In comparing long-bone cross-sectional geometric properties between individuals, percentages of bone length are often used to identify equivalent locations along the diaphysis. In fragmentary specimens where bone lengths cannot be measured, however, these locations must be estimated more indirectly. In this study, we examine the effect of inaccurately located femoral and tibial midshafts on estimation of geometric properties. The error ranges were compared on 30 femora and tibiae from the Eneolithic and Bronze Age. Cross-sections were obtained at each 1% interval from 60 to 40% of length using CT scans. Five percent of deviation from midshaft properties was used as the maximum acceptable error. Reliability was expressed by mean percentage differences, standard deviation of percentage differences, mean percentage absolute differences, limits of agreement, and mean accuracy range (MAR) (range within which mean deviation from true midshaft values was less than 5%). On average, tibial cortical area and femoral second moments of area are the least sensitive to positioning error, with mean accuracy ranges wide enough for practical application in fragmentary specimens (MAR = 40–130 mm). In contrast, tibial second moments of area are the most sensitive to error in midshaft location (MAR = 14–20 mm). Individuals present significant variation in morphology and thus in error ranges for different properties. For highly damaged fossil femora and tibiae we recommend carrying out additional tests to better establish specific errors associated with uncertain length estimates. Am J Phys Anthropol 2010. © 2009 Wiley-Liss, Inc.