Canadian Journal of Philosophy

I Introduction

Philosophical interest in thought experiments has grown over the last couple of decades. Several positions have emerged, defined largely by their differing responses to a perceived epistemological challenge: how do thought experiments yield justified belief revision, even in science, when they provide no new empirical data? Attitudes towards this supposed explanandum differ. Many philosophers accept that it poses a genuine puzzle and hence seek to provide a substantive explanation.¹ [End Page 55] Others reject or deflate the epistemic claims made for thought experiments.²

In this paper I present a model for many thought experiments in philosophy and science.³ The model doesn’t assume that thought experiments in fact manage to achieve epistemic justification, but it allows us to see how they aspire to do so. It also emphasises both the parallels and the discrepancies between thought experiments and ordinary scientific experiments. And it indicates that there is a systematic mismatch [End Page 56] between the epistemic pretensions of many thought experiments and what they deliver. So although the alleged epistemic merits of thought experiments is not the principal focus of this paper, the view of them that I propose bears on that issue.

The term ‘thought experiment’ is quite vague. What I am interested in, and shall denote by the term, are hypothetical cases intended to function as experiments, in the following sense: they aspire to test hypotheses or theories.⁴ So I will not discuss hypothetical cases intended merely to illustrate or convey a theory, although many of these are plausibly called ‘thought experiments’ too.

The outline of the paper is as follows. In the next section, I defend my stipulative restriction to hypothetical examples functioning as test cases. Section III argues that although thought experiments are not identical to arguments, they have to be seen as intimately connected to certain arguments. In section IV, I make a suggestion concerning the form of such arguments. Section V illustrates how the schema works by applying it to two influential thought experiments, Putnam’s Twin Earth case and Einstein’s clock-in-a-box thought experiment. The final section draws some morals concerning the epistemic value of thought experiments in general.

To some, this paper may seem under-ambitious in that it doesn’t head-on address the issues where most of the action is in current work on thought experiments: their epistemic merits and their underlying psychological implementation.⁵ However, I believe that having a — somewhat abstract — framework for thinking about what thought experiments aim at and how they work is both a useful propedaeutic to discussing questions of merit or implementation, and interesting in its own right. It is also useful for thinking about the dialectical situation regarding particular thought experiments. And as will become clear in the final section, the model itself does offer some insights about the reliability of thought experiments.

II What are these things called thought experiments?

It may seem that anyone setting out to make general claims about the workings of thought experiments ought to start by considering what they are. If this is taken to involve a general characterisation of the [End Page 57] plethora of things that have on some occasion been called ‘thought experiments,’ however, the task seems both daunting and misguided. For these are very different things indeed, ranging from mathematical arguments, pre-Socratic reasoning⁶ and Husserlian eidetic variation⁷ to Harvey’s discovery of the circulation of the blood.⁸ It seems quite obvious that the class of things to which the term ‘thought experiment’ has been applied does not constitute any natural kind or category.

Hence I am also skeptical of James Robert Brown’s optimism when he states that we may rely on our intuitive ability to correctly sort cases into the extension and counterextension of the term, and therefore need not bother with defining it. Brown writes: ‘We know them when we see them, and that’s enough to make talking about them possible.’⁹ He goes on to argue that we shouldn’t build into our conception of thought experiments e.g. that they involve idealization or be essentially...