Principal component analysis was used to demonstrate the main associations between patterns of resistance to antibiotic drugs in 670 gram-negative bacteria consecutively isolated from blood cultures over a period of two years. Six factors were derived, which accounted for 84% of the total variance of the original matrix. Each factor represented an association between resistance to certain antibiotics as follows: factor 1: aztreonam, third generation cephalosporins and aminoglycosides; factor 2: first and second generation cephalosporins; factor 3: tetracycline and chloramphenicol; factor 4: ampicillin and ureidopenicillins; factor 5: trimethoprim/sulfamethoxazole; factor 6: fluoroquinolones. On two-way analysis of variance the difference in the factor scores was significant between bacteria for all factors except factor 5. The difference in factor scores between community and hospital acquired strains was significant only for factors 1, 2 and 6. Only the score of factor 6 showed a clear trend to increase with time during the two-year study period. Patients who were treated with antibiotics prior to bacteremia had higher scores for all factors, the difference being most marked in patients treated with fluoroquinolones. Factor analysis can be used to describe phenotypic associations between resistance to antibiotics, and the factor score used to compare groups of isolates and to demonstrate temporal and other trends.