Abstract A pertinent interpretation of thermal infrared (TIR) information to characterize crop water status requires at least to consider the fraction of crop cover. Even if the crop cover is known, such an interpretation remains difficult and the current issues to be overcome in the field of TIR remote sensing applications stands on bare soil effects. An experiment was conducted during summer 1999 in Montpellier (France) on a row-cotton crop in order to acquire a data set relating thermal and optical multidirectional measurements to crop structure and water status. The crop was monitored all along its development. Three plots were delimited: a reference plot with no water limitation and two plots without water supply respectively at flowering and cutout stage. On three dates, directional TIR and optical images were acquired both on the reference plot and on the one with limited water supply. Directional averaged temperatures (Ts) and Normalized Difference Vegetation Index (NDVI) values showed a strong dependence on canopy gap fraction. Ts appeared particularly influenced by directional sunlit soil fraction variability, depending on both sun/sensor angle configuration, crop structure and water status. Leaves at different levels in the canopy (with different ages and spectral properties) could be observed by the sensor, but the impact of the sunlit/shaded leaves ratio on directional temperature measurements was weak in comparison to soil effects. The different directional influence of sunlit soil fractions on Ts and NDVI values explains in a large part the limits encountered by water stress indices approaches, aiming at relating linearly such variables, when applied to partially covering crops. Such results provide an exhaustive experiment-based biophysical analysis of very high resolution multidirectional TIR signal. They point out further ways of investigations to be explored in the field of water stress indices improvement or performing. This comes as a preamble of an experiment-based analysis of the limits and opportunities of water stress indices methods, complemented with a 3D model-based analysis that allows confirmation and extrapolation of the results to larger ranges of crop characteristics and directional configurations.