1. Stevens conceived this issue as one of `permissible' mathematical operations. RTM saw here a problem of `meaningfulness': if a claim is not invariant to admissible transformations, it is not empirically `meaningful' (Krantz et al. [1971]; Roberts [1985]). As Michell ([1990], pp.40-46) argues, the issue is better understood as one of valid inferences (versus permissibility or meaningfulness).