Alexander James Bird

The usual, comparative, conception of inference to the best explanation (IBE) takes it to be ampliative. In this paper I propose a conception of IBE ('Holmesian inference') that takes it to be a species of eliminative induction and hence not ampliative. This avoids several problems for comparative IBE (for example, how could it be reliable enough to generate knowledge?). My account of Holmesian inference raises the suspicion that it could never be applied, on the grounds that scientific hypotheses are inevitably underdetermined by the evidence (i.e. are inevitably ampliative). I argue that this concern may be resisted by acknowledging, as Timothy Williamson has shown, that all knowledge is evidence. The latter suggests an approach to resisting scepticism different from those (e.g. the reliabilist approach) that embrace fallibilism.

Debates concerning the analysis of the concept of law of nature must address the following problem. On the one hand, our grasp of laws of nature is via our knowledge of their instances. And this seems not only an epistemological truth but also a semantic one. The concept of a law of nature must be explicated in terms of the things that instantiate the law. It is not simply that a piece of metal that conducts electricity is evidence for a law that metals conduct electricity. It is also the case that to explicate what it is for there to be such a law requires, and requires little more than, alluding to the fact that the piece of metal conducting electricity is an instance of that law. This is the driving intuition behind regularity theories of laws — to understand the concept ‘law,’ as in ‘it is a law that metals conduct electricity’ one need only understand little more than what it is for something to be a metal and to conduct electricity and the concept of universal generalization. On this view a law just is a regularity (or some kind of regularity) among its instances.

Thomas Kuhn was undoubtedly the strongest influence on the philosophy of science in the last third of the twentieth century. Yet today, at the beginning of the twenty-first century it is unclear what his legacy really is. In the philosophy of science there is no characteristically Kuhnian school. This could be because we are all Kuhnians now. But it might also be because Kuhn’s thought, although revolutionary in its time, has since been superseded. In a sense both may be true. We are all Copernicans—yet almost everything Copernicus believed we now disbelieve. In this paper I shall examine the development of Kuhn’s thought in connection with changes in the philosophy of science during the second half of the twentieth century. Now that philosophy in general, philosophy of science in particular, is in a post-positivist era, we all share Kuhn’s rejection of positivism. But we do not, for the most part, share Kuhn’s belief in incommensurability, or his scepticism about truth and objective knowledge. Just as in Copernicus’ case, Kuhn initiated a revolution that went far beyond what he himself envisaged or even properly understood.

1. Contextualism seeks to acknowledge the power of sceptical arguments while permitting to be true at least some of the assertions of knowledge and justification we commonly make. It seems to me now just as if I am in an office in Edinburgh. According to the sceptic the claim that I am in fact in an office in Edinburgh is unjustified, since there is no reason I can give for this belief that is not also consistent with (or undermined by) the alternative hypothesis that I am in fact on a beach in Hawaii being deceived by an evil demon into thinking that I am in an office in Edinburgh. Yet if I were..

In this paper I argue that it is not a priori that all the laws of nature are contingent. I assume that the fundamental laws are contingent and show that some non-trivial, a posteriori, non-basic laws may nonetheless be necessary in the sense of having no counterinstances in any possible world. I consider a law LS (such as 'salt dissolves in water') that concerns a substance S. Kripke's arguments concerning constitution show that the existence of S requires that a certain deeper level law or variants thereof hold. At the same time, that law and its variants may each entail the truth of LS. Thus the existence of S entails LS. Consequently there is no world in which S exists and fails to obey LS. I consider the conditions concerning the fundamental laws that would make this phenomenon ubiquitous. I conclude with some consequences for metaphysics.

The first obstacle that confronts the student of induction is that of defining the subject matter. One initial point is to note that much of the relevant subject matter goes under the description ‘the theory of confirmation’. The distinction is primarily that the study of induction concerns inference, i.e. cases where one takes the conclusion to be established by the evidence, whereas confirmation concerns the weight of evidence, which one may take to be something like the credibility of a hypothesis in the light of the evidence. Discussions of confirmation often concern incremental confirmation, i.e. cases where the evidence is taken to increase the credibility of some hypothesis, even if not sufficiently to warrant inferring the truth of that hypothesis. However, some uses of ‘confirmation’ clearly refer to absolute confirmation, cases where the credibility of the hypothesis in the light of the evidence exceeds some (high) threshold. One may ask whether inductive inference corresponds to the case of absolute confirmation for some suitable threshold. I shall discuss inference and confirmation together, though it should be noted that some approaches eschew inference altogether. For example, the Bayesian takes scientific reasoning to be a matter of adjusting credences in propositions in the light of evidence, and says nothing about unqualified belief in a proposition. However, if we are interested in inductive knowledge then we must consider inference, since only then do we have a detached proposition that is the possible content of a mental state of knowing. A more pressing question concerns which inferences (or allegedly confirmatory relations) should be classed as inductive. A natural and straightforward approach is to define induction as encompassing any form of reasoning that extrapolates from one population to another, usually from a sample of a population to the whole population. For example, one might note that all observations of the position of some planet fall on an ellipse that has the Sun at one of its foci; from this one concludes that all the positions that planet takes fall on this ellipse (i.e..

Frege's logicism consists of two theses: the truths of arithmetic are truths of logic; the natural numbers are objects. In this paper I pose the question: what conception of logic is required to defend these theses? I hold that there exists an appropriate and natural conception of logic in virtue of which Hume's principle is a logical truth. Hume's principle, which states that the number of Fs is the number of Gs iff the concepts F and G are equinumerous is the central plank in the neo-logicist argument for and. I defend this position against two objections Hume's principle canot be both a logical truth as required by and also have the ontological import required by ; and the use of Hume's principle by the logicist is in effect an ontological proof of a kind which is not valid.

In this paper I examine two aspects of Hempel’s covering-law models of explanation. These are (i) nomic subsumption and (ii) explication by models. Nomic subsumption is the idea that to explain a fact is to show how it falls under some appropriate law. This conception of explanation Hempel explicates using a pair of models, where, in this context, a model is a template or pattern such that if something fits it, then that thing is an explanation. A range of well-known counter-examples to Hempel’s models has led his successors to seek alternatives. Problems with limited amendments have encouraged some theorists of explanation to abandon nomic subsumption. So, in particular, causal components have come to be regarded as essential, even though Hempel had intended his model to capture causal explanation as well.1 Here I want to examine the prospects for retaining nomic subsumption by rejecting the other feature of Hempel’s approach – explication by models. An examination of the counter-examples will suggest that it is a mistake to imagine that a limited quantity of information about laws and antecedent conditions will be able to provide an actual explanation – other information, about explanations, may be relevant. This in turn leads me to examine what I shall call structural approaches. They are structural because the status of something as an explanation depends on its fitting into a structure of explanations. There are two structural approaches I shall examine. One is holistic – it proposes that we consider explanation hand-in-hand with the concept of law. This account of explanation inherits its holistic nature from the holistic (or sys- tematic) character of laws of nature. The second supervenience view I shall consider is not global as the holistic approach is. Instead it concentrates on the ‘vertical’ structure of explanations, whereby the existence of a nomic explanation at one level reflects explanations on lower levels on which it supervenes. These structural approaches were first proposed in Bird Synthese 120: 1–18, 1999. © 1999 Kluwer Academic Publishers..

This paper sketches a dispositionalist conception of laws and shows how the dispositionalist should respond to certain objections. The view that properties are essentially dispositional is able to provide an account of laws that avoids the problems that face the two views of laws (the regularity and the contingent nomic necessitation views) that regard properties as categorical and laws as contingent. I discuss and reject the objections that (i) this view makes laws necessary whereas they are contingent; (ii) this view cannot account for certain kinds of laws of nature and their properties.

It is widely agreed that many causal relations can be regarded as dependent upon causal relations that are in some way more basic. For example, knocking down the first domino in a row of one hundred dominoes will be the cause of the hundredth domino falling. But this causal relation exists in virtue of the knocking of the first domino causing the falling of the second domino, and so forth. In such a case, A causes B in virtue of there being intermediate events I1... In such that A causes I1, I1 causes I2,..., In−1 causes In, and In causes B. Cases of this sort include my putting my foot on the brake causing the car to slow, the smoke from a fire causing the fire brigade to be alerted, and so forth. In other cases the more basic causal relations may not be intermediate (or at least it is controversial that they are). My seeing that it is raining may cause me to want to stay inside, and this causal relation depends upon more basic causal relations among various components of my brain. But it does not seem possible to analyze this in terms of my perception causing certain brain events, which cause other brain events, which eventually cause my desire. Rather it seems as if the principle causal relation, between perception and desire, is constituted, rather than mediated, by the more basic causal relations in the brain. The same is true of the operation of the dynamo causing the current to flow. Again there are not intermediate events, but rather the causal relation between them is constituted by the motion of the charged particles in the wires moving though a magnetic field, which causes an electric field, which causes the charges to move in the wire. There are thus at least two kinds of complex causal relation: the chain kind and the constitution kind. If we wish to understand causation, we need to understand the basic causal relations, at least as found in the chain kind. That is, to understand what it is for A to cause B when the latter is a causal relation of the chain kind, requires understanding what it is for the intermediate, basic causal relations to hold. In the case of a complex causal relation of the constitution kind, it is may be that understanding what it is for A and B to be causally related does not require understanding what it is for the constituting causal relations to hold..