The use of substances considered toxic by definition, in much lower doses than used can stimulate plant development. This effect is known as "hormesis effect" (CALABRESE; BALDWIN, 2002). The hormesis effect was observed in all groups of organisms like bacteria and fungi, higher plants and animals (Calabrese 2005) may become a tool to be employed in the production system of cotton or agriculture in general. The aim of this study was to determine the effects of application of under levels of glyphosate on cotton plants in field conditions. This work was developed in the growing season of 2010/11 out at the  the experimental area of São Paulo State University, in the county of Selvíria, State of Mato Grosso do Sul, Brazil. The experimental design was the randomized completely blocks with 6 treatments and 4 repetitions, totaling 24 plots. The treatments were a combination of sublethal rates of glyphosate, being: 0.0 (control), 26.0, 52.0, 78.0, 104.0, and 130.0 g ha^-1 .   Significant differences were verified for potassium, the leaf content was increased with a difference of 3.26 g kg^-1 comparing the dose and 130 g ha-1 in control . Analyzing the results of the variables in the agricultural season of 2010/11,  it was found that the doses did not influence significantly the number of bolls per plant.  The mass of 20 bolls were not affected by the rates of glyphosate in the study, probably because it is a genetic characteristic of the cultivar.  The Yield was affected by doses studied, with quadratic fit of the medium, with maximum increase yield in sub-dose of 65 g ha-1, with an increase of approximately 693 kg ha^-1 compared to control.  It was verified that the greater level of 130 g ha^-1 generated no increase in the yield, so this sub-dose was phytotoxic to cotton plants.