Evan Fales

Theories of how organisms learn about cause-effect relations have a history dating back at least to the associationist/mechanistic hypothesis of David Hume. Some contemporary theories of causal learning are descendants of Hume's mechanistic models of conditioning, but others impute principled, rule-based reasoning. Since even primitive animals are conditionable, it is clear that there are built-in mechanical algorithms that respond to cause/effect relations. The evidence suggests that humans retain the use of such algorithms, which are surely adaptive when causal judgments must be rapidly made. But we know very little about what these algorithms are and about when and with what ratiocinative procedures they are sometimes replaced. Nor do we know how the concept of causation originates in humans. To clarify some of these issues, this paper surveys the literature and explores the behavioral predictions made by two contrasting theories of causal learning: the mechanical Rescorla-Wagner model and the sophisticated reasoning codified in Bayes' Theorem

It has been nearly a decade and a half since Fred Dretske, David Armstrong and Michael Tooley, having each rejected the Regularity theory, independently proposed that natural laws are grounded in a second-order relation that somehow binds together universals.' (l shall call this the ‘DTA theory’). In this way they sought to overcome the major - and notorious — shortcomings of every version of the Regularity theory: how to provide truth conditions for laws that lack instances; how to distinguish laws from accidental generalizations; how to provide truth conditions for the counterfactuals and disposition statements that laws apparently ‘support’; how to justify inductive inferences from past events to laws and future events. For each of these puzzles, an apparently key element in the solution seems to be missing from Regularity theories. That missing element is a genuine connection, a relation with more than merely spatial and/or temporal content, linking the antecedent of a law to its consequent. Once such an additional objective element - however understood — is admitted to be essential to the analysis of laws, one is forced to give up the idea that the logical form of laws can be given in terms of quantifiers ranging over events or states of alfairs, and truth-functions.