An important step in studies relating worker health to industrial exposure is the estimation of mean exposure levels. The investigator frequently has to rely on industrial hygiene measurements collected for other purposes. Samples may have been taken at several companies on different dates, and on each occasion multiple individual samplers may have been employed. Often it is not recognized that readings from such a hierarchical arrangement are correlated; for example, samples taken at the same time and location are more alike than samples taken on different days. This correlation invalidates the commonly used standard errors of sample means and the usual sample standard deviation. A component of variance analysis is suggested which quantifies within-day, between-day and between-company variation. Estimators of mean exposure are presented with correct standard errors. The techniques are illustrated by a small set of data and by a recent study of exposures to styrene in 36 companies manufacturing reinforced plastics.