Eighteen ruminally cannulated dairy cattle were fed a series of diets (in 28-d periods) designed to elicit different degrees of milk fat depression (MFD) for the purpose of relating MFD to ruminal bacterial populations. Cows were fed a TMR containing 25% starch (DM basis) supplied as corn silage, a slowly fermented starch (SFS treatment, period 1), then switched to a TMR containing 27% starch, much of it supplied as ground high-moisture corn, a rapidly fermented starch (RFS treatment, period 2). In period 3, the RFS diet was amended with 13.6 mg of monensin/kg of DM (RFS/Mon treatment), and in period 4, the cows were returned to the RFS diet without monensin (RFS/Post treatment). Effect of both starch source and monensin on milk fat percentage varied by cow, and cluster analysis identified 4 pairs of cows having distinct milk fat patterns. Archived ruminal liquors and solids from the 4 pairs were processed to isolate bacterial DNA, which was subjected to automated ribosomal intergenic spacer analysis followed by correspondence analysis to visualize bacterial community composition (BCC). One pair of cows (S-responsive) showed MFD on RFS feeding, but displayed no additional MFD upon monensin feeding and a fat rebound upon monensin withdrawal. The second pair of cows (M-responsive) showed no MFD upon switch from the SFS diet to the RFS diet, but displayed strong MFD upon monensin feeding and no recovery after monensin withdrawal. Both groups displayed major shifts in BCC upon dietary shifts, including dietary shifts that both did and did not change milk fat production. The third pair of cows (SM-responsive) displayed reduction of milk fat on both RFS and RFS/Mon diets, and fat returned to the levels on the RFS diet upon monensin withdrawal; these cows showed a more gradual shift in BCC in response to both starch source and monensin. The fourth pair of cows (nonresponsive) did not display changes in milk fat percentage with dietary treatment and showed only minor shifts in BCC with dietary treatment. Regardless of milk fat response, BCC did not reassemble its original state upon monensin withdrawal, though the difference was strongest in M-responsive cows. One amplicon length (representing a single bacterial species) was elevated in most, but not all, MFD-susceptible (S-, M-, or SM-responsive) cows relative to milk fat-nonresponsive cows, whereas 2 amplicon lengths displayed reduced abundance under MFD conditions. Overall, this study demonstrates an association between MFD and wholesale shifts of microbial communities in the rumen.