Lakatos and MacIntyre on Incommensurability and the Rationality of Theory-change
Although he is not primarily a philosopher of science, Alasdair MacIntyre has drawn on post-Kuhnian methodological reflection in his formulation of an historicist theory of knowledge (1984a: 271) or what his more recent work terms tradition-constituted inquiry (1988: 354). In many respects, MacIntyres traditions are similar to the research programs described in the work of Imre Lakatos (1977). Both thinkers propose a shift in focus from atomic propositions to some type of holism by making an entire theory, or series of theories, the proper object of evaluation. Each argues that the issues investigated by participants in research traditions are not timeless questions, but are crucially shaped by their own problematics. Without devaluing consistency and logical rigor, each supposes that incoherence of a certain sort is the motor of intellectual progress. And finally, both philosophers adhere to a realist conception of truth. In short, MacIntyre and Lakatos want to abandon positivist methodological assumptions and acknowledge the historical dimension of scientific enquiry, without succumbing to any species of anti-realism.
These similarities create the appearance of essential continuity between Lakatos and MacIntyre. One might be tempted to view MacIntyres theory of traditions as an application of Lakatos philosophy of science to his own interests in ethics. Closer examination, however, shows that MacIntyres theory of traditions and Lakatos methodology of scientific research programs are incompatible, because they take divergent positions on the relation between incommensurability and rationality in theory-change. Lakatos holds that incommensurability is impossible, because it prohibits an observer from affirming that movement from one research program to another is ever rational. MacIntyre, by contrast, defends both incommensurability and rationality in theory-change, and employs several strategies to dissolve any assumed tension between the two. After probing this disagreement, I will conclude that MacIntyre has the better of the argument, while leaving us with some questions about his own research program.
I. Lakatos' Rejection of the Incommensurability Thesis
Lakatos rejects the incommensurability of rival scientific theories, a thesis advanced in the work of Thomas Kuhn and Paul Feyerabend. The thesis, as characterized by Dudley Shapere, is that far from it being the case that an earlier such theory is encompassed in and derivable from a later one, the later one introduces entirely new meanings, standards, methods, and goals, making any deductive relationship between the two impossible (Shapere 1995: 139). (1) We shall label this version of the incommensurability thesis IT. Lakatos advances two objections against IT. The first is the empirical fact that a single scientist can simultaneously work on two rival and ostensibly incommensurable research programs. This fact undermines Kuhns thesis of the psychological incommensurability of rival programs (Lakatos 1977: 112). We need not dwell on this objection, for it misses the point altogether. Despite his use of psychological metaphors, Kuhn is not denying the ability of a scientist to entertain two rival paradigms. IT is a thesis about linguistic structure, not the workings of the human brain.
Lakatos second objection to IT is considerably more interesting. It assumes the form of an indirect attack on IT. If rival and ostensibly incommensurable theories are neither inconsistent with each other, nor comparable for content, then it becomes impossible to say that allegiance to one research program is more rational than allegiance to another. If such comparative judgments are impossible, then it appears that while scientists may decide to switch from one large-scale theory to another, there are never any reasons that mandate or justify the switch. The conclusion is that incommensurability and rationality in theory-change are simply incompatible.
By itself, this argument does not refute IT. It simply holds that if one affirms IT, then one must deny the appearance of rationality in theory-change. If one wants to save the appearance, however, then one must deny IT. This conclusion is considerably weaker than the claim that IT is false, but it is sufficient to separate Lakatos from MacIntyre. Against Lakatos, MacIntyre denies that IT and rationality in theory-change are incompatible, and argues that one can simultaneously affirm the existence of genuine incommensurability between traditions and the rational character of switch in allegianceat least on occasionfrom one incommensurable tradition to another.
After making the claim that IT and rationality in theory-change are antithetical, Lakatos proceeds to argue that even if the historical record seems to show that large-scale bodies of theory are incommensurable, and thereby confirm IT, it remains the case that some research programs can be rationally chosen against others. This argument is based on a distinction between the internal history of science and its actual history. The elements of the internal history are the rational reconstructions of theories that the seeker of scientific truth, unlike the purely descriptive historian, can and must devise. The actual history may be safely relegated to the footnotes (Lakatos 1977: 120). Theories as described by actual history may well be incommensurable. But the rational reconstructor can place incommensurable paradigms in a relationship of logical inconsistency, and make their content comparable, by using a dictionary (of her own creation) that enables translation between rival theories (1977: 91n). Lakatos strategy is not a straightforward denial of IT (although he does leave this Davidsonian door open), but a restriction of its scope to actual, descriptive history. This move emasculates IT by rendering it irrelevant in the normative context of evaluation and adjudication.
II. MacIntyre's Reply to Lakatos
In forcing himself to distort the historical record, Lakatos can never provide rational justification for changes from one theory to another. At best, he can only justify the switch from one caricature to another. If this is the case, then it appears that there are no rational grounds for eliminating scientific theories. We can eliminate only their caricatured simulacra, which are all too likely to be drawn from a perspective that already knows the correct outcome. Something like this is the objection of MacIntyre, who in response to Lakatos dismissal of actual history argues that it matters enormously that histories should be true, just as it matters that our scientific theories make truth one of their goals (1977: 469).
Lakatos might reply that the decision to privilege internal history over actual history is simply the price that one must pay, if one wants to preserve rationality in theory-change. MacIntyre denies this, in effect, by proposing a view that argues for the mutual compatibility of actual history, IT, and rationality in theory-change. If MacIntyre can deny the necessity of dispensing with the historical record, then Lakatos argument will lose much of its force, and his intention to take seriously the historical dimension of scientific change will be better honored by MacIntyre.
The first part of MacIntyres argument is to deny that the incommensurability of rival paradigms entails their incomparability. The version of IT assumed by Lakatos does seem to imply the total incomparability of paradigms. In Shaperes formulation, IT holds that incommensurable theories have entirely different meanings, criteria and goals. Here we have not one but several incommensurability theses. One may, prima facie, speak of incommensurability of meaning or criteria without postulating incommensurability of goals. The first step is to disambiguate IT by distinguishing several possible versions.
Incommensurability of meaning entails incommensurability of criteria and incommensurability of goals. There are no shared goals or criteria, since there is no common language in which shared goals or criteria can be formulated. This is the most radical version of the incommensurability thesis, IT1.
There is no need to postulate incommensurability of meaning. Rival theories are perfectly capable of communicating with one another. What they cannot do is agree with another on the basis of shared criteria, either because the criteria are nonexistent, or because they are so weak as to be consistent with both theories. This version of incommensurability is less a semantic thesis than a thesis about truth. Let this be IT2.
Incommensurability of meaning is present in many of the key terms that comprise rival theories, at IT1 holds. And there are no shared criteria that enable rational adjudication, as both IT1 and IT2 affirm. However, rival theories do share the same goals. (IT1 denies this, IT2 is neutral.) Common goals ensure the comparability of incommensurable paradigms, and make a switch from one to the other potentially rational. Call this IT3.
No less than Lakatos, MacIntyre denies IT1, and seems to agree that it entails the irrationality of scientific theory-change. IT2 is a weaker supposition, a possible option for the advocate of an incommensurability thesis who wants to save the appearance of rationality in theory-change. MacIntyre, however, affirms incommensurability both as a semantic thesis, and as a thesis about the absence of criteria that are at once common and decisive. His position is closest to IT3, and must be examined more carefully.
IT3 concedes that the referent of terms such as mass will differ from one theory to the next. MacIntyre agrees with Kuhn that mass as defined by Newton and mass as defined within quantum mechanics are concepts embodied in incommensurable bodies of theory (MacIntyre 1984b: 42). He adds, however, that both concepts of mass must be understood as concepts of that property of bodies which determines their relative motion, if we are to be able to understand what makes those bodies of theory contending rivals (1984b: 43, italics in original). This distances MacIntyre from IT1, because it posits a higher-level language in which a common subject-matter can be formulated. If there no such language existed, then the theories would be incommensurable, but not incommensurable rivals. Rivalry requires the ability to specify the subject matter in a common language. Incommensurability requires that any common criteria are too weak to provide ground for rational choice between them.
If rival theories necessarily have the same subject matter, they will also have the same goals. The assumption is that goals are defined in terms of the shared subject matter. This seems questionable, because two theories can take as their subject the relative motion of bodies, and yet have different goals. One theory might want to predict novel facts, while another might privilege simplicity. If shifting Kuhnian values (1971: 185) are understood as the goals of rival theories, then it appears that the possession of shared subject matter does not entail identity of goals. The problem may disappear, however, if we follow Ernan McMullin and make a distinction between primary goals and secondary goals (1993: 67-68). Secondary goals are epistemic values that serve as means to endse.g. logical consistency, compatibility with other knowledge, fertility, unifying power, coherence. They are valued because they enable research programs to achieve goals that do not change over time, the primary goals of empirical predictive accuracy and explanatory power (goals whose fulfillment is truth-indicative, for scientific realism). If this distinction is valid, then MacIntyre can say that competing scientific theories have constant goals, even as they lack common criteria. For instance: rival paradigms in physics, despite differences in criteria, each have as their goal the formulation of general and accurate equations for motion.
In keeping with IT3, then, MacIntyre contends rival traditions share basic goals, even if they do not possess common criteria that are strong enough to enable decision in favor of one of them. Two theories may be incommensurable, as the term is understood by proponents of IT3, yet genuine rivals. Still lacking is an account of how rational theory-change is possible between two incommensurable theories, if they lack common decision criteria. Lakatos assumes that without common criteria, no rational adjudication between rival theories is possible. To vindicate his position against Lakatos, MacIntyre must show that incommensurability, even in the refined sense of IT3, is compatible with rationality in theory-change.
MacIntyre argues that, at least in some cases, traditions fail not by external criteria, and not by criteria ostensibly held in common with other traditionsbut by their own standards. If this is true, then one can imagine a comparison between two rival incommensurable theories. One theory fails by its own standards, and the other does not. It is rational, then, to abandon the failed theory. This by itself does not preserve rationality in theory-change in the full sense. It saves only the rationality of theory abandonment. MacIntyre is aware of this, and adds that it is rational to adopt the theory that does not fail only if can explain the reasons for the defeat of its rival (1984b: 43).
By refining the notion of incommensurability, and appropriating the idea of internal failure, MacIntyre has provided us with the outlines of a conception on which one can adjudicate between two rival incommensurable theories, with no common decision criteria between them. His account does not resort to Lakatos expedient of imposing common criteria onto rival theories, and thereby falsifying them. By making the failure of paradigms a matter of their internal logic, it avoids the hollowness of critiques that rely on standards external to the theories that are the object of criticism.
III. Some Questions
MacIntyres solution to the problem left by Lakatos is not without difficulties of its own. I will first respond to a common objection to MacIntyres conception of tradition-constituted enquiry, and then proceed to raise some questions.
The objection is that MacIntyres own evaluations require a criterion that is not internal to any particular tradition, but that is common to all of them. In other words, MacIntyre covertly uses a meta-criterion, viz. the demand that a tradition succeed, or not fail, by its own standards. This criterion, it seems, does not derive from a single tradition, but is the neutral standing ground whose existence MacIntyre affirms in practice, even as he denies it in theory (see MacIntyre 1988: 350).
This objection rests on a confusion. The claim that the meta-criterion is not unique to a single tradition does not imply that its source lies outside of any tradition whatever. It may be that the demand to aim at the accomplishment of particular goals (whose proper description cannot detain us here) is a constitutive part of all scientific traditions, and that the criterion has no existence outside of them. MacIntyre can easily deny that he is invoking a meta-criterion which hovers outside of the relevant traditions.
The common objection to MacIntyre fails, but some questions remain. Epistemological problems attend the notion of a traditions internal failure. How does one know that what appears to be a failed theory has genuinely failed by its own standards? At many points during the history of a tradition, a gap will appear between what the theory does and what its criteria say it ought to do. When is the gap so wide that it constitutes a failure? Certainly in narratives written by the opposition the size of the gap will render the theory dead, never to rise again. But proponents of the theory can often tell another story, in which the alleged failures are transitory setbacks, or even specimens of the anomalies that, on MacIntyres own account, are the condition of all intellectual progress (1984b: 42). MacIntyre can respond that there are criteria for distinguishing fruitful tensions from incoherences that cause internal failure. This response, however, merely raises the same question once again. Do the criteria intended to distinguish between temporary fatigue and total collapse constitute impersonal standards, or are they also expressions of parochial bias? Knowing precisely when a tradition has failed by its own standards, without begging the question against its adherentswhose own perspective may convince them that the tradition has stalled without having failedmay be more difficult than MacIntyre allows. Alongside narratives of failure will be narratives whose interpretations are more charitable. Only rarely is the truth value of these interpretations not a matter of controversy.
But even it were possible to establish that a particular tradition has unmistakably failed, and failed by its own standards, further questions remain. Why should success or failure in terms of internal criteria be normative for rationality? Why is this type of failure a decisive reason for rejecting a tradition? We can imagine a scenario in which two rival, incommensurable theories differ. One theory has criteria that are easy to meet; those of the second are more difficult to satisfy. Yet it is possible that the second theory is better than the first, even if the first succeeds in terms of its own criteria and the second fails. Why abandon the second tradition, if the only reason for doing so is the cheap success of the first? MacIntyre may respond that traditions whose standards are too high do not collapse butif they are able to flourishreformulate their criteria, without losing their essential identity. This line of response may be promising, but it raises some questions. What are the identity conditions of a tradition? Who determines these, and how? What distinguishes the substance of a tradition from its accidents? How does one map the logical relations among various traditions? When does a body of doctrine become a tradition? The persistence of these issues as disputed questions requires MacIntyre to say more about the ontology of traditions.
To concludeit must be said that while Lakatos philosophy of science is vulnerable, since it depends on a strong dismissal of actual history, his conception of research programs broke much ground for later accounts that attempt to preserve a thoroughgoing holism and realism without denying the existence of incommensurability. Among such accounts, MacIntyres tradition-constituted inquiry is perhaps the most promising. Certainly it leaves unanswered questions, some of which I have tried to indicate. This is not necessarily a strike against it. To generate unresolved issues, unanswered questions, unsettled objections is a feature of all fertile research programsincluding MacIntyres own.
(1) I adopt Shaperes formulation as a preliminary characterization of IT because it captures in brief compass the version of the thesis that many (including Lakatos) have attributed to Kuhn and Feyerabend. I bracket the issue of whether Shapere accurately characterizes the views of either (For the suggestion that to equate incommensurable and incomparable is to misrepresent Kuhn, see McMullin 1993:56, and the recent essays by Kuhn cited in McMullin 1993: 76.).
Lakatos, I. 1977. The Methodology of Scientific Research Programmes: Philosophical Papers, volume 1, ed. J. Worrall and G. Currie. Cambridge: Cambridge University Press.
MacIntyre, A. 1977. Epistemological crises, dramatic narrative, and the philosophy of science. Monist 60: 453-72.
________. 1984a. After Virtue: A Study in Moral Theory, 2nd edition. Notre Dame: University of Notre Dame Press.
________. 1984b. The relationship of philosophy to its past. In Philosophy in History., ed. R. Rorty, J.B. Schneewind and Q. Skinner, 31-48. Cambridge: Cambridge University Press.
________. 1988. Whose Justice? Which Rationality? Notre Dame: University of Notre Dame Press.
McMullin, E. 1993. Rationality and paradigm change in science. In World Changes: Thomas Kuhn and the Nature of Science, ed. P. Horwich, 55-78. Cambridge: The MIT Press.
Shapere, D. 1995. Review of H. Sankey, The Incommensurability Thesis (Aldeshot: Avebury, 1994). In British Journal of the Philosophy of Science 46: 139-141.