This study explores differences in body mass index (BMI), diet, and lifestyle between women of varying socio-economic status in Karnataka, India, using data from the National Family Health Survey 2 (n ¼ 4374), in-depth interviews (n ¼ 20) and six focus group discussions (n ¼ 40) completed in Bangalore. Predictors of overweight (BMI Z25kg/m 2 ) were modelled using logistic regression. A content analysis of the qualitative data was undertaken to understand behavioural factors affecting overweight. Prevalence of undernutrition was high for rural women (48%) and for overweight, prevalence was high in large urban areas (44%). In Bangalore knowledge of unhealthy foods was higher for richer women, although their diet and activity levels were less healthy than poorer women, who did not have the resources to purchase less healthy diets. Although some knowledge of the need for activity and healthy diets exists for wealthier Bangalore women, this does not translate into healthy behaviour, suggesting a need for obesity prevention programmes. Keywords: nutrition; transition; India; socio-economic status; women While India is one of the poorest countries in the world with extremely high rates of malnutrition, there is evidence of an emerging nutrition transition (Singh et al, 1997; 1999;Griffiths & Bentley, 2001). A higher prevalence of overweight (Body mass index (BMI) Z25 kg/m 2 ) is observed nationally in females of reproductive ages in urban areas (24%) compared to rural areas (6%) (International Institute for Population Sciences & ORC Macro, 2000b). This study uses quantitative and qualitative methods to further understand predictors of overweight in Karnataka, South India.Quantitative data come from the stratified random sample of the Karnataka state National Family Health Survey, 1998/ 99 (NFHS 2) collected between March and September 1999 from questionnaires administered to women of reproductive ages (n ¼ 4374) (International Institute for Population Sciences & ORC Macro, 2001). All women with height and weight data who were not pregnant or delivered within 2 months of the survey were included. Logistic regression using maximum likelihood estimation and accounting for the complex survey design (Stata Release 6, 1999) was used to assess factors associated with being overweight (BMI Z25 kg/ m 2 ) vs normal weight. Only significant variables were retained (P-value o0.05).After observing an association between urban residence and overweight from the NFHS data, in-depth qualitative data were collected in January and February 2001 from women living in Bangalore city. The aim of the qualitative component was to identify current knowledge regarding healthy diet and physical activity behaviours as well as to investigate the determinants of engagement in healthy lifestyle behaviours. Participants were recruited through women's organizations, using a 'snowball' technique to recruit further women. In all, 20 women aged 20-65 y were interviewed individually and an additional 40 took part in six focus groups (5-8 per group). Women ...