Most acceleration studies estimate a subject's G-level tolerance by taking only one determination (test) for a given condition. The purpose of this study was to examine the error structure and reliability of an individual's acceleration tolerance and to provide design considerations for future experimentation. A hierarchical (nested) design was used to estimate the sources of variation in measuring G-level tolerance. Six males rode relaxed in the USAF School of Aerospace Medicine human-use centrifuge and were exposed to a 0.1 G/s onset rate profile until greyout. Each subject was tested on three randomly selected days with three repeated determinations within a day. This design allowed for an estimate of both day-to-day and measurement error within a testing session. A single +Gz tolerance determination was found to be moderately unreliable (reliability coefficient = 0.74). Under the best of circumstances a subject's G-level tolerance cannot be estimated with any more accuracy than about +/- 0.3 G with 95% confidence. This degree of accuracy can only be obtained with multiple measurements.