Approximate coherentism suggests that imperfectly rational agents should hold approximately coherent credences. This norm is intended as a generalization of ordinary coherence. I argue that it may be unable to play this role by considering its application under learning experiences. While it is unclear how imperfect agents should revise their beliefs, I suggest a plausible route is through Bayesian updating. However, Bayesian updating can take an incoherent agent from relatively more coherent credences to relatively less coherent credences, depending on the data observed. Thus, comparative rationality judgments among incoherent agents are unduly sensitive to luck.

Only one philosopher has succeeded in building his reputation on silence—Cratylus. It is said that under no circumstance would he say anything, but would merely crook his finger. Nor is it known whether he achieved this unique glory by a persistence that lasted from toothless youth to toothless old age, or whether he merely petered out into withering speechlessness before the follies of man and the grandeur of God. More likely it was something like the latter, for otherwise we would be confronted by the delicate riddle: why was he called a philosopher?

Nevertheless, a tenable explanation is that Cratylus' reputation was arrived at by an experiment. Arranging philosophers on a scale of self .control in speech, one might derive the plausible law: the less one says the more he is of a philosopher. Ergo, say nothing and be declared the worthiest of philosophers. The difficulties in such an extrapolation, however, are very great. It is hard to see what harm just one word can do, and even two or three, or several, if they but fall short of being a sentence. Why absolute silence? Then, again, what harm is there in repeating even a multitude of words, the entire dictionary if you please, plus whatever sounds you can think of, provided you say nothing? And since we have progressed so far with the argument, what harm can there be even in connected speech, grammatical, fluent and mellifluous, provided you still say nothing? Cratylus, apparently, was playing a safe bet with a vengeance. He must have feared being understood (or misunderstood) by accident!

Several aspects of the work of George H. Mead have attracted a rather considerable amount of attention in recent years. The attention devoted by several disciplines to the relevant aspects of his work, however, has not involved an explicit focusing upon his philosophy of science. This is deemed regrettable, at least by this writer, in view of the fact that Mead's philosophy of science appears to be a rather substantial conceptualization of the actual research process.

How does a complex organism develop from a relatively simple, homogeneous mass? The usual answer is: through the (context-dependent) execution of species-specific genetic instructions specifying the development of that organism. Commentators are sometimes skeptical of this usual answer, but of course not all commentators, and not always for the same reasons. Here I attempt to lay bare the logical structure of the usual answer through an extended analysis of the heuristics and methodological principles at play in the exploration and explanation of development—and also to show a critical ambiguity that renders the usual answer suspect.

Jonathan Livengood and Justin Sytsma have published a series of studies on “Actual Causation and Compositionality,” in which they investigate causal attributions of laypeople. We use one of their vignettes to follow up on their research. Our findings cast doubt on their conclusion that ordinary causal attributions tend to violate the compositionality constraint if one looks at cases in which someone is responsible for an effect by way of an intermediary that does not share in the responsibility.

This paper does not purport to present a fully developed argument. Rather, it intends to delineate, in broad strokes and in a synoptic manner, the total frame of reference of Scientific Empiricism as far as it is necessary for an integration of the several contributory clarifications of the complex, usually referred to as the psychophysical problem. Several considerations seem to justify such an attempt. It is desirable, from time to time, to free the results of logical and methodological analysis from their specialized contexts. Only thus can they be presented in a way which brings to the fore their continuity with more traditionally worded epistemological thought. By doing this such a presentation should also help to dissolve the current objections of formalism, scientism, and philosophical irrelevance so frequently based on misunderstanding. It is only fair to admit, however, that these misunderstandings have been facilitated by the preoccupation with the technical, and the brevity with regard to the nontechnical, which characterize the modern analysts.

In order to determine what are the metaphysical presuppositions of science (if any), we have to first define the terms ‘presupposition’ and ‘metaphysics’ or ‘metaphysical’. We shall begin with the former.

Arthur Pap in his article Does Science Have Metaphysical Presuppositions? discusses two senses of the term. A presupposition may be (a) a necessary condition: p presupposes q, dt = p implies q, i.e. not-g implies not-p. (b) a premise or a rule (or a set of rules) of inference used in deriving a belief. Thus in this sense, p presupposes q means that if q were false, then the belief in p would cease to be grounded.

A physical theory intends to provide us with a picture of a certain domain of physical phenomena. We have become accustomed to deal rather with physical pictures than with phenomena themselves. This is indispensable, even if the number of phenomena contained in the picture is so large that we cannot have them in mind all at the same time or if the phenomena are not directly accessible to our senses and cannot be studied without the help of a set of very complicated apparatus.

Ramsey's device [2], for eliminating theoretical terms from a theory assumes that the theory has only finitely many axioms. Paul Berent [1], has recently suggested a modification of Ramsey's method that is designed to cover the infinite case. But, his modified method is quite unlike Ramsey's in that it involves an ascent from the original theory to a suitable metatheory. Thus, it is perhaps of some interest to note that there is an extension of Ramsey's method that covers the infinite case and at the same time is much closer in spirit to his original suggestion.

Comparative genomicists seem to be convinced that the unit of measurement employed in their studies is a gene that drives the function of cells and ultimately organisms. As a result, they have come to some substantive conclusions about how similar humans are to other organisms based on the percentage of genetic makeup they share. I argue that the actual unit of measurement employed in the studies corresponds to a structural rather than a functional gene concept, thus rendering many of the implications drawn from comparative genomic studies largely unwarranted, if not completely mistaken.

This paper is an assessment of an attempt, by James Greeno, to measure the explanatory power of statistical theories by means of the notion of transmitted information (It). It is argued that It has certain features that are inappropriate in a measure of explanatory power. In particular, given a statistical theory T with explanans variables Si and explanandum variables Mj, it is argued that no plausible measure of explanatory power should depend on the probability P(Si) of occurrence of initial conditions in the systems to which T applies or the magnitudes of the conditional probabilities P(Mj/Si), in the manner in which IT does.

Alasdair MacIntyre's expression of the “profound gratitude we all owe to Althusser for having brought French Marxism back into dialogue with the rest of French philosophy” (as quoted by Elliott in his introduction to the selection under discussion, 1990, xi), important as it may be for a host of reasons, did not try to make Althusser's philosophy particularly attractive to philosophers of science. However, the present selection of essays does precisely this: It is almost ideally designed to mark the beginnings of an effective encounter of this particular brand of “French Marxism” with philosophy of science.

Relationist theories of space or space-time based on embedding of a physical relational system A into a corresponding geometrical system B raise problems associated with the degree of uniqueness of the embedding. Such uniqueness problems are familiar in the representational theory of measurement (RTM), and are dealt with by imposing a condition of uniqueness of embeddings up to composition with an “admissible transformation” of the space B. Friedman (1983) presents an alternative treatment of the uniqueness problem for embedding relationist theories, developed independently of RTM. Friedman's approach differs from that of RTM in securing uniqueness by adding new primitives to the physical system A in contrast to the RTM approach which adds new axioms. Friedman's proposal has recently been developed and defended by Catton and Solomon (1988).

This method of solving the uniqueness problem is here argued to be substantially inferior to the RTM method, both in practice and in principle. In practice we find that in none of the concrete examples offered to illustrate the method is the uniqueness problem actually solved in general. Moreover we find that in the most interesting case (addition to the system A of a finite number of relations of finite degree) the method is in principle incapable of success for mathematical reasons. In addition to these technical difficulties there are compelling methodological reasons for preferring the RTM method to the method of adding primitives.

Carnap's system of inductive logic (1) has very often been criticized on the ground that “degree of confirmation” is defined only for languages which are extremely over-simplified. Allegedly, it would be very difficult—and perhaps impossible—to define it adequately for languages formalized within the higher predicate calculi, or languages equivalent to these in richness, and it is such languages that would be needed were we ever to formalize the language of empirical science as a whole. Thus, this criticism bears not only on Carnap's work, but on all attempts to construct an exact analysis of induction from this particular standpoint; that is, from the standpoint of “logical measure functions.”

We consider the nature of quantum randomness and how one might have empirical evidence for it. We will see why, depending on one’s computational resources, it may be impossible to determine whether a particular notion of randomness properly characterizes one’s empirical data. Indeed, we will see why even an ideal observer under ideal epistemic conditions may never have any empirical evidence whatsoever for believing that the results of one’s quantum-mechanical experiments are randomly determined. This illustrates a radical sort of empirical underdetermination faced by fundamentally stochastic theories like quantum mechanics.

After briefly presenting Ronald Giere's (1979, 1980) recent counterfactual characterization of population-level causation, I present two counterexamples to the characterization. The difficulty discussed stems from nonaccidental correlations that can obtain between causally effective and causally neutral factors.

Two essentially different approaches to “truth” should be recognized, although they are not in every respect and on all occasions mutually exclusive. If truth be taken to mean the “most adequate idea (statement, explanation, interpretation, expression) a mind (plural number of minds) is able to perceive at any given time,” the process may, first, be one of immediate intuition, insight or inspiration in which conclusions are arrived at without consciousness of the mediate, detailed steps between recognition of a problem and its solution and without considering all of the possible solutions to the problem; or, it may, second, be one of detailed examination and traversal of the various direct and simple steps to the “truth”, as well as the testing of all other methods of arriving at the same truth. The first method may bring about its result “in a flash”, instantaneously, unpredictably; but the second method is never so rapid, always more laborious, and always consumes more time and energy than the first. That is not to say that the first method always involves less elapsed time between inception and completion than the second; often the first method is marked by delays because the mind in question leaps to other trains of thought before obtaining a solution to the first problem. However, the amount of total time used exclusively to answer the question or solve the problem by the first method is much less than for the second.

Wesley Salmon provided three classic criteria of adequacy for satisfactory interpretations of probability. A fourth criterion is suggested here. A distinction is drawn between frequency-driven probability models and theory-driven probability models and it is argued that single case accounts of chance are superior to frequency accounts at least for the latter. Finally it is suggested that theories of chance should be required only to be contingently true, a position which is a natural extension of Salmon's ontic account of probabilistic causality and his own later views on propensities.