A limitation of the principal components method is its scale dependence. This note shows that the method is scale invariant if the normalization is modified in an obvious way. Then the effect of a change in units is as transparent as in linear regression, and principal components can be used without apology. Most researchers who use multivariate methods take for granted the invariance property of linear regression, and they can now feel the same assurance about principal components. It is not likely that the researcher would ever need to compute the renormalization proposed. The adjustment is hypothetical, but it demonstrates that the choice of units is inconsequential. (Contains eight references.) (SLD)