Twenty eight isolates of Macrophomina phaseolina were tested for levels of pathogenicity on 45-day-old greenhouse-grown seedlings of trhee cotton cultivars. Isolate X cultivar interaction was a very highly significant (0.0000) source of variation in infected seedlings suggesting that isolates responded differently to different cultivars. Due to the significance of isolate X cultivar interaction, a least significant difference (LSD) was used to compare between the individual isolate means within cultivars, based on these comparisons, it was easy to differentiate between some of M. phaseolina isolates by their differential pathogenicity on some of the tested cultivars. On the other hand, some isolates were indistinguishable because they showed nonsignificant differences on any of the cultivars. Proteins of the isolates were compared by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Gels were stained with silver nitrate (SN). Cluster analysis of the protein banding patterns by the unweighted pair-group method based on arithmetic means placed the isolates in several groups; however, grouping the isolates was not related to their virulence, geographic origin, or host. Some isolates, which were indistinguishable by pathogenicity test, were easily distinguished by their protein banding patterns. The results of the present study indicate that isolates of M. phaseolina pathogenic on cotton could be identified by their differential pathogenicity on a set of cotton cultivars, combined with their specific protein banding patterns separated by SDS-PAGE and stained with SN.