This paper deals with BMI and morbidity in relation to body-fat mass (BFM) and fat-free mass (FFM). The analysis was based on cross-sectional data concerning the age, household income, anthropometry and morbidity of 575 males aged 18–59 years from a rural community in North-East India. Data on morbidity were based on the self-reported morbidity (SRM) of the subjects during the last 4 weeks before the survey, whereas data on BMI and body composition were estimated from anthropometry. It was found that SRM was significantly associated with age and income. However, the relationship between BMI and SRM was not significant after adjusting for age and income. Separating the BMI into body-fat mass index (BFMI being BFM in kg divided by height squared in metres) and fat-free mass index (FFMI being FFM in kg divided by height squared in metres), it was found that BFMI was significantly associated with SRM after adjusting for age, income and FFMI. The subjects with a low (<2·9 kg/m2) BFMI were about 4·7 times (odds ratio 4·7, 95 % CI 2·6, 8·6) more likely to become sick than those with a normal (2·9–5·0 kg/m2) BFMI. In addition, the risk of becoming sick was higher in the subjects with a high (>5·0 kg/m2) BFMI than in those with a normal BFMI (odds ratio 3·9, 95 % CI 1·3, 9·8). However, the relationship between FFMI and morbidity was not clearly perceptible. It is therefore speculated that BMI may not always provide accurate information about the variation in body fat and body composition that is associated with morbidity.