Fiber quality variability is one of the key determinants of the profitability of the cotton crop. As this variability increases, the risk associated with production increases and the expected return decreases. Prior to our studies, variability had been characterized at the field or crop lever. New technologies permit the analysis of very small samples such that multi-scales studies are possible. We characterized the variability in fiber properties and associated with genetic, environmental, and production-linked sources. This research uses exclusively statistical models with the objective of studying the link between fiber quality variability and different levers at which this variability can be assessed in a cotton population. The treatments imposed on the different plots (water stress, nutrient levers, plasticulture) had as objective the induction of different variances in fiber properties and thus allowing the statistical associations to be made between sources and nature of the variability. We were able to measure the effect of treatments on the number of bolls/m2 and the vegetative development of the plants. Subsequently we want to specifically associate and characterize the changes in this variability at different scales in a production plot.