Structuralism as a form of scientific realism

Abstract

Structural realism has recently re‐entered mainstream discussions in the philosophy of science. The central notion of structure, however, is contested by both advocates and critics. This paper briefly reviews currently prominent structuralist accounts en route to proposing a metaphysics of structure that is capable of supporting the epistemic aspirations of realists, and that is immune to the charge most commonly levelled against structuralism. This account provides an alternative to the existing epistemic and ontic forms of the position, incorporating elements of both. Structures are here identified with relations between first order, causal properties: properties that confer specific dispositions for relations. This form of structuralism constitutes an explicit proposal for what seem implicit structuralist tendencies in sophisticated but more traditional characterizations of realism. An outline of the proposal's response to the anti‐realist's pessimistic induction on the history of scientific theories is considered.

Notes

Correspondence: Institute for the History and Philosophy of Science and Technology, Victoria College, University of Toronto, 91 Charles Street West, Toronto, Ontario M55 1K7, Canada. E‐mail: [email protected]

See Gower (2000) for an historical account of early structuralism.

The distinction is due to Ladyman (1998). The same distinction is found in Psillos (2001a) under the labels ‘restrictive’ and ‘eliminative’ SR, respectively.

For a defence of the epistemic view, see Worrall (1989, 1994), Zahar (1996) and Worrall and Zahar (2001). The ontic view is favoured by Ladyman (1998), French (1998, 1999) and French and Ladyman (2003).

Maxwell (1970b, p. 181), like Russell, believes that ‘the observable is instantiated only in inner events of observers’. Contemporary scientific realists generally take observables to be external objects and processes detectable by the unaided senses. However, as Maxwell himself notes, one may adopt the Ramsey sentence approach whatever one's view of where to draw the line separating observables and unobservables.

I am, however, sceptical of these rebuttals. Redhead's argument is open to the criticism that Newman's objection cannot be dissolved merely by invoking objects with first order properties and relations as a postulate. It is only through some detailed knowledge of these properties and relations that we have substantive (as opposed to trivial) knowledge of Russellian structures (cf. Psillos, 2001b, p. 369). Worrall and Zahar's argument depends crucially on the contestable distinction between observable and unobservable terms. They also accept that no empirically adequate Ramsey sentence can be false, which renders Maxwell's approach unattractive to most realists. For a critique of this approach see Ketland (2004).

Compare Redhead (2001a, p. 345): ‘The vague reference to mathematical equations is not sufficient to get a proper handle on this notion [i.e. structure]’.

As will become clear in sections 4–6, the relata I have in mind are determinate properties in the first instance. How kinds of objects are then constituted from kinds of properties is a further question.

This suggests a commitment to universals, but the language of instantiation is merely expedient. One might refer instead to the class of concrete structures. I will continue to speak of properties simpliciter, but everything here might be understood in terms of transcendent universals, immanent universals, tropes or resemblance nominalism.

For detailed views of this sort see Shoemaker (1980) and Swoyer (1982). In these accounts the identities of properties were determined by the dispositions they confer, but this further claim is not crucial to the current proposal.

See Mumford (1998) for a comprehensive survey of accounts of dispositions and a defence of realism, and Cartwright (1989) for a defence of dispositional realism (‘capacities’) in the interpretation of scientific theories.

Maxwell (1970a, n. 19, pp. 33–34) himself mentions dispositions en passant. He states that if we were to redefine higher order property terms by means of ‘a viable causal redefinition’, we could predicate them of physical entities. For example, we might redefine ‘red’ so that it refers not to a property of visual experience, but rather to a disposition on the part of the object that appears red to us. But he concludes that such a disposition, though a structural property, would not be a first order property of the object in question. This conclusion is precisely what I deny.

For details of the physics see French and Redhead (1988), French (1989), van Fraassen (1991) and Huggett (1997). My primary concern is with the possible import of underdetermination, discussed in French (1998), Ladyman (1998) and French and Ladyman (2003).

See also Bueno and French (1999) and French and Ladyman (2003).

See Lavine (1991), who argues for a putatively different conception of quantum particles. Redhead and Teller (1992) and Teller (1995, 2001) support the non‐individuals view, while van Fraassen (1991) appears to dispute it. For work on related issues see Castellani (1998) and Huggett (1997).

French (2003) notes Eddington's view that in the relativistic framework the relata are events, another potentially useful category for collecting groups of causal properties.

Morrison (1990) argues that entity realists cannot restrict knowledge to ‘low level’ generalizations concerning manipulated entities, since in order to generate the causal processes by which they are known we require further theoretical knowledge. I agree; the proposal here will take a different approach.

See Worrall (1989, pp. 153–154) for a discussion of maturity and ad hocness. Psillos (1999, pp. 105–108) discusses ad hoc theories and the importance of novel predictions. See also Leplin (1997).

The distinction between detection and auxiliary properties is central to Chakravartty (1998).

R/I=tan(i – r)/tan(i + r); R′/I′=sin(i – r)/sin(i + r); X/I=(2sinr · cosi)/[sin(i + r) cos(i – r)]; X′/I′=2sinr · cosi/sin(i + r). I ², R ² and X ² are the intensities of the incident, reflected, and refracted components polarized in the plane of incidence; I′², R′² and X′² are the components polarized at right angles to the incident plane; i and r are the angles made by the incident and refracted beams with a normal to the plane of reflection.

This was suggested to me by Steven French (private communication).

Post's examples come primarily from the history of physics. See French and Kamminga (1993) for studies which exemplify and discuss Post's retentionist views, including examples from chemistry and biology.

See Putnam (1978, pp. 22–25). This is one of two strategies suggested by Hardin and Rosenberg (1982).

See Laudan (1983) for a discussion of this and related concerns. Niiniluoto (1997) argues that the similarity theory of approximate truth can be used to make the benefit of the doubt precise, but one still requires a criterion with which to establish when terms have the same reference.