There is a paucity of data on the referral of children with meningitis in developing countries, and on the relationship of presentation and outcome to previous care. Referral and previous care were investigated in 281 post-neonatal children treated in two tertiary centres. Data were obtained through the review of referral notes from orthodox health facilities, interview of parents/guardians and review of admission notes. Forty-four (16%) children were facility-referred and 81 (29%) self-referred from orthodox facilities while 156 (55%) were self-referred without previous care in these facilities. The facility-referrals (n=44) included 19 (43%) with meningitis on treatment, 13 (30%) with suspected meningitis and 12 (27%) with unsuspected meningitis. Twenty-two (50%) were referred because of deterioration, partial response or non-response to treatment, 5 (11%) on request by the parents, 9 (21%) on the suspicion of meningitis or other neurological disorder and 7 (16%) for mixed reasons. No reason was given in 1 case of meningitis on treatment. Among the 19 children referred with meningitis on treatment, only 1 was referred within 24 h of diagnosis, a confirmatory lumbar puncture was done only in 7, and only 10 of 18 (no data in 1 case) were on reasonably appropriate antibiotic regimens. Previous care in orthodox facilities was significantly associated with delayed presentation (>3 days of illness, p<0.001), partial treatment (p<0.001), lack of typical signs (p<0.05), severe illness (p<0.01), and adverse outcome (death or recovery with neurological sequelae, p<0.05). Limited recognition of the possibility of meningitis in acutely ill children and an inadequate referral practice may account for these effects. A clear delineation of referral needs might reduce the magnitude of these problems.