Spatially variable yield estimates for cotton fields would be beneficial to farmers in terms of marketing their crops and optimizing input vs. output costs. Remote sensing may be useful in producing such estimates. Multispectral satellite images, acquired during the growing season in 1998, were examined for their relationship with cotton yield in a particular Mississippi-Delta cotton field. It was found that one infrared image of the field, acquired 3.5 months in advance of harvest, was a good predictor of cotton yield.