In cross-sectional epidemiological studies, blood pressure (BP) is often found to be positively correlated with fatness. Usually sphygmomanometers with only one cuff size for adults are used to measure BP while arm circumference (AC) influences BP readings. We have studied cross-sectional anthropometric and BP data of adult men and women from three populations: Cook Islanders (n = 259), Papua New Guinean: Purari (n = 295), and Ok Tedi (n = 274). These were selected because of their diverse socio-economic, anthropometric, and BP characteristics. Partial correlations and regressions were used to analyze these data. Systolic and diastolic pressures (SBP, DBP) showed dependence on AC, body mass index (BMI), and skinfold thickness. Stature had some effect on SBP and DBP, independent of BMI and AC. When effects of AC and stature were statistically controlled, BMI did not correlate with either SBP or DBP. People of larger body mass have greater AC, and this biases BP readings. Average values of SBP and DBP in groups of underweight, normal, overweight, and obese people predicted by AC (sex, age, and BMI being statistically controlled) closely matched observed SBP and DBP averages in those groups. Out of 24 pairwise comparisons (3 samples from different populations × 4 groups of BMI × 2 pressure readings) of predicted and actual BP, only two produced statistically significant differences while 21 of the differences were 5 mm Hg or less. Correlations between BP and obesity found in epidemiological studies may be severely biased by effects of variation in AC. Sphygmomanometric measurements of BP should be corrected for continuous variation in AC.