +/-2 SD) than expected (p < 0.05) was generated using the stretched technique (SD = 0.19 cm; 14 outliers) compared the unstretched (SD = 0.23 cm; 53 outliers). Measurement technique did not affect the height and height velocity regression curves and the results of time series analysis. In the evaluation of short-term growth we recommend that children should be measured by a standard stretch technique between 1800 and 2100 hours when the diurnal variation in height is smallest.