Evaluations of energy-price policies are necessarily based on measures of the substitution of energy and non-energy inputs. Facing a variety of substitution elasticities, the central question arises which measure would be appropriate. Apparently, for a long time, this question has not been at issue: Allen's elasticities of substitution (AES) have been the most-used measures in applied production analysis. This paper's main contribution is an instructive survey of the origin of substitution measures and of the trinity of empirical substitution elasticities - AES, cross-price elasticities, and the Morishima elasticities of substitution (MES) - with particular emphasis on their interpretations and the perspectives that will be captured by these measures. This survey clarifies why classical cross-price elasticities are to be preferred for many practical purposes. Berndt and Wood's (Rev. Econom. Stat. 57 (1975) 259) frequently applied data set of US manufacturing is used to illustrate why assessments of energy-price policies would be better based on cross-price elasticities like the energy-price elasticity of capital, rather than on AES or MES. (C) 2003 Elsevier Ltd. All rights reserved.