HVI, utilized in the cotton industry to determine the qualities and classifications of cotton fibers, is time consuming and sometimes destructive. UV/visible/NIR spectroscopy, a rapid and easy sampling technique, was investigated as a potential method for the prediction of such key cotton color and physical attributes as reflectance (Rd), yellowness (+b), micronaire, strength, mean length, upper-half mean length, short fiber index, and uniformity index. Cotton fibers were scanned in the region of 220-2200 nm, and HVI values were measured as the references. PLS regression models were individually developed and then compared for each property in three spectral ranges. The best performances for nearly all properties were obtained from the region covering the UV/visible absorptions, which was in consistent agreement with Pearson correlations from HVI data alone. On the basis of RPD value in the validation set, the suitability of UV/visible/NIR predictive models could be in the descending order of micronaire, +b, Rd, mean length, upper-half mean length, uniformity index, short fiber index, and strength. In addition, to limit the possibility of misclassification for boundary samples from the micronaire PLS model, a 3-class SIMCA/PCA model was developed and the classification efficiency was compared. The comparison indicated that the discrimination model utilizing the UV/visible region could assign one cotton fiber to an appropriate micronaire class of Discount Range, Base Range, or Premium Range with a success rate of 100% for the samples under investigation. Both prediction and classification results suggested that the UV/visible/NIR technique is an accurate means of determining fiber micronaire for cotton quality grading and classification.