![]()
Part I
Procedural Foundationalism
The final arbitrator in philosophy is not how we think but what we do.
(Ian Hacking)
![]()
Ā§ 1
Introduction to Part I
Once upon a time epistemology was soothing. To stay out of trouble it counseled reliance on undoubtable sense experience, and inference from that basis by a priori truth. Testing beliefs, therefore, was easy. You deduced observational consequences from them, and then observed if they were true.
Quine inherited from Carnap a diminished version of this view, and repudiated it. No foundation for our knowledge claims is possible, he argued, since no purely observational language is possible. The commonsense conceptual scheme in which we live and breathe is impure, each precious concept, however observational, tainted by theory. We may replace the scheme with something better, and, indeed, replace our native tongue with a formal prosthesis; but any synthetic product, whatever its virtues, will not cleanly divide the unassailable observation from the dubious inference. As for insight into incorrigible mathematical and logical truth, holism dictates that in surrendering to the demands of recalcitrant experience, any purported truth, no matter how treasured, may end its day on the chopping block.
The result is that testing your beliefs is not as easy as the positivist view made it seem: When you look around to see if your āobservationalā consequences are true, you are, in principle, testing your entire set of beliefs. The flip-side of the claim, one paragraph above, that any purported truth can go, is that any purported truth can stay (regardless of how the world intrudes), provided we tinker with other bits of what we believe. Our methods of confirmation and infirmation, thus, are tainted by vague methodological canons such as simplicity and conservation.
Quine has company now: due in part to his influence, these views have become clichƩs of contemporary epistemology. Parts I and II, therefore, challenge this picture and substitute another. In Part I, I revive a form of epistemic foundationalism, although not the linguistic form crushed so thoroughly by Quinean considerations. Rather, I shift from attempts to separate linguistic components of our conceptual scheme which are (more or less) observational from those which are (more or less) theoretical, and attend instead to the methods used to connect our terms to the things we refer to.
Hereās the script for Part I. Ā§Ā§ 2 and 3 attack confirmation holism and theoretical deductivism, at least in the forms they take at the hands of recent philosophers of science. I show that the impact of confirmation holism on the epistemology of science is far less than philosophers make of it. In particular, confirmation holism is muted across the traditional fault lines between sciences; indeed, these fault lines are not merely ones of subject matter: they are epistemic fault lines.1 I next mount an attack on theoretical deductivism, and then in Ā§ 4 discuss what the autonomy of the special sciences comes to. Ā§ 5 examines the role of the truth predicate in our web of belief and uses the results against Nancy Cartwrightās brand of scientific anti-realism. In Ā§ 6, I turn attention to a class of regularities (what I call āgross regularitiesā) which are the soil from which the evidential procedures for the sciences spring. It is the peculiar fact that some gross regularities must simultaneously function as an evidential basis for the sciences while remaining (for the most part) impermeable to scientific theory which gives procedural foundationalism its unusual character.
At this point Iām positioned to discuss the methods, mentioned one paragraph back, used to connect our terms to the world. These methods, called procedures, are ways of forging and exploiting causal connections between ourselves and what we talk about: their operations are licensed largely by gross regularities, and as a result they enjoy partial epistemic independence from empirical science, too. This is the central tenet of procedural foundationalism.2
A further claim (Ā§ 7) is that certain special (perceptual) procedures are foundational to all our causal practices. I mean by this, first, that such procedures are in a strong sense (how strong, I make clear in Ā§ 8) not replaceable, and, second, that they are (literally) proximate parts of all the procedures we use to connect our terms to the world.
Procedures should be studied carefully by those interested in the ācausal theory of reference.ā But my concern in Part I is not so much with what empirical terms refer to as with the general epistemic properties of the methods we use to get at what such terms refer to. The term will not be neglected, however: I turn to it in Part II.
These days, meaning holism,3 as itās called, gets a lot of attention because of its apparent impact on central doctrines in linguistics and the philosophy of mind. Almost everyone who distinguishes between confirmation holism and meaning holism unconcernedly concedes the former to Quine, while going on to worry about what argument can enable a deduction from that (uncontroversial) holism to the interesting (if possibly incoherent) claim of meaning holism. My concern, by lonely contrast, is not with meaning holism at all: I investigate the scope of confirmation holism, for that holism is the one bearing on the epistemological issues involved here.
Despite my adoption of procedural foundationalism, I call the general epistemic position developed in Part II ātwo-tiered coherentism.ā That is, the epistemology for sentences (the methods by which sentences are confirmed and disconfirmed) is coherentist in contrast to that for procedures.
Hereās why. Procedural foundationalism recognizes that the causal mechanisms used to connect ourselves to the world have a certain (epistemic) rigidity. But this does not require terms to have fixed sets of causal mechanisms (āoperationsā) associated with them, as operationalists hoped.4 Vocabulary can change a lot even if the epistemic engineering embodied in our causal resources remains untouched.
A corollary is that procedural foundationalism has only indirect impact on the plasticity of conceptual change. Although the broad sorts of claims one finds in, say, Churchland (1979), Feyerabend (1962), Kuhn (1970), Quine (1951), etc., cannot be sustained, the argument is complex, as Iāll show in Part II, and doesnāt require rigid connections between sets of perceptual procedures and āobservation terms.ā
I conclude with an important methodological point illustrated in Parts I and II: our view of how we gather evidence and do science is seriously distorted if we concentrate only on how and whether vocabulary changes, and not on how and whether our evidential procedures do.
1 Contrast this with Fodor (1990a:183); I also recommend the reader take a peek at Hackingās discussion (1990:157) of Boutroux. Although I am not entirely unsympathetic to those who see metaphysical reasons for the existence of special sciences, I intend to concentrate much more, although not entirely, on the (generally underrated) epistemic reasons for such sciences.
2 I discuss procedures in pretty fair detail in what is to follow: for now, for concreteness, the reader may imagine seeing amoebas through a microscope. The procedures, in this case, are the methods the viewer uses via the microscope to learn about the properties of amoebas (I will not be very precise about how to individuate procedures); the gross regularities are, among others, various hands-on tricks a viewer uses to distinguish, for example, artefacts of the microscopic procedures from actual properties of amoebas.
3 The doctrine, roughly, that sentences or statements are not meaningful on their own, but only in a larger context such as a group of sentences, a theory, or an entire language.
4 Part IV shows how the presence of this slippage, between terms and the procedures used to pick out what such terms refer to, creates insurmountable difficulties for causal theorists, such as Fodor, who hope for nomological connections between our use of empirical terms and what they refer to.
![]()
Ā§ 2
Program and Scope
By a program for a science I mean a (sometimes vague) description of what its terms hold of, and on which its laws are supposed to operate. In practice, what falls under a program turns, broadly speaking, on a (fairly unified) set of methods available to practitioners, and on certain unities in the domain the science is applied to (which explain why the methods work where they do). Ultimately, however, the possible range of these methods (however refined) is taken to be underwritten by physical fact. Thus, the program of physics is widely believed to be āfull coverage,ā as Quine puts itāeverything is supposed to fall under its sway;1 chemistry studies the composition, structure, properties, and reactions of matter insofar as this turns on what we now know to be electromagnetic forces; organic chemistry studies the compounds of carbon; biology studies living organisms and life processes; etc.2
Scope is the domain of application of the laws and techniques of the science achieved to date. It depends less on having the right laws for a science than it does on technical and mathematical tools available which mark out how far the laws in question can be tested and applied.3
A brief digression. My subsequent concern is with the link between intractability (that is, the factors blocking the direct application of laws and truths of a science to areas we would otherwise expect them to apply) and the methodological independence of special sciences. I concentrate, therefore, on how scope limitations give rise to special sciences. But of course there would not be any special sciences if there were no domains obeying general regularities susceptible to recognition in a direct way without needing a reduction to something more fundamental. Here purely physical considerations, not epistemological ones, provide an explanation of what is going on. Programs in special sciences (and in physics, too) are ultimately to be explained physically.
Consider chemistry first. The reason chemical processes obey laws that can be studied somewhat independently of physics is that nuclear forces decay many magnitudes faster than electromagnetic ones. Chemical processes operate by means of atomic bonding via electron exchanges of various sorts; and the properties of elements, and the chemicals made from them, can be explained by generalizations relying on these exchanges alone.
Something similar may be said about biology. Regularities arise here because there are self-replicating (modulo factors which introduce variations) organisms which respond selectively to external pressures. Therefore we can explain why such regularities (and laws) arise without necessarily being able to reduce them to laws or regularities of the underlying sciences; we can, for example, explain why teleological explanations arise in biological contexts.4 The relation between psychology and neurophysiology is similar. End of digression.
The Program-Scope Gap: Illustrations from Physics
I now illustrate the program-scope gap in empirical science indirectly by showing how successful confirmation of scientific theory, and successful application of such theory to special contexts, indicate this gap. I discuss five approaches: the development and application of sophisticated mathematics; the use of calculational short-cuts; the deployment of idealized models; the design of laboratory experiments; and the use of formal heuristics.
The Development and Application of Sophisticated Mathematics
An enormous amount of eighteenth- and nineteenth-century mathematics arose from the need for methods of solving differential equations, and, where solutions are not to be had, methods of deriving theorems which shed light on the motions described by such equations.5
Differential equations arise in all branches of physics: celestial mechanics; the equilibrium of rigid bodies; elasticity; vibrating strings and membranes; pendula; heat; fluid dynamics; electricity; electromagnetic theory; relativistic and nonrelativistic quantum mechanics; etc.6 Such equations can (but donāt always) characterize what they describe deterministically. For example, consider the Newtonian n-body problem. Initial conditions plus the inversesquare law completely determine the locations, velocities, and accelerations of n point-masses for all time. This describes the program of (this bit of) Newtonian mechanics. Its scope, however, is limited by, among other things, our capacity to mathematically manipulate the differential equations in question to extract the physical information we need.7
The absence of general methods for solving differential equations leads naturally to an explosion in mathematics: Numerous techniques must be developed for solving specific classes of such equations. Such techniques are often quite restricted in their application, and offer little insight into why they apply where they do, and what (if anything) can used instead where they donāt. Since broad and comprehensive principles for the solution of differential equations are lacking, the result is a morass of separate techniques for the various types of equation.8
On the other hand, one also searches for general theorems that can be applied to differential equations, or to sets of such equations, when they canāt be solved, and which can answer important physical questions despite this. We can wonder, for example, under what situations solutions exist (for what sets of parameters) when they are unique, and how their properties depend on small changes in the values of these parameters. We may also be curious wh...
