The quality of cotton fibers in Brazil has been studied. This was done by studying the fiber samples obtained from bolls removed from the middle third of the plants. These fiber samples are referred to as "standard sample". This way to collect fiber data requires lots of labor, and may disguise the results obtained in experimental appraisals, due to human errors in gathering boll. Besides, cotton yield and quality is influenced by water availability, especially during abiotic tests with water deficit, in which, fiber quality samples may be affected by boll position. Thus, the objective of this study was to evaluate the influence of sampling method on the technological characteristics of cotton fibers, in irrigated and water stress tests at different stages of the crop cycle. Two methods were made to collect cotton fiber. The first method was the standard sample and the second way was gathering sample of bolls in randomized position, through all experimental plot (called randomized sample). The results show that the analysis performed by standard sample tend to overestimate the values of the fiber quality parameters, differing from the results obtained with the randomized sample that is representative of all plot. It was observed that the variability of cotton fiber quality affected by water stress treatments were best represented using bolls obtained by randomized method. Consequently, in the case of experiments with water stress, the most representative method to collect cotton fiber, is through a sampling of all the plant, and not only of the middle third.