The nonstress test is commonly used as a test of fetal well-being, despite the fact that both false-positive and false-negative results are known to occur. Although some of the errors are due to biologic variability among fetuses, some may be due to variability of reading the test. The purpose of the present study was to assess the reliability and reproducibility of nonstress test readings. Fifty nonstress test monitor strips were read by five perinatologists on two occasions, for a total of 500 readings. Nonstress tests were read as reactive, equivocal, nonreactive, or technically unsatisfactory. Of the 50 tracings, only 11 had identical readings on all 10 occasions, and readers had different readings for the first and second readings 28% of the time. Estimates of interobserver and intraobserver agreement ranged from 0.41 to 0.55. These results indicate that there is considerable interobserver and intraobserver variability that can be expected to have a serious adverse effect on the sensitivity of the nonstress test.