The National TB Control Programme of Malawi registers and treats large numbers of patients with chronic cough for smear-negative pulmonary tuberculosis (PTB). Smear-negative PTB is diagnosed according to clinical and radiographic criteria, as mycobacterial cultures are not routinely available. In an area of high HIV seroprevalence there is a concern that other opportunistic infections apart from TB, such as Pneumocystis carinii, may be missed owing to lack of diagnostic facilities. The aims of this study were to investigate (i) the extent of P. carinii pneumonia (PCP) in patients about to be registered for smear-negative PTB; (ii) whether there were any clinical or radiological features that could help identify PCP in the absence of more detailed investigations; and (iii) the treatment outcome of PCP patients. A cohort of 352 patients who were about to be started on treatment for smear-negative PTB were investigated further in 1997–1999 by clinical assessment, HIV testing and bronchoscopy. HIV sero-prevalence was 89% (278/313). A total of 186 patients underwent bronchoscopy and bronchoalveolar lavage, and PCP was diagnosed by indirect immunofluorescence or polymerase chain reaction in 17 (9%) of this subgroup. Dyspnoea was significantly more common in PCP cases compared to non-PCP cases (RR 1·35; 95% CI 1·24–1·48; P = 0·008), but discrimination between the groups was difficult using clinical criteria alone. The outcome of PCP cases was poor despite management with high-dose co-trimoxazole and secondary co-trimoxazole prophylaxis, with a median survival of 4 months (25–75% range: 2–12 months).