With the use of three-dimensional whole body scanning technology, this study compared the 'traditional' anthropometric model [one-dimensional (1D) measurements] to a 'new' model [1D, two-dimensional (2D), and three-dimensional (3D) measurements] to determine: (1) which model predicted more of the variance in self-reported best 2000-m ergometry rowing performance; and (2) what were the best anthropometric predictors of ergometry performance, for junior rowers competing at the 2007 and 2008 Australian Rowing Championships. Each rower (257 females, 16.3 ± 1.4 years and 243 males, 16.6 ± 1.5 years) completed a performance and demographic questionnaire, had their mass, standing and sitting height physically measured and were landmarked and scanned using the Vitus Smart® 3D whole body scanner. Absolute and proportional anthropometric measurements were extracted from the scan files. Partial least squares regression analysis, with anthropometric measurements and age as predictor variables and self-reported best 2000-m ergometer time as the response variable, was used to first compare the two models and then to determine the best performance predictors. The variance explained by each model was similar for both male [76.1% (new) vs. 73.5% (traditional)] and female [72.3% (new) vs. 68.6% (traditional)] rowers. Overall, absolute rather than proportional measurements, and 2D and 3D rather than 1D measurements, were the best predictors of rowing ergometry performance, with whole body volume and surface area, standing height, mass and leg length the strongest individual predictors.