Notes on van Fraassen's Constructive Empiricism

Philosophy of Science
Curtis Brown

I. Background: Logical Positivism

The logical positivists were anti-realists about so-called "theoretical entities." A. J. Ayer, for instance, explicitly defends a version of phenomenalism not far removed from Berkeley's in Language, Truth, and Logic. The general idea was that we could distinguish between two kinds of vocabulary, observational and theoretical.  Observational terms simply corresponded to particular experiences or "sense-data", while theoretical terms only acquired meaning by being defined in terms of observation terms. Thus theoretical terms, e.g. "electron," actually referred, not to tiny particles too small to see, but rather to collections of experiences!  (Or, somewhat more precisely, to "logical constructions" out of experiences.)

Maxwell actually discusses three somewhat different versions of the positivist approach. They all have in common that theoretical terms do not refer to unobservable entities. (1) fictionalism: theoretical terms don't really refer to unobservables, but it's convenient to pretend that they do, so in practice we maintain the fiction that they do. When we're being careful, however, we'll admit that they don't. (2) phenomenalism: theoretical terms can be defined in terms of observational ones. As a result, theoretical terms refer, not to unobservables, but to (logical constructions out of) "sense data". (3) instrumentalism: theoretical terms are literally meaningless; they are merely useful tools or devices ("instruments") for helping us to make predictions. In passing (in discussing the fictional psychologist "Pelter"), Maxwell also mentions a fourth view: (4) theoretical terms may or may not refer to anything real, but either way it doesn't matter: methodologically science will be better off if it simply doesn't use such terms, and instead limits itself to making connections between observable phenomena. [Pop quiz: what real psychologist is "Pelter" supposed to make us think of?]

II. Maxwell

Maxwell, responding to the positivist account, argues that there is no sharp distinction between observational and theoretical vocabulary, so that any reasons we have for being realists about the entities referred to by observational vocabulary also apply to entities referred to by theoretical vocabulary. (See my separate handout on Maxwell for a summary of his arguments.)

III. van Fraassen

van Fraassen wants to return to a view in some ways closer to that of the positivists -- a view which is realistic about some sorts of entities but not realistic about others. However, van Fraassen's view concedes that in some ways the positivist version of empiricism was unsatisfactory.

Here are some key features of van Fraassen's view:

1. It is a mistake to speak of an "observational/theoretical" distinction. There are really two different distinctions at work here: (a) a linguistic distinction between theoretical and nontheoretical vocabulary, and (b) an epistemological distinction between observable and unobservable entities. van Fraassen agrees with Maxwell that the theoretical/nontheoretical distinction is bogus, because all our vocabulary is "theory-laden." Nevertheless, he argues that there a real and important distinction between observable and unobservable entities. Observable entities are entities which can be directly detected by the human sensory apparatus; unobservable entities are entities which cannot be directly detected by the human sensory apparatus, although they may produce effects that can be observed (e.g. we cannot see an electron in a cloud chamber directly, although we may see the trail it leaves).

2. van Fraassen agrees with realists that we should interpret scientific claims about unobservables literally. The positivists were forced to interpret such claims as really being about sensory experiences, but van Fraassen thinks that this is a mistake.

3. van Fraassen says that we should understand realism as holding that it is part of the aim of science that the claims it makes about unobservable as well as observable entities are literally true. Realism holds further that to accept a scientific theory is to believe that these claims are true. To understand the contrast he draws between realism and constructive empiricism, we need to know what he means by empirical adequacy:  a theory is empirically adequate if the claims it makes about observable entities (whether actually observed or not) are true, regardless of whether its claims about unobservable phenomena are true or false. With this notion in hand, van Fraassen can explain that constructive empiricism holds that science aims at theories that are empirically adequate (rather than true simpliciter), and that the only belief involved in acceptance of a scientific theory is belief that the theory is empirically adequate. Thus the constructive empiricist agrees with the realist concerning observable entities, but is an agnostic about scientific claims about unobservables.

4. A feature of van Fraassen's view that some find puzzling is the sharp contrast he draws between unobservable entities, and entities that are observable but have not actually been observed. It might seem that van Fraassen's arguments that we should have beliefs about the truth of claims about observable things that we have not yet observed would push him in the direction of full-fledged realism, while his arguments for not having beliefs about unobservable entities would push him toward skepticism about anything we haven't observed. It is an interesting issue how stable his position is; one might think it will have a tendency to tip in one direction or the other.

A theological analogy van Fraassen mentions briefly may be helpful in thinking about his view. Scientific realism is somewhat like religious fundamentalism: the doctrines in question are interpreted literally, and believed to be true. Some forms of anti-realism, e.g. fictionalism, are analogous to atheism: the doctrines are interpreted literally, and believed to be false. Still others are like liberal theology (e.g. Bultmann's "demythologized" Christianity): the doctrines in question are held to be true, but only if interpreted nonliterally. van Fraassen's version of anti-realism is most like agnosticism, which interprets the doctrines literally but withholds both belief and disbelief.

claims about unobservables are:



can't tell (and don't care)

interpreted literally

[cf. fundamentalism]
[cf. atheism]
constructive empiricism
[cf. agnosticism]

interpreted nonliterally

[cf. liberal theology]


Curd, Martin, J.A. Cover, and Christoher Pincock, Philosophy of Science: The Central Issues, Second Edition. New York: Norton, 2013.

van Fraassen, Bas. "Arguments Concerning Scientific Realism." In Curd et al., 2013. (First published 1980, in van Fraassen's book The Scientific Image.)

Maxwell, Grover. "The Ontological Status of Theoretical Entities." In Curd et al., 2013. (First published 1962.)

Last update: September 12, 2016.
Curtis Brown | Philosophy of Science | Philosophy Department | Trinity University