Reichenbach worked in an era when philosophers were hopeful about the unity of science, and particularly about unity of method. He looked for universal tests of causal connectedness that could be applied across disciplines and independently of specific modeling assumptions. The hunt for quantum causes reminds us that his hopes were too optimistic. The mark method is not even a starter in testing for causal links between outcomes in E.P.R., because our background hypotheses about these links are too thin to (...) supply the kind of information we need to put the method into play. When we turn to conventional statistical methods, we have seen that one test proposed — robustness of the conditional probabilities — can be conclusive only when we know that there are no other causal factors at work. In the particular case of E.P.R., it has often been assumed that this antecedent question can be settled by applying Reichenbach's conjunctive fork condition. But that application is in no way free of further modeling assumptions. Cartwright (1989) has shown that the conjunctive fork is only a necessary condition on a common cause under very limiting restrictions (restrictions that take one far from the case of maximal commonality); and she has argued that these special conditions are not satisfied in E.P.R.Finally, even given the assumption that there are no other causes at work, the significance of robustness for E.P.R. is unclear. We need a model which tells us how one outcome would influence the other; without that, there is no way of interpreting the results of the robustness test so that it is decisive. In each of the approaches we discussed, Reichenbach provided the crucial guiding ideas that underlay our construction of a causality test; but the articulation of a specific criterion depends on the other details of the model. What is a criterion for a specific kind of link in one model need not be in another. We have illustrated with robustness and E.P.R., but we take the point to be perfectly general: there are no tests of causality outside of models which already have significant causal structure built in. (shrink)

** The primary topic of this dissertation is the study of the relationships between parts and wholes as described by particular physical theories, namely generalized probability theories in a quasi-classical physics framework and non-relativistic quantum theory. ** A large part of this dissertation is devoted to understanding different aspects of four different kinds of correlations: local, partially-local, no-signaling and quantum mechanical correlations. Novel characteristics of these correlations have been used to study how they are related and how they can be (...) discerned via Bell-type inequalities that give non-trivial bounds on the strength of the correlations. ** The study of quantum correlations has also prompted us to study a) the multi-partite qubit state space with respect to its entanglement and separability characteristics, and b) the differing strength of the correlations in separable and entangled qubit states. Results include a novel classification of multipartite (partial) separability and entanglement, strong constraints on the monogamy of entanglement and of non-local correlations, and many new entanglement detection criteria that are directly experimentally accessible. ** Because of the generality of the investigation these results also have strong foundational as well as philosophical repercussions for the different sorts of physical theories as a whole; notably for the viability of hidden variable theories for quantum mechanics, for the possibility of doing experimental metaphysics, for the question of holism in physical theories, and for the classical vs. quantum dichotomy. (shrink)

Three claims about what makes a theory “physically complete” are (1) Shimony's assertion that a complete theory says “all there is to say” about nature; (2) EPR's requirement that a complete theory describe all “elements of reality”; and (3) Ballentine and Jarrett's claim that a “predictively complete” theory must obey a condition used in Bell deviations. After introducing “statistical completeness” as a partial formalization of (1), we explore the logical and motivational relationships connecting these completeness conditions. We find that statistical (...) completeness motivates but does not imply Jarrett's completeness condition, because Jarrett's condition encodes further intuitions about locality and causality. We also dispute Ballentine and Jarrett's claim that EPR-completeness implies Jarrett's completeness condition. (shrink)