CERTAIN types of controlled experiments may enable causal inferences to be made about the relationship between two variables; other types of experiments can be used to deal with the (not completely unrelated) notions of consistency in estimation, exogeneity of variables and with temporal ordering. Generally, however, causal interpretations of a relationship are simply imposed on the data by appeal to an underlying theory. In this paper we shall examine the theoretical basis of profits-concentration correlations, arguing that such correlations cannot be given a causal interpretation of the conventional kind running from structure to performance [Section III. 1 Lest this conclusion seem to undermine the basis for anti-trust policy, we shall also argue that such erroneous causal interpretations are unnecessary in making a case for intervention based on structure (although they do direct attention towards inappropriate targets). Profits concentration regressions play a role, it will be argued, mainly in the derivation of intervention rules [Section III]. While the causal interpretation of co-efficients is our primary concern here, there is also a sense in which correlations can be interpreted as measuring the known but unobserved parameters of a model imposed on the data, and we shall close [Section IV] with a few remarks on this second issue.