The objectives of this study were to assess the agreement between endometrial cytology and uterine biopsy for the diagnosis of subclinical endometritis (SEND) in grazing dairy cows, the interobserver agreement of the biopsy's readings, and the bacterial population isolated from the uterus of cows having SEND. In experiment 1, lactating Holstein cows (n=44) 31 to 59 d in milk (DIM) at sampling were enrolled. Clinical endometritis was diagnosed by direct evaluation of vaginal discharge and SEND by endometrial cytology evaluation. Two hundred cells per smear were counted to determine the percentage of polymorphonuclear neutrophilic leukocytes (PMNL). Cut-off values used were ≥8% PMNL at ≤33 DIM, ≥6% PMNL at 34 to 47 DIM, and ≥4% PMNL at ≥48 DIM. Biopsies were assessed blindly by 2 observers who categorized them into 4 groups according to their inflammatory changes: none, minimal, moderate, and severe inflammatory changes. Data were analyzed using the kappa coefficient and logistic regression. In experiment 2, lactating Holstein cows (n=60) 21 to 62 DIM were enrolled. Clinical endometritis and SEND were diagnosed as previously described. Samples were cultured for aerobic and anaerobic bacteria by routine methods of bacteriological testing. Data were analyzed with logistic regression. In experiment 1, little agreement was observed between cytology and biopsy outputs (kappa=0.151), and strong agreement between the 2 operators (kappa=0.854). The likelihood of having a normal biopsy (no inflammatory change) was greater for healthy cows than for those having SEND (odds ratio=13.145). The probability for getting normal uterine tissue decreased 2.1% for every increasing percentage point in PMNL. In experiment 2, no bacteria were isolated from cows with SEND, coagulase-negative staphylococci were commonly isolated from healthy cows, and Trueperella pyogenes was frequently isolated from cows with clinical endometritis. The likelihood of isolating T. pyogenes from uterine samples increased with the percentage of PMNL (odds ratio=1.100). In conclusion, biopsy showed low agreement with cytology for the diagnosis of SEND. Nevertheless, fertility trials using uterine biopsies to predict pregnancy outcomes are needed to determine its diagnostic usefulness. Finally, bacteriology would not be recommended as a diagnostic tool because no bacteria were isolated from cows with SEND.