Determining the reliability and predictive validity of office blood pressure (OBP), ambulatory BP (ABP), and home BP (HBP) can inform which is best for diagnosing hypertension and estimating risk of cardiovascular disease. This study aimed to assess the reliability of OBP, HBP, and ABP and evaluate their associations with left ventricular mass index (LVMI) in untreated persons. The Improving the Detection of Hypertension (IDH) study, a community-based observational study, enrolled 408 participants who had OBP assessed at 3 visits, and completed 3 weeks of HBP, 2 24-h ABP recordings, and a 2-dimensional echocardiogram. Mean age was 41.2 ± 13.1 years, 59.5% were women, 25.5% African American, and 64.0% Hispanic. The reliability of 1 week of HBP, 3 office visits with mercury sphygmomanometry, and 24-h ABP were 0.938, 0.894, and 0.846 for systolic and 0.918, 0.847, and 0.843 for diastolic BP, respectively. The correlations among OBP, HBP, and ABP, corrected for regression dilution bias, were 0.74 to 0.89. After multivariable adjustment including OBP and 24-h ABP, 10 mm Hg higher systolic and diastolic HBP were associated with 5.07 (standard error [SE]: 1.48) and 3.92 (SE: 2.14) g/m2 higher LVMI, respectively. After adjustment for HBP, neither systolic or diastolic OBP nor ABP was associated with LVMI. OBP, HBP, and ABP assess somewhat distinct parameters. Compared with OBP (3 visits) or 24-h ABP, systolic and diastolic HBP (1 week) were more reliable and more strongly associated with LVMI. These data suggest that 1 week of HBP monitoring may be the best approach for diagnosing hypertension.