The cost of resistance has been measured by a competition experiment over a range of densities, in the absence of herbicide treatment, on two nearly isogenic lines of Foxtail millet, differing in a chloroplastic resistance to herbicide. Three characters have been measured: shoot height, shoot weight, and seed production. Sensitive individuals were better competitors despite a larger decrease in production under within-biotype competition. The cost of resistance was density dependent and increased with density. The cost was higher when measured on seed production and reached 65% at the higher density for resistant individuals. This is compatible with the low frequency or the absence of that gene in natural populations. This work illustrates that the cost is easiest to observe when high levels of constraints are used.