Passive diffusion samplers containing tri-ethanolamine (TEA) as adsorbent are widely deployed for measuring NO2 concentrations in urban and rural air. In this study, a one-dimensional model of the sampler incorporating diffusive transport and chemical reaction has shown that NO2 concentrations are systematically overestimated as a result of chemical reactions in the tube, dominated by the reaction of NO with O3 to give NO2. The contribution of PAN and other potential interferents is shown to be small for British conditions.The extent of overestimation depends on the instantaneous ratio and on the relative concentrations of NO and O3. In rural air, where [NO]<[NO2], [NO] ⪡ [O3] and total NOx (NO+NO2) concentrations are typically less than 5 ppbV (parts in 109 by volume), the model shows that diffusion tubes effectively measure NOx rather than NO2 concentrations, although in practice the difference between NO2 and NOx concentrations is likely to be within the measurement uncertainty. In urban air, however, close to sources of NO, where average [NO] is of a similar magnitude to [NO2], a large overestimate of NO2 is made. Using hourly measured gas concentration data for a summer month in Edinburgh, the model predicted 28% more NO2 captured by the diffusion tube than measured as NO2 by a continuous monitor. The overestimate at rural sites, and at the urban site in winter, was between 8 and 14%.The systematic error of within-tube chemistry is independent of, and acts in the same direction as, the possible overestimation caused by an effective shortening of diffusion path length at high wind speeds and may be one explanation for diffusion sampler overestimations previously attributed to wind effects. There appears to be no simple method for retrospectively correcting NO2 measurement data obtained using diffusion tubes since the overestimate is not simply related to the average nor ratios, but reflects the dynamical behaviour throughout the sampling period. Although passive diffusion samplers may give reliable concentration data for NOx in rural air, they cannot be used with confidence in polluted urban air to estimate NO2 concentrations.