There is a very plausible transitivity principle for theory choice. It says that if all criteria of theory evaluation are considered, and theory A is a better theory than theory B, and theory B is a better theory than theory C, then A is a better theory than C. I argue against this principle. It turns out that whenever there are two or more relevant and independent criteria of theory evaluation, and that whenever at least of one the criteria is ‘nonlinear’ in a certain sense, there may be violations of transitivity that do not violate any standards of rationality (of theory choice). This shows, again, that theory choice cannot be seen as merely the application of given rules of rational theory choice.

Faced with the paradox of undermining futures, Humeans have resigned themselves to accounts of chance that severely conflict with our intuitions. However, such resignation is premature: The problem is Humean Supervenience (HS), not Humeanism. This paper develops a projectivist Humeanism on which chance claims are understood as normative, rather than fact stating. Rationality constraints on the cotenability of norms and factual claims ground a factual-normative worlds semantics that, in addition to solving the Frege-Geach problem, delivers the intuitive set of possibilia for each chance law. Hence, the account does not entail HS, and the paradox does not arise. A confirmation theory is developed, and the Principal Principle is justified.

Geometry was a main source of inspiration for Carnap's conventionalism. Taking Poincaré as his witness, Carnap asserted in his dissertation Der Raum ([1922] 1978) that the metrical structure of space is conventional while the underlying topological structure describes ‘objective’ facts. With only minor modifications he stuck to this account throughout his life. The aim of this paper is to disprove Carnap's contention by invoking some classical theorems of differential topology. It is shown that his metrical conventionalism is indefensible for mathematical reasons. This implies that the relation between topology and geometry cannot be conceptualized as analogous to the relation between the meaning of a proposition and its expression in some language as the logical empiricists used to say.

Two central and well-known philosophical goals of the logical empiricists are the unification of science and the elimination of metaphysics. Textual analysis shows, however, that these two apparently distinct planks of the logical empiricist party platform are actually intimately related. From the 1920s through 1950, one abiding criterion for judging whether an apparently declarative assertion or descriptive term is metaphysical is that that assertion or term cannot be incorporated into a language of unified science. I explore various versions of this criterion throughout the works of Carnap and Neurath.

Scientists typically use a variety of representations, including different kinds of figures, to present and defend hypotheses. In order to understand the justification of scientific hypotheses, it is essential to understand how visual representations contribute to scientific arguments. Since the logical understanding of arguments involves the truth or falsity of the representations involved, visual representations must have the capacity to bear truth in order to be genuine components of arguments. By drawing on Goodman's analysis of symbol systems, and on Tarski's work on truth, I show that the figures used in science meet this criterion.

In the nineteenth century, William Whewell claimed that his confirmation criterion of consilience was a truth-guarantor: we could, he believed, be certain that a consilient theory was true. Since that time Whewell has been much ridiculed for this claim by critics such as J. S. Mill and Bas van Fraassen. I have argued elsewhere that, while Whewell's claim that consilience can guarantee the truth of a theory is clearly wrong, consilience is indeed quite useful as a confirmation criterion (Snyder 2005). Here I will show that, even when consilience gives evidence for a theory that turns out to be false, there is an important sense in which consilience shows that the theory has captured something correct about the natural-kind structure of the physical world. Whewell was therefore correct to claim that consilience provides a “criterion of reality” (Whewell [1847] 1967, vol. 2, 68). Consilience provides this by giving justification for the claim that we have really ‘cut nature at its causal joints’, to adapt Plato's famous phrase. Because of this, consilience can play a role in an argument for scientific realism.

The kinds of models discussed in this paper function as measuring instruments. We will concentrate on two necessary steps for measurement: (1) one should search for a mathematical representation of the phenomenon; (2) this representation about the phenomenon should cover an invariant relationship between properties of the phenomenon to be measured and observable associated attributes of a measuring instrument. Therefore, the measuring instrument should function as a nomological machine. However, invariant relationships are not necessarily ceteris paribus regularities, but could also occur when the influence of the environment is negligible. Then we are able to achieve accurate measurements outside the laboratory.

Starting with a discussion of what I call ‘Koyré's paradox of conceptual novelty’, I introduce the ideas of Damerow et al. on the establishment of classical mechanics in Galileo's work. I then argue that although their view on the nature of Galileo's conceptual innovation is convincing, it misses an essential element: Galileo's use of the experiments described in the first day of the Two New Sciences. I describe these experiments and analyze their function. Central to my analysis is the idea that Galileo's pendulum experiments serve to secure the reference of his theoretical models in actually occurring cases of free fall. In this way, Galileo's experiments constitute an essential part of the meaning of the new concepts of classical mechanics.

I develop a concept of observability that pertains to qualities rather than objects: A quality is observable if it can be registered by human sensation (possibly with the aid of instruments) without involving optional interpretations. This concept supports a better description of observations in science and everyday life than the object-based observability concepts presupposing causal information-transfer from the object to the observer. It also allows a rehabilitation of the traditional empiricist distinction between observations and their interpretations, but without a presumption that observations are infallible. Using this concept of observability, I also propose a re-formulation of constructive empiricism that is easier to defend against realist attacks, while open to reasonable realist intuitions.

Philosophers of experiment have acknowledged that experiments are often more than mere hypothesis-tests, once thought to be an experiment's exclusive calling. Drawing on examples from contemporary biology, I make an additional amendment to our understanding of experiment by examining the way that `wide' instrumentation can, for reasons of efficiency, lead scientists away from traditional hypothesis-directed methods of experimentation and towards exploratory methods.

Standard statistical measures of strength of association, although pioneered by Pearson deliberately to be acausal, nowadays are routinely used to measure causal efficacy. But their acausal origins have left them ill suited to this latter purpose. I distinguish between two different conceptions of causal efficacy, and argue that: (1) Both conceptions can be useful; (2) The statistical measures only attempt to capture the first of them; (3) They are not fully successful even at this; (4) An alternative definition based more squarely on causal thinking not only captures the second conception, but also can capture the first one better too.