Jump to ContentJump to Main Navigation
Putting Metaphysics FirstEssays on Metaphysics and Epistemology$

Michael Devitt

Print publication date: 2009

Print ISBN-13: 9780199280803

Published to Oxford Scholarship Online: February 2010

DOI: 10.1093/acprof:oso/9780199280803.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 23 June 2018

Underdetermination and Commonsense Realism *

Underdetermination and Commonsense Realism *

(p.57) 3 Underdetermination and Commonsense Realism*
Putting Metaphysics First

Michael Devitt (Contributor Webpage)

Oxford University Press

Abstract and Keywords

This chapter defends commonsense realism from the underdetermination theses of extreme skepticism. If true, these would count against nearly all our knowledge and hence undermine realism about the observable world. The traditional responses to these theses rest on a priori knowledge. The chapter rejects a priori knowledge, but even if there were such knowledge, these traditional responses tend to involve bizarre metaphysics and to be otherwise unsatisfactory. Instead, it offers a Moorean response: realism is much more firmly based than the epistemological theses that are thought to undermine it. The Moorean response is supported by a naturalistic one that appeals to scientific practice. Rather than proceeding form an a priori epistemology to an a priori metaphysics, we should proceed from an empirical metaphysics to an empirical epistemology. We should put metaphysics first.

Keywords:   underdetermination, commonsense realism, skepticism, a priori, Moore, naturalism, Moorean response

1. Introduction

The underdetermination of theories by evidence often leads to skepticism about the theories and hence to antirealism about the worlds described by the theories. At its most extreme, this skepticism is about “Commonsense Realism”, about our knowledge of the observable world of stones, trees, cats, and the like. This skepticism must spread to “Scientific Realism”, to our knowledge of the unobservable world of atoms, viruses, photons, and the like. But underdetermination sometimes leads to a less extreme view aimed only at Scientific Realism: against a background acceptance of realism about the observable world, realism about unobservables is rejected. I shall set this criticism of Scientific Realism aside until Chapter 4 of this volume.

The underdetermination theses that concern us claim that a theory (belief) has rivals that stand in some sort of equivalence relation to it with respect to certain evidence. The theses vary with the equivalence relation and with the evidence. In particular, sometimes the relation is deductive and sometimes ampliative; and sometimes the evidence is the actual given evidence and sometimes it is some sort of possible evidence.

There are two dimensions to the realisms that are challenged by underdetermination. “The existence dimension” of Commonsense Realism is a commitment to the existence of most observables such as stones, trees, and cats and to these entities having, for the most part, the properties attributed to them by science (p.58) and common sense. Typically, idealists, the traditional opponent of realists, have not denied this dimension; or, at least, have not straightforwardly denied it. What they have typically denied in response to the skeptical challenge is “the independence dimension”. According to some idealists, the entities identified by the first dimension are made up of mental items, “ideas” or “sense data”, and so are not external to the mind. In recent times, under the influence of Kant, another sort of idealist has been much more common. According to these idealists, the entities are not, in a certain respect, “objective”: they depend for their existence and nature on the cognitive activities and capacities of our minds. Realists reject all such mind dependencies. Relations between minds and those entities are limited to familiar causal interactions long noted by the folk: we throw stones, plant trees, kick cats, and so on. We could say a lot more to make these doctrines precise and I have done so elsewhere.1 But these definitions will suffice for our purposes.

Let us now consider whether underdetermination has any consequence for Commonsense Realism.

2. Extreme Skepticism

In the First Meditation Descartes famously doubted the evidence of his senses. We can see this as an argument about underdetermination. Descartes believes that he is sitting by the fire. But perhaps he is suffering from an illusion, perhaps he is dreaming, perhaps he is being stimulated by an evil demon. There seem to be a range of alternative hypotheses to Descartes's belief, each equally compatible with the evidence available to him: his belief is underdetermined by the sensory evidence.

This is the first step in the argument of the extreme skeptic. We do not need to imagine evil demons and the like to find support for this underdetermination: it is supported by our psychological and neurophysiological theories of perception. Consider this example: I look in front of me and come to believe that there is a cat there. Our scientific explanation of this is roughly: the cat reflects light waves which provide stimulus to my retina causing sensory neurons to fire leading to my belief. But that scientific account also tells us that the cat is not necessary for the belief, for a multitude of reasons. The cat is not necessary for the light waves that strike the retina: the waves might have (p.59) other causes. The light waves are not necessary for the stimulus: there are other ways of providing that stimulus. Similarly the stimulus is not necessary for the neuron firings and those firings are not necessary for the belief. So the belief has a range of rivals that are compatible with the sensory evidence.

Let us call beliefs like Descartes's and mine—beliefs about the external environment caused by perception—“observational beliefs”. We seem to have established the following basic underdetermination thesis:

  • D1: Any observational belief has rivals that are equally compatible with the actual given sensory evidence for that belief.2

D1 needs some clarification. What exactly is “the sensory evidence”? We might take it to be the perceptual experiences themselves, but then it is not clear what it is for an observational belief to be “compatible with” the evidence. If we suppose that perceptual experiences produced beliefs about ideas or sense data then we could take beliefs about them to be the sensory evidence; for example, the belief that the sense datum I am now perceiving is cat‐like. But the supposition is controversial at best. We can be more ontologically cautious, taking the sensory evidence to be beliefs, prompted by perceptual experience, about how things appear (with no commitment to mental entities); for example, the belief that it appears to me that there is a cat in front of me. Then we can take D1 to be claiming that this belief is logically consistent not only with the belief that there is a cat in front of me but with many other beliefs. That is surely true.

We move to the second step in the extreme skeptic's argument. Given D1, how could an observational belief be justified? What basis is there for eliminating rivals that are equally compatible with the evidence? The skeptic's position is that there is no basis. She infers the troubling epistemological thesis:

  • A1: Any observational belief has rivals that are equally supported by the actual given sensory evidence for that belief.3

D1 is concerned with the deductive relation of logical compatibility between the sensory evidence and observational beliefs; hence the ‘D’ in its name. A1 is a much stronger underdetermination thesis because it is concerned with the ampliative relation of epistemic support between the sensory evidence and observational beliefs; hence the ‘A’ in its name.4 If A1 is (p.60) correct, even ampliative inferences like induction and abduction (inference to the best explanation) do not provide a basis for preferring any observational belief over many alternative hypotheses. Indeed, the skeptic is dubious of all such inferences. Our observational beliefs about stones, trees, cats and the like are all unjustified. Commonsense Realism must be abandoned.

Suppose that we found some way to reject A1 and justify our observational beliefs. We still have a way to go to escape extreme skepticism. We want to move from observational beliefs to singular beliefs about unobserved objects and to general beliefs that cover such objects; to take a boring but familiar example, we want to move from the evidence of many ravens all observed to be black to the belief that Oscar, an unobserved raven, is black, indeed to the “theory” that all ravens are black. The skeptical tradition once again presents us with an underdetermination problem:

  • D2a: Any theory has rivals equally compatible with the actual given observational evidence for that theory.

This seems to be indubitably so. Thus our evidence of many sightings of ravens is compatible not only with the theory that all ravens are black but also with the theory that all ravens are black except the ones on island X which has never been investigated. Another related underdetermination thesis also causes trouble:
  • D2b: Any theory has rivals that entail the same actual given observational evidence.

Thus the theories that all ravens are black and that they are all black except on X both entail our evidence of raven sightings.

These deductive underdetermination theses are the first step in the extreme skeptic's argument. She then, once again, infers an ampliative thesis. Given these deductive theses, how could a theory be justified? She arrives at another troubling epistemological thesis:

  • A2: Any theory has rivals that are equally supported by the actual given observational evidence for that theory.

According to this ampliative underdetermination thesis, a theory enjoys no epistemic support over some of its rivals.5 Our general beliefs about (p.61) stones, trees, cats and the like are unjustified. Again Realism must be abandoned.

We note that the pattern of the skeptic's argument is:

  • Deductive underdetermination → ampliative underdetermination → antirealism.

Now, it might well be objected that my discussion does not do skepticism justice for it takes skepticism to be committed to A1 and A2. The cautious skeptic would avoid this commitment, like almost all others, simply putting the onus on the nonskeptic to undermine A1 and A2. I think this is a good objection but, in the end (section 6), we shall see that it makes no difference whether the skeptic is committed or not. Meanwhile, for convenience, I shall take the skeptic to be committed.

3. “First Philosophy” Responses

We could accept the inferences to A1 and A2 and give up on our knowledge of the world: we do not know what there is and what it is like. This abandons the existence dimension of Realism and is very unappealing. The traditional responses of “First Philosophy” were different.

One response started by seeking a more basic area of knowledge than (what I have called) our observational beliefs, an area that was not open to skeptical doubt and that could serve as a foundation for all or most claims to knowledge. This foundation was found in the sensory evidence for our observational beliefs: we were thought to have indubitable knowledge of our own ideas (sense data); this knowledge was not underdetermined. Even if we go along with this highly dubious claim, we still have to get from this foundation to knowledge of the world of stones, trees, cats, and the like, thus rejecting A1 and A2. In attempting to solve this problem, foundationalists nearly always gave up the view that the world is external to the mind, thus abandoning the independence dimension of Realism. It was thought that only by constituting the world somehow out of ideas could we hope to save our knowledge of it. Realism leaves a “gap” between our ideas and the world that makes knowledge of the world impossible. Idealism closes the gap by bringing the world into the mind.

Another traditional response, currently much more popular, is also idealist. It seeks to reject A1 and A2 by taking the world to be partly constituted by the mind's imposition of concepts, theories, or languages. We can know about (p.62) that world because we partly create it. Realism's independence dimension is abandoned once again.6

So the price of saving our knowledge in the face of underdetermination and skepticism was typically an idealist metaphysics of one sort or another. Even if we were prepared to pay the price of such a bizarre metaphysics, these responses would not be too convincing. Although First Philosophy aims to take skepticism seriously and hence meet the very demanding skeptical standards for rational belief, it often seems to fall short of those standards: it assumes what no self‐respecting skeptic should allow (for example, indubitable knowledge of ideas). So even with an idealist metaphysics we still seem not to have the knowledge we want. I have argued for this, and against idealism, elsewhere (1984, 1991b, 1997a, 1999a, and 2001a, which is Ch. 5 in the present volume).

We are faced with a choice between skepticism and idealism. Surely something has gone seriously wrong. It is time to think again. I shall first make a Moorean response to the skeptical challenge, then a naturalistic one.

4. A Moorean Response7

The disaster has come from the epistemological theses A1 and A2 which make Commonsense Realism untenable: we are supposed to doubt our commonsense beliefs in an external world. But why should we accept these skeptical theses? How much confidence should we have in a view that undermines Realism? Realism is a compelling doctrine almost universally held outside intellectual circles. From an early age we come to believe that such objects as stones, cats, and trees exist. Furthermore, we believe that these objects exist even when we are not perceiving them, and that they do not depend for their existence on our opinions nor on anything mental. This Realism about ordinary objects is confirmed day by day in our experience. It is central to our whole way of viewing the world, the very core of common sense. A Moorean point is appropriate: Realism is much more firmly based than the epistemological theses A1 and A2 that are thought to undermine it. We have started the argument in the wrong place: rather than using A1 and A2 as evidence against Realism, we should use Realism as evidence against A1 and A2. We should, as I like to say, “put metaphysics first”.

(p.63) Descartes puts us in an armchair and asks us to start by clearing our minds of all knowledge and doing some epistemology. The Moorean puts us in an armchair and asks us to start by assessing the evidence for Realism. In so doing we must resolutely decline to theorize about standards of good and bad evidence, for that epistemological path was what led to the disaster: we simply apply our ordinary evidential standards, just as we presumably did in childhood when we became Realists in the first place.8 Once we have done that, we turn to the epistemological theses A1 and A2. In assessing them we have little to go on but the skeptical argument that puts these theses at odds with Realism. We ask: is it more likely that the theses are mistaken than that Realism is? Is it more likely that there is a flaw in the skeptical argument than that, contrary to what we have always supposed, we lack knowledge of the external world? The Moorean answers these questions with a resounding “Yes”.

5. A Naturalistic Response

What else do we have to go on in assessing A1 and A2? D1 does not entail A1, nor do D2a and D2b entail A2. Where can we look for more evidence? Not, First Philosophy assumes, to empirical science, for science itself is doubted by the extreme Cartesian skeptic. So the evidence must be of some nonempirical sort. Thus the various idealist positions that rejected the theses were thought to be, like mathematics and logic, known a priori. The a priori approach is the very essence of First Philosophy and its response to underdetermination. Reflecting from the comfort of armchairs First Philosophers decide what knowledge must be like, and from this infer what the world must be like. If the world were the way the Realist says it is, we could not know about it. Yet, it is typically thought, we surely do know about it. So the Realist cannot be right.

The Moorean response alone casts doubt on any such arguments. Given the strength of Realism, it is simply not plausible that we could know something a priori that undermined it, whether that something is the skeptic's A1 and A2 or the idealist's response to them. But the Moorean response is not, of course, sufficient. We need to do better and we can. Manifestly we could not have a priori knowledge damaging to Realism if we could not have a priori knowledge at all. According to Quinean naturalism, we could not: there is no a priori knowledge. There is only one way of knowing, the empirical way that (p.64) is the basis of science (whatever that way may be); in Quine's vivid metaphor, the web of belief is seamless.9 So we could not know A1 and A2 a priori because we could not know anything a priori (1996a, 1998, which is Ch. 12 in the present volume, 2005a, b, and 2009c, which is Ch. 13 in the present volume).

So, we should now consider A1 and A2 from a naturalistic perspective.

6. Naturalism and the Underdetermination Arguments

Naturalism is an overarching epistemological doctrine claiming that the only way of knowing anything is the empirical way of science: for each area of knowledge x, naturalized x. When the area is physics, this yields naturalized physics, when the area is biology, it yields naturalized biology, and when the area is epistemology, it yields naturalized epistemology. Everyone believes in a naturalized physics. Everyone but a few benighted creationists in places like Kansas believes in a naturalized biology. But those in the tradition of First Philosophy do not believe in a naturalized epistemology. The radical consequence of naturalism is that philosophy, including epistemology, becomes continuous with science.

From this naturalistic perspective, the troubling epistemological theses, A1 and A2, have no special status. They have to be assessed empirically, contrary to the assumptions of First Philosophy, because there is no other way to assess them. As empirical theses, they do not compare in evidential support with our view of stones, trees, cats, and the like. Experience has taught us a great deal about such objects but rather little about how we know about them. So epistemology is just the wrong place to start the argument: it is one of the weakest threads in the web of belief (cf. Quine's vivid image of Neurath's boat).10 Instead, we should start with an empirical metaphysics and use that as the basis for our naturalized epistemology, as the basis for our empirical study of what we can know and how we can know it. Instead of the traditional pattern of argument, exemplified by the underdetermination arguments against Realism,

  • a priori epistemology → a priori metaphysics,

(p.65) we should follow the pattern,
  • empirical metaphysics → empirical epistemology.

The underdetermination arguments not only use the wrong methodology, they proceed in the wrong direction.

Proceeding empirically in the right direction, we start with metaphysics. Realism is then irresistible. Indeed, it faces no rival we should take seriously.11 We then turn to naturalized epistemology. This is a very difficult matter. Still, with Realism established, we already know that A1 and A2 are false. We can go a bit further. It is clear from scientific practice that we are entitled, despite D1, to dismiss implausible theses like the evil demon one; and that we are usually entitled, despite the equivalence that a theory has with some rivals according to D2a and D2b, to prefer that theory to its rivals; we are entitled to believe that all ravens are black, for example. The epistemic standards implicit in scientific practice clearly give us these entitlements. We would like to know, of course, exactly what those standards are but it has proved notoriously difficult to say. Nevertheless, it is indubitable that, whatever the standards are, they give us these entitlements.

Finally, we must consider the objection raised in section 2. For convenience, we have taken the skeptic to be committed to A1 and A2. The objection is that this does not do the skeptic justice. The cautious skeptic would be dubious of these epistemological theses as of all other substantive theses. After all, she is a skeptic. Since she is not committed to the theses she is not committed to knowing them a priori. So what then is the nature of her challenge to Realism? She puts the onus on the Realist to justify his rejection of A1 and A2. She does not boldly assert the badness of ampliative inferences, thus embracing skeptical standards of justification. She simply points to these epistemic standards, which yield A1 and A2, and asks for a justification for ruling the standards out in favor of Realist alternatives that reject A1 and A2.

From our naturalistic perspective the challenge of this cautious skeptic is no more difficult to meet than that of the incautious one. For what I have presented is just what the cautious skeptic wants: a case against A1 and A2 and against any epistemic standards that would sustain them. On the one hand, the strength of the case for Realism counts against them. On the other hand, the practices of science count against them. And these practices are the only place to look in assessing epistemic standards because there is no a priori knowledge. These practices support the use of ampliative inferences for preferring a theory (p.66) to many, if not all, of its rivals. Sometimes, of course, a theory will face a rival that cannot be ruled out in this way but it is not the case that all theories always face such rivals. Commonsense Realism is not threatened by the underdetermination that remains.

7. Conclusion

We have considered the underdetermination theses A1 and A2 of extreme skepticism. If true these would count against nearly all our knowledge and hence undermine Commonsense and Scientific Realism. The traditional responses of First Philosophy to these theses rest on a priori knowledge. I have argued elsewhere that there is no such knowledge. Even if there were, these traditional responses tend to involve bizarre metaphysics and to be otherwise unsatisfactory. Instead, I urge a naturalistic response that gives priority to metaphysics over epistemology. The case for Realism is then irresistible and alone shows that A1 and A2 must be false. Furthermore, the epistemic standards implicit in science count against these theses and hence against extreme skepticism.


(*) This chapter draws heavily on part I of “Underdetermination and Realism” (Devitt 2002a) and appears with the kind permission from Wiley‐Blackwell. Part I of that publication includes a discussion of a priori knowledge which has been omitted from this chapter; but see Chs. 12 and 13. Part II discusses underdetermination and scientific realism, on which see Ch. 4, sec. 3.1 and Postscript.

(1) 1991b, 1997a, 1999a. My definitions are unfashionable in not being, or even appearing to be, semantic. I argue that it is very important to disentangle the metaphysical doctrine of realism from any semantic doctrine. See also Ch. 2, part I, in the present volume.

(2) Note that the thesis concerns the actual given evidence. There are much stronger underdetermination theses that concern possible evidence. See Ch. 4, sec. 4.1, in the present volume.

(3) Quine thinks this likely even given all the sensory evidence that there will ever be, “man's surface irritations even unto eternity” (1960: 23).

(4) Larry Laudan emphasizes the importance of the distinction between the deductive and the ampliative in his helpful discussion of underdetermination theses (1996: ch. 2).

(5) Laudan calls this “the nonuniqueness thesis”. A stronger thesis, “the egalitarian thesis” claims that a theory enjoys no support over all of its rivals. He is sadly persuasive in attributing the latter thesis to Quine, Thomas Kuhn, and Mary Hesse (1996: 33–43).

(6) E.g. Kuhn 1962 (on which see Hoyningen‐Huene 1993); Feyerabend 1975, 1981a; Goodman 1978; Putnam 1981; Latour and Woolgar 1986. See Ch. 5 of the present volume for discussion.

(7) The discussion in the next three sections takes a similar line to that against constructivism in Ch. 5, sec. 6, and against semantic ascent in Ch. 15, sec. 4, of the present volume.

(8) Devitt 1984/1991b: sec. 5.7, in effect, does this (albeit without emphasizing that the independence of the world is also confirmed day by day in experience).

(9) See particularly Quine 1952: pp. xi–xvii; 1961: 42–6. Quine uses ‘naturalism’ to stand for this epistemological doctrine. Others use it to stand for a reductive metaphysical doctrine like physicalism.

(10) For details of the image, see Ch. 5, sec. 6, of the present volume.

(11) Devitt 1984/1991b: sec. 5.9, demonstrates the difficulties of arguing for antirealism from a naturalistic perspective.