Investigations have suggested that owing to the effects of autocorrelation and/or non-stationary behaviour exposure variability increases with the number of days between measurements. This study confirmed such increasing variability with the interval between observations in a collection of occupational data sets after controlling for factors likely to contribute to variability and for sample size. Consecutive shift-long exposure measurements for 53 workers from five different data sets in 123 time series were analysed. When the data were combined a clear increasing trend in the variance was observed with interval, but a breakdown by data set revealed that this trend was present in only two of the five data sets. The effect was further isolated in 30% of the workers who contributed data and in 29% of the total number of time series analysed. Amongst the data where the trend was evident the combination of autocorrelation and non-stationary behavior explained the increase in 64% of the time series. Significant autocorrelation was detected for a small group of workers in only one of the data sets and for a minority of cases amongst workers who contributed more than one time series to the analysis. Thus, autocorrelation of shift-long exposures does not appear to be pervasive and is unlikely to present significant problems when implementing statistically-based sampling strategies. On the other hand, the issue of non-stationarity remains equivocal. Although only a small proportion of time series was found to be non-stationary, the period investigated was short (around 30 days) and it remains to be seen whether the problem is more pronounced over longer time scales.