Antibiotic resistance of 1515 consecutive laboratory isolates of Streptococcus pneumoniae between 1989 and 1994 was analyzed. Overall, 39 (2.6%) isolates were resistant to penicillin, 102 (6.7%) resistant to erythromycin and 52 (3.4%) resistant to tetracycline. There was a higher proportion of penicillin resistant isolates from sterile sites compared with "non-sterile sites" (5% vs. 2.2%; P < 0.02). This same pattern occurred with erythromycin (12.5% vs. 5.6%; P < 0.001). From 1989-90 to 1993-94 the penicillin resistance rate increased from 0.8% to 8% and the erythromycin from 5.7% to 8.4%, whereas the tetracycline resistance rate fell from 3.7% to 2.8%. The increase in resistance to penicillin largely occurred in the final 12 months of this study period. One hundred and fifty isolates (9.9%) were serotyped, including isolates from sterile sites and those with penicillin resistance. The commonest serotypes of penicillin-sensitive pneumococci were 14, 19, 9 and 6. The majority of penicillin-resistant pneumococci (PRP) were of serotype 9 (64%) followed by 6, 23 and 19. Overall 95% of these isolates were of serotypes represented in the 23-valent pneumococcal polysaccharide vaccine (Pneumovax II). PRP were more likely to have resistance to erythromycin (23%) or tetracycline (23%) compared to penicillin-sensitive pneumococci (6% and 3% respectively). Most of the PRP were isolated from patients aged over 50 years including 11 isolates from blood cultures of patients with pneumonia or septicaemia. There was a possible epidemiological association between four patients with PRP but surveillance cultures of hospital contacts revealed no extra cases. These results show a worrying increase in infections due to PRP which has implications for clinical and laboratory staff in the diagnosis and treatment of serious pneumococcal infections.