Incomplete vegetation canopies are complex remote-sensing targets. The directional variability of sensor response when viewing a cotton row crop with 48 per cent ground cover was documented for the thermal infrared region. The geometric structure of the canopy was described and radiometric temperatures of four components, sunlit and shaded vegetation and soil, were measured. These data were used to validate and verify a geometric, row projection model. The model predicts the thermal infrared response of a sensor as a function of sensor view angle, component temperature and geometry structure of the canopy. The field data showed sensor response differentials as great as 16.2°C when going from a zenith view angle of 0° to one of 80° in a plane normal to the row direction. The root-mean-square deviation between the model prediction and measured sensor response for all measurement periods and view angles (n = 90) was 0.96°C. The model serves as a sound mathematical basis for interpreting remotely-sensed data from row crops, and optimizing the directional view for specific canopy structures and environments.