Sir: In his letter, Dr. Voracek claims that the data of our multiple regression analysis were inappropriately analyzed. His claim is based on the fact that most β coefficients were greater than 1. We agree that such results may indicate problems with regard to multicollinearity, thus requiring a careful interpretation. However, what counts against this claim is the fact that the backward-elimination method did work and not all variables remained in the model, which would not have been possible in a major problem with regard to multicollinearity. Furthermore, Dr. Voracek’s claim that “correctly specified multiple regressions never yield β > 1” is not correct. In relatively highly correlated variables, values above 1 are possible all the same; regression models do not have to be wrong because of such values. Dr. Voracek seems to object to the number of correlated variables included in our model. In our view, however, such an argument seems to be too simplistic because it neglects practical aspects of aesthetics and plastic surgery. Of course, most horizontal body measurements are relatively highly correlated because every measurement depends on the particular body weight. A slim woman tends to have not only a narrower waist but usually also smaller hips, a smaller-sized bust, and so forth. What would have been a methodical alternative? Should we have, for example, included waistline measurements but left out the other variables because of their high correlation? Such an approach would have been necessary to exclude any risk of multicollinearity. However, the attractiveness of a female body is much more complex and does not depend on slimness only; thus, female attractiveness cannot be explained by a single variable. We agree that our regression model represents the other extreme, because we used not only several horizontal measurements but also ratios of these measurements. We would like to point out the usefulness of each of these variables, because the ratios act as additional interactive terms that are always highly correlated with the original variables also. Furthermore, our study did not aim at quantifying the contribution of individual body measurements with regard to the attractiveness of a female figure. The main aim was to investigate, in an exploratory manner and in detail by means of highly standardized stimulus material, which variables are relevant factors of attractiveness. The quality of a statistical model, however, can be reliably decided on neither the type of variables used nor β coefficients. A statistical model needs to be tested in a new data set in practice. For several months, we have worked on a validation study with photographs of female bodies (n = 500 or more planned). First results are promising and show that the interrelations identified in our multiple regression analysis are also present in the new data set. We would very much like to submit our new study for publication in due course. Lucas Prantl, M.D. Department of Trauma and Plastic Surgery University Hospital of Regensburg Marita Eisenmann-Klein, M.D. Department of Trauma and Plastic Surgery University Hospital of Regensburg Martin Gründl, Ph.D. Department of Experimental and Applied Psychology University of Regensburg Regensburg, Germany