The aim of this study was to quantify the relationship between heart rate determined physical activity levels (HR-PAL), and estimates of cumulative low back loads estimated from video records of non-occupational activities. Subjects were videotaped while performing self-determined, non-occupational activities within their own homes, for a period of 2 hours. Subject HR was continuously recorded during the data collection period. The 2-hour HR profile, along with subjects' height, mass, age, gender, and median sitting HR, were used as inputs into a regression-based mathematical model to estimate HR-PAL. The video data were captured to digital format at 3 samples/sec; the video was trimmed to match the heart rate file, and each task was then trimmed out into separate video image clips. The video clips were analyzed with the 3-D Match video analysis tool. At an alpha level of 0.05, quadratic regression equations were able to account for a significant amount of variance in cumulative compression force (R²=0.817), cumulative flexion moment (R²=0.757), and cumulative right axial twist moment (R²=0.769). Significant differences were found between predicted and actual compression force (-3.502, p=0.04), joint anterior shear force (t=-22.527, p = 0.00), reaction anterior shear force (t=-17.471, p=0.00), and flexion moment (E 14.016, p = 0.00). When an alpha level of 0.1 was used, quadratic regression equations were able to account for a significant amount of variance in four additional cumulative output variables: cumulative reaction anterior shear force (R²=0.573), cumulative extension moment (R²=0.492), cumulative left lateral bend moment (R²=0.535), and cumulative left axial twist moment (R²=0.492). Significant differences were also found between predicted and actual joint anterior shear force (t = -2.437, p = 0.09). While HR-PAL shows promise in its ability to predict cumulative L4/L5 spine loading, further investigation of this relationship is needed.