Francis Bacon (1561–1626) wrote that good scientists are not like ants (mindlessly gathering data) or spiders (spinning empty theories). Instead, they are like bees, transforming nature into a nourishing product. This essay examines Bacon's “middle way” by elucidating the means he proposes to turn experience and insight into understanding. The human intellect relies on “machines” to extend perceptual limits, check impulsive imaginations, and reveal nature's latent causal structure, or “forms.” This constructivist interpretation is not intended to supplant inductivist or experimentalist interpretations, but is designed to explicate Bacon's account of science as a collaborative project with several interdependent methodological goals.

This paper examines some common measures of complexity, structure, and information, with an eye toward understanding the extent to which complexity or information-content may be regarded as objective properties of individual objects. A form of contextual objectivity is proposed which renders the measures objective, and which largely resolves the puzzle of Maxwell's Demon.

Reflection and martingale principles are central to models of rational learning. They can be justified in a variety of ways. In what follows we study martingale and reflection principles in the context of measure theory. We give special attention to two approaches for justifying these principles that have not been studied in that context before: diachronic coherence and the value of information. Together with an extant argument based on expected accuracy, these arguments lend support to the thesis that reflection and martingale principles govern rational learning.

The disagreement with regard to the validity of the principle of causality, existing to-day among scientists, has its roots in the diversity of definitions of the principle itself rather than in a problematic scientific situation. As far as the formulation of quantum theory is complete its bearing upon philosophical questions can be fixed with precision provided the questions are phrased intelligibly. But a question is intelligible from a scientific point of view only if it satisfies two conditions: (1) the meaning of its terms must be fixed; (2) it must be in accord with the conventions of the science to which the question is put.

In this paper I consider two objections raised by Nick Smith (1997) to an argument against the probability of time travel given by Paul Horwich (1995, 1987). Horwich argues that time travel leads to inexplicable and improbable coincidences. I argue that one of Smith's objections fails, but that another is correct. I also consider an instructive way to defend Horwich's argument against the second of Smith's objections, but show that it too fails. I conclude that unless there is something faulty in the conception of explanation implicit in Horwich's argument, time travel presents us with nothing that is inexplicable.

This paper contrasts two information-theoretic approaches to statistical explanation: namely, (1) an analysis, which originated in my earlier research on problems of testing stochastic models of learning, based on an entropy-like measure of expected transmitted-information (and here referred to as the Expected-Information Model), and (2) the analysis, which was proposed by James Greeno (and which is closely related to Wesley Salmon's Statistical Relevance Model), based on the information-transmitted-by-a-system. The substantial differences between these analyses can be traced to the following basic difference. On Greeno's view, the essence of explanation lies in the relevance relations expressed by the conditional probabilities that relate the explanans variables to the explanandum variables; on my view, in contrast, the essence of explanation lies in theories viewed as hypothetical structures which deductively entail conditional probability distributions linking the explanans variables and the explanandum variables. The explanatory power of a stochastic theory is identified with information (regarding the values of explanandum variables) which is “absorbed from” the explanans variables. While other information which is “absorbed from” the explanandum variables (through the process of parameter estimation, for example) reflects descriptive power of the theory. I prove that Greeno's measure of transmitted information is a limiting special case of the E-I model, but that the former, unlike the latter, makes no distinction between explanatory power and descriptive power.

Some have suggested that certain classical physical systems have undecidable long-term behavior, without specifying an appropriate notion of decidability over the reals. We introduce such a notion, decidability in μ (or d-μ) for any measure μ, which is particularly appropriate for physics and in some ways more intuitive than Ko's (1991) recursive approximability (r.a.). For Lebesgue measure λ, d-λ implies r.a. Sets with positive λ-measure that are sufficiently “riddled” with holes are never d-λ but are often r.a. This explicates Sommerer and Ott's (1996) claim of uncomputable behavior in a system with riddled basins of attraction. Furthermore, it clarifies speculations that the stability of the solar system (and similar systems) may be undecidable, for the invariant tori established by KAM theory form sets that are not d-λ.

Ronald Giere's analysis of causal effectiveness in populations involves the comparison of two hypothetical populations, one in which every individual has the suspected causal factor, and the other in which none do. Elliott Sober has argued that in cases where causal effectiveness depends on relative population sizes, Giere's analysis breaks down. I take issue with this claim, and argue to the contrary that Giere's analysis can help provide insight into these cases.

The Doomsday argument and anthropic reasoning are two puzzling examples of probabilistic confirmation. In both cases, a lack of knowledge apparently yields surprising conclusions. Since they are formulated within a Bayesian framework, they constitute a challenge to Bayesianism. Several attempts, some successful, have been made in a Bayesian framework that represents credal states by single credence functions to avoid these conclusions, but none of them can do so for all versions of the Doomsday argument. I show that adopting an imprecise framework of probabilistic reasoning allows for a more adequate representation of ignorance and explains away these puzzles.

A common criticism of punctuated equilibria as an evolutionary theory is that it erects a straw man by characterizing the modern synthesis as being devoid of mechanisms that bring about rapid speciation and abrupt changes in morphology. Thompson supports this view and argues that the modern synthesis does not entail gradualism, all-pervasive adaptationism, or extrapolationism and that punctuationists have mischaracterized the theory on all these points; properly understood the synthetic theory is hierarchical and able to explain phenomena at all levels of the hierarchy, thus rendering macroevolutionary theories, such as punctuated equilibria, unnecessary. I argue in this paper that Thompson's approach is overly dependent upon rational reconstruction in the style of the logical empiricists, and as such ignores important sociological and historical factors that when taken into account justify punctuational criticism of the synthetic theory.

The Aristotelian distinction between actual and potential presence of a substance in a mixture forms part of a conception of mixture which stands in contrast to atomist and Stoic theories as propounded by the ancients. But the central ideas on which these theories are built need not be combined and opposed to one another in precisely the ways envisaged by these ancient theories. This is well illustrated by Duhem, who maintained the Aristotelian idea that the original ingredients are only potentially, and not actually, present in a mixture, but who sided with the Stoics and against Aristotle on the possibility of co-occupancy. I have argued that the Stoic theory cannot dispense with some such notion as the Aristotelian conception of potentiality in allowing the elements to be actually present in a mixture. Here I suggest that some such Aristotelian notion must be at work in a more modern atomic conception of matter if this is to allow elemental substances to be actually present in compounds, which I think is how compounds are usually understood. Analogous issues arise regarding the status of solutions and their components.

Quine and Putnam argued for mathematical realism on the basis of the indispensability of mathematics to science. They claimed that the mathematics that is used in physical theories is confirmed along with those theories and that scientific realism entails mathematical realism. I argue here that current theories of confirmation suggest that mathematics does not receive empirical support simply in virtue of being a part of well confirmed scientific theories and that the reasons for adopting a realist view of scientific theories do not support realism about mathematical entities, despite the use of mathematics in formulating scientific theory.

The literary and poetic turn in philosophy exudes contempt for science and hostility against technology, both allegedly justified on grounds established by Nietzsche and Heidegger. I try to show that these grounds instead call for an extremely positive assessment of science and technology. It turns out that what Nietzsche and Heidegger describe as our highest achievement, namely, human autonomy, is really made possible by modern experimental physics. And I show how this assessment is borne out by what Heisenberg describes as the lesson of quantum physics.

Case studies of diverse scientific fields show how scientists use a range of resources to generate new interpretative models and to establish their plausibility as explanations of a domain. They accomplish this by manipulating imagistic representations in particular ways. I show that scientists in different domains use the same basic transformations. Common features of these transformations indicate that general cognitive strategies of interpretation, simplification, elaboration, and argumentation are at work. Social and historical studies of science emphasize the diversity of local contexts of practice. However, the existence of common strategies shows that this diversity masks an important repertoire of cognitive strategies. Scientists use this repertoire to adapt their representations to meet the cognitive demands of different contexts of practice. This paper considers the implications of this finding for the notion of scientists as cognitive agents in distributed knowledge-producing systems.

I defend the theory that the reasons why some event occurred are its causes. Many “counterexamples” to this theory turn on confusing two levels of reasons. We should distinguish the reasons why an event occurred (“first-level reasons”) from the reasons why those reasons are reasons (“second-level reasons”). An example that treats a second-level reason as a first-level reason will look like a counterexample if that second-level reason is not a cause. But second-level reasons need not be first-level reasons.

From late 1684 through mid-1685, Isaac Newton turned to developing and refining the conceptual foundations presupposed by his emerging physics. Analysis of his manuscripts from this period reveals that Newton’s understanding of the relativity of acceleration led him to seek a spatiotemporally invariant quantity of matter. He found two such quantities and then designed an experiment to discover their relationship. Interpreting the experiment, however, required distinguishing a new notion of force. Others have recognized the conceptual distinction between inertial and gravitational mass. I show here that Newton clearly saw this distinction and that he provisionally established their equivalence.

The requirement of total evidence is a mainstay of Bayesian epistemology. Peter Fisher Epstein argues that the requirement generates mistaken conclusions about several examples that he devises. Here we examine the example of Epstein’s that we find most interesting and argue that Epstein’s analysis of it is flawed.

1. Outline of the Project. This paper is intended to be a progress report on a project in philosophy of science. The immediate stimulus of this report is the eightieth birthday of E. A. Singer, Jr., who was the inspiration of the project, and, needless to say, though responsible for the whole is not responsible for the misconceptions in the specific parts.

Positivists and phenomenalists of all sorts maintain, and long have maintained, some variant of the following thesis concerning the existence of physical objects: Such statements as ‘There is (exists) now a wall behind my back’ are synonymous with a class of statements of which the following is representative ‘If I shall turn my head (have certain kinesthetic experiences), then I shall also have the visual experience called ‘seeing a wall'.’ This amounts to proposing what many of us call a philosophical analysis of ‘exist’ or, more precisely, of one meaning of ‘exist'; for the thesis implies that this verb, in the sense in which we use it when we say ‘This wall exists', is dispensable in the sense of being definable. Realists, who oppose the thesis, hold that ‘exist', in the sense mentioned, is what I would call an undefined descriptive predicate; and then they go on to recommend that instead of defining existence in terms of what we (shall) see, we had better say that we shall, if we turn, see a wall because there is a wall (and because we have put ourselves in a position to perceive it). Thus one could say, perhaps, that the realists wish to convert the positivistic position; instead of founding existence upon experience, they want to found experience (among other things) upon existence. To be sure, this is but a bare and crudely formulated schema of an issue that has been argued for a long time. Also, contemporary analysts for the most part do not discuss the issue directly; they give their attention to preliminary and, therefore, more fundamental questions. Like experienced chess players, who know to which characteristic situations in the middle game certain openings will eventually lead, we try to convince each other of the excellence of our respective openings. There is, in particular, one fundamental or opening move that is now widely discussed and which, I believe, most of us examine with a view to the position in which we shall find ourselves—in the middle game—with respect to the realism issue. I refer to the clarification of the relations between meaning and verification or, to put it the way positivists do, to the formulation of an adequate meaning criterion. Let me indicate the connection between these questions and the crude schema I have given for the realism issue. According to current garden varieties of positivism, a statement is (empirically) meaningful if it is verifiable by (future) experience. ‘There is a wall behind my back’ is, in this view, meaningful because it is synonymous with a class of statements each of which is verifiable by future experience. More pointedly, the familiar common sense statements about the existence of physical objects are considered as meaningful because they are verifiable. This has again the subjectivistic ring realists dislike so much; so they may again be inclined to convert the positivistic position, if only tentatively and because they feel that a piece of realistic common sense would thus be preserved. But to make the conversion is to say that ‘There is a wall behind my back’ is verifiable because it is meaningful.