Wesley C. Salmon

Starting in 1983, Salmon became interested in theory choice in science, and sought to resolve the enduring conflict between the logical empiricist view, whereby theories undergo a logical process of confirmation and comparison, as against the Kuhnian historical perspective, whereby theory choice and comparison are troubled by incommensurability, the inability of scientists to even effectively communicate and compare theories across differing paradigms.[14] Recognizing that Kuhn's 1962 thesis in Structure of Scientific Revolutions was largely misunderstood—that Kuhn had not meant that scientific theory change is irrational but merely relative to the scientific community where the change occurs—Salmon believed that Bayesianism, which quantifies decisionmaking via subjective probability or "degree of belief", could help close the seemingly unbridgeable gap between the logical empiricist view versus the Kuhnian historical view of theory choice and change.[14]

According to the empiricist view associated with the 18th-century Scottish philosopher David Hume, we do not actually observe causes and effects, but merely experience constant conjunction of sensory events, and impute causality between the observations.[15] More precisely, one finds merely counterfactual causality—that altering condition A prevents or produces state B—but finds no further causal relation between A and B, since one has witnessed no either logical or natural necessity connecting A and B.[15]

In the 20th century, as a formula to scientifically answer Why? questions, logical empiricist Carl Hempel and Paul Oppenheim explicated the deductive-nomological model (DN model). Concerning deterministic laws, the DN model characterizes scientific explanation as a logical form, whereby initial conditions plus universal laws entail an outcome via deductive inference, but no reference to causal relations.[16] Concerning ceteris paribus, which are probabilistic, not deterministic, Hempel introduced the inductive-statistical model (IS model). The IS model, too, indicates correlations, not causation.[16]

By 1970, Salmon had found that when seeking to explain probabilistic phenomena, we seek not merely high probability, but screen for causal influence by removing components of a system to find ones that alter the probability. Salmon sought to replace Hempel's IS model with Salmon's statistical-relevance model (SR model).

In 1948 when explicating DN model, Carl Hempel and Paul Oppenheim had stated scientific explanation's semiformal conditions of adequacy (CA), but acknowledged redundancy of the third, empirical content (CA3), implied by the other three: derivability (CA1), lawlikeness (CA2), and truth (CA4).[17] In the early 1980s, Salmon called for returning cause to because,[18] and helped replace CA3 empirical content with CA3' strict maximal specificity.[9] Yet ultimately, Salmon found mere modifications to the covering law model to be unsatisfactory.[8]

As conventionally conceived by philosophers of science, scientific explanation of a phenomenon was simply epistemic (concerning knowledge), and centered on the phenomenon's counterfactual[19] derivability from initial conditions plus natural laws (Hempel's covering law model).[20] Yet Salmon found causality ubiquitous in scientific explanation,[21] which identifies not only natural laws (empirical regularities), but accounts for them via nature's structure and thereby involves the ontic (concerning reality),[20] how the phenomenon "fits into the causal nexus" of the world (Salmon's causal/mechanical explanation).[8] For instance, Boyle's law relates temperature, pressure, and volume of an ideal gas (epistemic), but this was later reduced to laws of statistical mechanics via average kinetic energy of colliding molecules composing the gas (ontic).[8] Thus, Salmon finds scientific explanation to be not merely nomological—that is, lawlike—but rather ontological, or causal/mechanical.[8] Though asserting the primacy of causal/mechanical explanation, Salmon was vague as to how scientists can attain it.[2] Still, consensus among philosophers of science is that causation is central in scientific explanation.

In metaphysics, Salmon sought a "process theory" of causality to model "causality without counterfactuals", yet meet the "Humean empirical strictures".[22] Salmon criticized Bertrand Russell's theory of causal lines—forerunner of today's theories of causal processes—for involving the epistemic but neglecting the ontic, which causation is.[5] Further, Hans Reichenbach had noted that Russell's causal lines must be distinguished from "unreal sequences": continuous phenomena that actually are not causal processes.[5] Salmon's explanation of causal processes drew a number of criticisms, whereupon Salmon explained that causal processes and causal interactions are "the basic causal mechanisms", while causal interactions are more fundamental than causal processes, but causal processes were discussed first for practical reasons.[22]

Salmon explained causal processes as "the means by which causal influence is transmitted", and thus what "constitute precisely the objective physical causal connections which Hume sought in vain".[22] Salmon explained that causal processes can transmit a mark or can transmit structure in a way continuous spatiotemporally.[8] Thereby, the marking principle sorts causal processes from pseudo processes (Reichenbach's "unreal sequences").[5] Marking a causal process modifies it,[8] a mark not transmitted by a pseudo process.[5] Meanwhile, causal forks are "the means by which causal structure is generated and modified".[22] Others have found Salmon's theory of mark transmission to have shortcomings, however, whereby it can fail to discern causal processes from pseudo processes.[5]