Serum retinol concentrations decrease during illness and thus may not accurately reflect the vitamin A status of populations with a high prevalence of illness. To quantify the contribution of illness to low serum retinol in a field study of children aged 6-59 mo in northern Ghana, serum retinol values were compared with two indicators of recent illness; symptoms reported by parents and acute-phase protein concentrations in serum. Serum retinol was not associated with symptoms of illness but showed a significant negative correlation with both alpha 1-acid glycoprotein (AGP) and serum amyloid A (SAA). Elevated AGP was associated with a 24% decrease in mean serum retinol. A large proportion of asymptomatic children had elevated AGP or SAA concentrations, suggesting that subclinical infections may have had important effects on serum retinol. A significant negative correlation between malaria parasite density and serum retinol indicated that malaria may have been one of the subclinical infections responsible. Measurement of AGP may improve interpretation of serum retinol data from populations with a high prevalence of morbidity.