Unhealthy dietary behaviours may contribute to obesity along with energy imbalance. Both positive and null associations of snacking and BMI have been reported, but the association between snacking and total adiposity or pattern of fat deposition remains unevaluated. The objective of this study was to investigate the associations between snacking frequency and detailed adiposity measurements. A total of 10 092 adults residing in Cambridgeshire, England, self-completed eating pattern snacking frequency, FFQ and physical activity questionnaires. Measurements included anthropometry, body composition using dual-energy X-ray absorptiometry scan and ultrasound and assessment of physical activity energy expenditure using heart rate and movement sensing. Linear regression analyses were conducted adjusted for age, socio-demographics, dietary quality, energy intake, PAEE and screen time by sex and BMI status. Among normal-weight individuals (BMI<25 kg/m2), each additional snack was inversely associated with obesity measures: lower total body fat in men and women (−0·41 (95 % CI −0·74, −0·07) %, −0·41 (−0·67, −0·15) %, respectively) and waist circumference (−0·52 (−0·90, −0·14) cm) in men. In contrast, among the overweight/obese (BMI≥25 kg/m2), there were positive associations: higher waist circumference (0·80 (0·34, 0·28) cm) and subcutaneous fat (0·06 (0·01, 0·110) cm) in women and waist circumference (0·37 (0·00, 0·73) cm) in men. Comparing intakes of snack-type foods showed that participants with BMI≥25 kg/m2 had higher intakes of crisps, sweets, chocolates and ice-creams and lower intakes of yoghurt and nuts compared with normal-weight participants. Adjusting for these foods in a model that included a BMI–snacking interaction term attenuated all the associations to null. Snacking frequency may be associated with higher or lower adiposity, with the direction of association being differential by BMI status and dependent on snack food choice. Improving snack choices could contribute to anti-obesity public health interventions.