Upper tropospheric NOx controls, in part, the distribution of ozone in this greenhouse-sensitive region of the atmosphere. Many factors control NOx in this region. As a result it is difficult to assess uncertainties in anthropogenic perturbations to NO from aircraft, for example, without understanding the role of the other major NOx sources in the upper troposphere. These include in situ sources (lighting, aircraft), convection from the surface (biomass burning, fossil fuels, soils), stratospheric intrusions, and photochemical recycling from HNO3. This work examines the separate contribution to upper tropospheric primary NOx from each source category and uses two different chemical transport models (CTMs) to represent a range of possible atmospheric transport. Because aircraft emissions are tied to particular pressure altitudes, it is important to understand whether those emissions are placed in the model stratosphere or troposphere and to assess whether the models can adequately differentiate stratospheric air from tropospheric air. We examine these issues by defining a point-by-point tracer tropopause in order to differentiate stratosphere from troposphere in terms of NOx perturbations. Both models predict similar zonal average peak enhancements of primary NOx due to aircraft (≈10--20 parts per trillion by volume (pptv) in both January and July); however, the placement of this peak is primarily in a region of large stratospheric influence in one model and centered near the level evaluated as the tracer tropopause in the second. Below the tracer tropopause, both models show negligible NOx derived directly from the stratospheric source. Also, they predict a typically low background of 1--20 pptv NOx when tropospheric HNO3 is constrained to be 100 pptv of HNO3. The two models calculate large differences in the total background NOx (defined as the source of NOx from lightning+stratosphere+surface+HNO3) when using identical loss frequencies for NOx. This difference is primarily due to differing treatments of vertical transport. An improved diagnosis of this transport that is relevant to NOx requires either measurements of a surface-based tracer with a substantially shorter lifetime than 222Rn or diagnosis and mapping of tracer correlations with different source signatures. Because of differences in transport by the two models we cannot constrain the source of NOx from lightning through comparison of average model concentrations with observations of NOx. ¿ 1998 American Geophysical Union