Measurement of nitrogen dioxide using passive diffusion tube over 22 months in Cambridge, U.K. are analysed as a function of sampler exposure time, and compared with NO2 concentrations obtained from a co-located chemiluminescence analyser. The average ratios of passive sampler to analyser NO2 at a city centre site (mean NO2 concentration 22 ppb) are 1.27 (n = 22), 1.16 (n = 34) and 1.11 (n = 7) for exposures of 1, 2 and 4-weeks, respectively. Modelling the generation of extra NO2 arising from chemical reaction between co-diffusing NO and O3 in the tube gave a ratio (modelled/measured) of 1.31 for 1-week exposures. Such overestimation is greatest when NO2 constitutes, on average, about half of total NOx (= NO + NO2) at the monitoring locality. Although 4- week exposures gave concentrations which were not significantly different from analyser NO2, there was no correlation between the datasets. At both the city-centre site and another semi-rural site (mean NO2 concentration 11 ppb) the average of the aggregate of four consecutive 1-week sampler exposures or of two consecutive 2-week sampler exposures was systematically greater than for a single 4-week exposure. The results indicate two independent and opposing systematic biases in measurement of NO2 by passive diffusion sampler: an exposure-time independent chemical overestimation with magnitude determined by local relative concentrations of NO and O3 to NO2, and an exposuretime dependent reduction in sampling efficiency. The impact of these and other potential sources of systematic bias on the application of passive diffusion tubes for assessing ambient concentrations of NO2 in short (1-week) or long (4-week) exposures are discussed in detail.