How do we go about weighing evidence, testing hypotheses, and making inferences? According to the model of _Inference to the Best Explanation_, we work out what to infer from the evidence by thinking about what would actually explain that evidence, and we take the ability of a hypothesis to explain the evidence as a sign that the hypothesis is correct. In _Inference to the Best Explanation_, Peter Lipton gives this important and influential idea the development and assessment it deserves. The (...) second edition has been substantially enlarged and reworked, with a new chapter on the relationship between explanation and Bayesianism, and an extension and defence of the account of contrastive explanation. It also includes an expanded defence of the claims that our inferences really are guided by diverse explanatory considerations, and that this pattern of inference can take us towards the truth. This edition of _Inference to the Best Explanation_ has also been updated throughout and includes a new bibliography. (shrink)

Several of these essays have been printed whole in journals; others are in varying degrees new. Two main themes run through them. One is the problem of meaning, particularly as involved in the notion of an analytic statement. The other is the notion of ontological, commitment, particularly as involved in the problem of universals.

Scientific reasoning is—and ought to be—conducted in accordance with the axioms of probability. This Bayesian view—so called because of the central role it accords to a theorem first proved by Thomas Bayes in the late eighteenth ...

An amended bootstrapping can avoid Christensen's counterexamples. Earman and Edidin argue that Christensen's examples to bootstrapping rely on his failure to analyze background knowledge. I add an additional condition to bootstrapping that is motivated by Glymour's remarks on variety of evidence. I argue that it avoids the problems that the examples raise. I defend the modification against the charge that it is holistic, and that it collapses into Bayesianism.

In a series of articles, Paul Thagard has developed a connectionist''s modelfor the evaluation of explanatory coherence for competing systems ofhypotheses. He has successfully applied it to various examples from thehistory of science and common language reasoning. However, I will argue thathis formalism does not adequately represent explanatory relations betweenmore than two propositions.In this paper, I develop a generalization of Thagard''s approach. It is notsubject to the connectionist paradigm of neural nets, but is based on fuzzylogic: Explanatory coherence increases with (...) the fuzzy truth value of theconjunction of explanans and explanandum and decreases with the value of theconjunction of explanans and the negation of the explanandum. (shrink)

Bonjour (1985: 101 and 1999: 124) and other coherence theorists of justification before him (e.g. Ewing, 1934: 246) have complained that we do not have a satisfactory analysis of the notion of coherence. The problem with existing accounts of coherence is that they try to bring precision to our intuitive notion of coherence independently of the particular role that it is meant to play within the coherence theory of justification (e.g Lewis, 1946: 338). This is a mistake: it does not (...) make any sense to ask what precisely makes for a more coherent information set independently of the particular role that coherence is supposed to play within the context in question. What is this context and what is this role? The coherence theory of justification rides on a particular common sense intuition: when we gather information from less than fully reliable sources, then the more coherent the story that materializes is, the more confident we may be, ceteris paribus. Within the context of information gathering from certain types of sources, coherence is a property of stories which plays a confidence boosting role. But what features should the information sources have, so that the coherence of the information set is indeed a determinant of our degree of confidence in question? And what goes into the ceteris paribus clause? In other words, what other factors affect our confidence in the information set in question? (shrink)

Your use of the JSTOR archive indicates your acceptance of J STOR’s Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. J STOR’s Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non—commercial use.

This paper develops a semantical model - theoretic account of (logical) content complementing the syntactically specified account of content developed in "A New Theory of Content I", JPL 23: 596-620, 1994. Proofs of Completeness are given for both propositional and quantificational languages (without identity). Means for handling a quantificational language with identity are also explored. Finally, this new notion of content is compared, in respect of both logical properties and philosophical applications, to alternative partitions of the standard consequence class relation (...) proposed by Stelzner, Schurz and Wiengartner. (shrink)

In a 1994 ANALYSIS article Peter Klein and Ted Warfield show that an epistemically more coherent set of beliefs often has a smaller unconditional probability of joint truth than some of its less coherent subsets. They conclude that epistemic justification, as understood in one version of a coherence theory of justification, is not truth conducive. After getting clear about what truth conduciveness requires, I show that their argument does not tell against BonJour's coherence theory.

Philosophers of science as divergent as the inductivist Carnap and the deductivist Popper share the notion that the (logical) content of a proposition is given by its consequence class. I claim that this notion of content is (a) unintuitive and (b) inappropriate for many of the formal needs of philosophers of science. The basic problem is that given this notion of content, for any arbitrary p and q, [(p ∨ q)] will count as part of the content of both p (...) and q. In other words, any arbitrary p and q share some common content. This notion of content has disastrous effects on, for instance, Carnap's attempts to explicate the notion of confirmation in terms of probabilistic favorable relevance, and Popper's attempts to define verisimilitude. After briefly reviewing some of the problems of the traditional notion of content I present an alternative notion of (basic) content which (a) better fits our intuitions about content and (b) better serves the formal needs of philosophers of science. (shrink)

It is shown that the criteria of T-theoreticity proposed by Balzer and Gähde lead to strongly counterintuitive and in this sense paradoxical results: most of the obviously empirical or at least nontheoretical terms come out as theoretical. This is demonstrated for a lot of theories in different areas. On the way, some improved and some new structuralist theory-reconstructions are given. The conclusion is drawn that the T-theoreticity of a term cannot possibly be proved on the basis of the mathematical structure (...) of theory T alone (as Gähde and Balzer suggest). Rather, an independent notion of pre-T-theoreticity and-more importantly-of empiricity is needed; i.e., not empirical and not pre-T-theoretical are independent, necessary but not sufficient conditions for T-theoretical (this is also a necessary complement of Sneed's original criterion). Finally it is asked whether the structuralist criterion of T-theoreticity complemented by such independent conditions would be a satisfactory answer to Putnam's challenge, and the answer again is negative: the criterion is not able to distinguish between empirically contentful and completely contentless (superfluous) theoretical terms. (shrink)