Empirically derived relationships between body size variables and cardiac dimensions have not been published previously for a large sample of male and female athletes. This process would inform scaling practice and facilitate intra- and inter-group comparisons of cardiac data. Therefore we investigated the relationships of body mass (BM), height and body surface area (BS) with a range of cardiac dimensions derived by echocardiography in 464 male and female elite junior athletes (age range 14-18 years; sporting allocation included rowers, cyclists, footballers, tennis players, swimmers and a miscellaneous group). Initial linearity checks suggested that most of the relationships between the body size variables and cardiac dimensions were non-linear, thus precluding the simple ratio standard approach to scaling. Multiple log-log least-squares linear regression confirmed commonality of slopes (between males and females, across the age range and between sporting groups) for all relationships involving BM and BS. Subsequent analyses of the slope exponent (b) for left ventricular dimensions supported previous data and were dimensionally consistent (LVM-BM, b=0.91+/-0.11; LVM-BS, b=1.44+/-0.19; where LVM is left ventricular mass), except for left ventricular internal dimension in diastole (LVIDd) (LVIDd-BM, b=0.25+/-0.04). Data for the left atria internal dimension (LA) were also dimensionally consistent (LA-BM, b=0.29+/-0.09); however, this was not the case for the right ventricular internal dimension in diastole (RVIDd) (RVIDd-BM, b=0.76+/-0.14). It is possible that these results were due to a study-specific limitation in the data range (LVIDd) and the geometric peculiarities of RVIDd compared with LVIDd. The gender/age/sporting groupxbody size interaction factor for virtually all relationships between height and cardiac dimensions was significant (P<0.05), and thus whole-group b exponents could not be generated. Generally these data support previous small-sample research with athletes, and suggest that allometric scaling, as opposed to simple ratio scaling, should be adopted in studies of cardiac dimensions in athletes. This should allow, with minimal mathematical difficulty, the production of body-size-independent cardiac indices to be evaluated in laboratory or clinical work. Further research is required to develop normative 'allometrically derived' cardiac indices, and care should be taken to determine relationships in specific population groups as well as to confirm commonality of slopes in multiple group comparisons. Caution is expressed regarding the use of height as a scaling variable in future research.