|Theory of Knowledge
Explanation, Understanding, and Subjectivity
Dwayne H. Mulder
ABSTRACT: Many theorists of explanation from Hempel onward have worked with the explicit or implicit assumption that considerations of the subjective sense of understanding should be kept out of the formulation of a proper theory of explanation. They claim that genuine understanding of an event comes only from being in an appropriate cognitive relation to the true explanation of that event. I argue that considerations of the subjective sense of understanding cannot be completely removed from the process of formulating and justifying an acceptable theory of explanation. Although understanding is neither a necessary nor sufficient condition for an explanation, understanding is necessary as an initial guide to the nature of explanation. The widespread method of providing counterexamples for criticizing theories of explanation presupposes that there is a neutral method of identifying at least some clear cases of explanation and some clear cases of non-explanations. I argue that the only plausible method to fill this role relies essentially on the subjective sense of understanding. Objective validation of judgments about explanatoriness comes only through a complex process of social correction of our initial intuitive judgments regarding explanation.
It is clear that understanding and explanation are related. It is unclear exactly how they are related. We speak both of explaining-why and understanding-why some event occurred. Explanations typically produce understanding in those who consider them, and the sense of increased understanding typically comes from consideration of an explanation. Consideration of an explanation can, however, fail to produce in someone an increased level of understanding of the explanandum. Further, the subjective sense of gaining increased understanding of an event can come from an account that is not the explanation of that event. Hence, we cannot say that anyone's sense of understanding is either necessary or sufficient for an account to be an explanation. However, I shall argue, we cannot completely avoid all reference to understanding in a correct theory of explanation. This situation presents a pressing problem for philosophical studies of the nature of explanation, for many theorists relegate the sense of understanding to a strictly derivative position by claiming that the subjective sense of understanding of an event comes, under appropriate (articulable) conditions, from consideration of a potential explanation, and that genuine understanding comes, under appropriate conditions, from consideration of the true explanation. (See, for example, Hempel 1948, 256-257.) According to such philosophers we should rely on a proper theory of explanation to delineate potential explanations from non-explanatory accounts and a delineation of understanding will follow. I shall argue that this is not a workable option.
One can also express the issue at hand in terms of the relative subjectivity or objectivity of explanation. Some theorists of explanation state an objectivity criterion for an account of explanation, and many others implicitly employ one. Wesley Salmon, for example, states clearly that the identifying criteria for scientific explanations must be objective, independent of personal, psychological considerations.
First, we must surely require that there be some sort of objective relationship between the explanatory facts and the fact-to-be-explained. Even if a person were perfectly content with an explanation of the occurrence of storms in terms of falling barometric readings, we should still say that the behavior of the barometer fails objectively to explain such facts. We must, instead, appeal to meteorological conditions. Second, not only is there the danger that people will feel satisfied with scientifically defective explanations; there is also the risk that they will be unsatisfied with legitimate scientific explanations. A yearning for anthropomorphic explanations of all kinds of natural phenomena for example, the demand that every explanation involve conscious purposes sometimes leads people to conclude that physics doesn't really explain anything at all.... Some people have rejected explanations furnished by general relativity on the ground that they cannot visualize a curved four-dimensional space-time. The psychological interpretation of scientific explanation is patentlyinadequate (1984, 13).
Salmon's rejection of the psychological interpretation of explanation can, when pushed, lead one to eliminate all considerations of the subjective sense of understanding from theorizing about the nature of explanation. I shall argue against this extreme position and clarify the proper place of considerations of understanding within theorizing about explanation.
I shall follow the common distinction between a potential explanation and a genuine explanation, or simply an explanation. A potential explanation is a set of propositions having all the characteristics of an explanation except, possibly, for truth. A genuine explanation, then, has all the characteristics of an explanation, and it is true (Hempel 1965, 338).
We should distinguish the mere sense of understanding from genuine understanding. One can get a sense of understanding from an account that is not wholly true. Genuine understanding can come only from a genuine explanation. By contrast, the sense of understanding one gets from a potential explanation containing false statements is not genuine. We can refer to it as sham understanding. At this point, the distinction between genuine understanding and sham understanding depends entirely on the distinction between a true explanation and a merely potential explanation. It is, in this sense, derivative. It depends on the notion of truth, however, not on any particular theory of explanation.
A sense of understanding can be sham understanding also because it is caused by an account that is not at all explanatory, regardless of whether its statements are true or false. Someone can claim to understand an event in virtue of considering a certain set of propositions but fail to have genuine understanding because those propositions, even if all true, do not explain the event. Hence, many philosophers would claim that criteria for what is explanatory (in a theory of explanation) are more basic, and criteria for understanding derive from them. If one has in hand an acceptable theory of explanation, one can impugn a sense of understanding as sham understanding either by showing (a) that it comes from an account that fails to measure up as being explanatory, or (b) that it comes from an account that is a potential explanation but untrue.
The foregoing description of the relationship between explanation and understanding sounds unassailable. It seems that one of the necessary conditions for understanding is that it be caused by consideration of an account that is at least potentially explanatory. The other necessary condition is that it come from consideration of a true account. One might go on to explain cases of sham understanding by finding, for example, similarities (perhaps only superficial similarities) between the accounts that generate the sham understanding and the genuine explanations of the relevant events or relevantly similar events. It would then appear that we can explain cases of sham understanding by reference to the true explanation, which is, to repeat, identified via a proper theory of explanation.
This cannot be the whole story about the relationship between understanding and explanation, however. Complications arise from a more careful consideration of the connections between the sense of understanding and explanation. We have noted that the sense of understanding need not come from an explanation or even from a potential explanation. A person might feel a sense of understanding a famine in terms of the appearance of a comet, for example, but this would be only sham understanding. Similarly, one can consider the true explanation for an event and fail to gain from it understanding of why the event occurred. Explanations citing very complex, abstract, or counter-intuitive principles of physics, for example, might fail to produce a sense of understanding. Looking only at these two points about the separability of explanation and understanding leaves the foregoing story about the conceptual primacy of a theory of explanation undisturbed.
We must note, in addition, that explanation and the sense of understanding cannot be completely isolated from each other. Consider the claim that some account, some set of propositions, is the explanation of an event E although no one ever has or could gain a sense of increased understanding of why E occurred by considering that account. We can imagine this account consisting only of true propositions, and we can imagine that the propositions are all topically relevant to E. What reason is there for calling the account an explanation of E rather than merely an accurate description of some facts related to E? I can find none.
Consider now an account consisting of propositions relevant to an event E such that nearly all of the very many psychologically healthy, intelligent, well-educated adults who think about it claim to gain an increased level of understanding of why E occurred. What reason could there be for denying that this account is at least a potential explanation of E? What good reason might one have for insisting that the account is merely a set of descriptive propositions somehow relevant to E? I can think of none.
I see no reason for calling some account an explanation of E if one has no idea even of how it could produce in someone an increased sense of understanding of E. And I see no reason for denying that an account is at least a potential explanation of E if many people claim that it provides them with an increased sense of understanding of E. A theorist of explanation might respond that the reasons I stubbornly fail to see come from an acceptable theory of explanation. We should call an account an explanation, such a theorist would claim, if it satisfies the criteria of a preferred theory of explanation, although we see no way for the account to increase anyone's understanding of the explanandum. Likewise, if an account fails to satisfy the conditions of some preferred theory of explanation, then it does not count as a potential explanation, even if many people seem to receive greatly increased understanding from it.
This last bold assertion of the conceptual priority of explanation over understanding clarifies why this strategy ultimately fails. To accept the ruling from a theory of explanation that a certain account is explanatory when one gains no sense of understanding from it, one needs reasons for accepting that theory of explanation as a correct theory. We therefore need to examine how a theory of explanation is justified, or defended against competing theories of explanation. The justification of a theory of explanation relies essentially, I shall argue, on using the sense of understanding as a guide to what accounts are potentially explanatory and what accounts are not.
There have been competing theories of explanation. Carl Hempel (1948, 1965) proposed the deductive-nomological model, the inductive-statistical model, and the deductive-statistical model as collectively capturing all the forms of scientific explanation. Richard Jeffrey (1969), James Greeno (1970), and Wesley Salmon (1970, 1971) objected to Hempel's claim that scientific explanations must take the form of an argument. Each proposed a competing theory of explanation (statistical-relevance or information-theoretic) as a correct account of scientific explanation. Later, Salmon revised his thinking about scientific explanation and proposed his causal theory of explanation with a statistical-relevance basis (Salmon 1984), thereby joining a long list of causal theorists of explanation. Criticisms and competing refinements have come from Peter Railton (1978, 1981), Philip Kitcher (1976, 1981, 1989), Peter Achinstein (1983), Bas van Fraassen (1980, chap. 5), and many others. The history of competing theories of explanation is complex, but a common theme lies in the methodology of defending a theory. A classic example will clearly illustrate the standard method of argumentation regarding theories of explanation.
The case of the flagpole and its shadow is an argument against Hempel's deductive-nomological theory of explanation. The example is attributed to Sylvain Bromberger, although he claims never to have published it (1992, 8). The argument states that one can deduce the height of a flagpole from the length of its shadow and the elevation of the sun (and appropriate laws regarding the propogation of light). The argument shows that such a deduction satisfies the conditions put forth in Hempel's D-N model of explanation for an acceptable scientific explanation of the height of the flagpole. The argument concludes that, since the length of its shadow clearly does not explain the height of the flagpole, Hempel's model does not present a set of sufficient conditions for scientific explanation.
The method illustrated here is the familiar method of counterexample. One argues against a theory of explanation either by presenting a set of propositions satisfying the criteria of the theory although it is not an explanation, or by presenting an explanation that fails to fit the criteria of the theory. The former method shows that the theory does not present a set of sufficient conditions for explanation, the latter shows that the theory does not present necessary conditions for an explanation. This method obviously requires some determination, independent of the theory under dispute, that the counterexample either is or is not an explanation. This determination depends, I claim, on the presence or absence of a sense of understanding generated in the counterexample.
A similar point applies to the defense of a theory of explanation. Theorists typically argue for their theory of explanation by demonstrating how it applies to a stock set of examples of explanation and a stock set of non-explanations, giving the right determination in every case. One might argue for the D-N model of explanation, for example, by showing that it does capture the sense in which one can explain the length of the flagpole's shadow by deducing it from the height of the flagpole, the elevation of the sun, and the appropriate laws. Such a defense of a theory of explanation must rely on some prior set of criteria for distinguishing explanations from non-explanations, and this set of criteria must be independent of the theory of explanation being tested in order to avoid an obvious circularity in the justification of the theory. It is of no use to show that one's theory of explanation succeeds in counting all explanations in a certain sample as explanations and all non-explanations in a certain sample as non-explanations if the samples were prepared using that theory of explanation itself.
The determination of stock examples of explanation and non-explanation, independent of a theory of explanation, must be based on the sense of understanding one gets from explanations but not from non-explanations. This follows from the peculiar nature of explanations as a class of statements. Explanations consist of descriptive statements. There is dispute about whether an explanation must take the form of an argument, whether an explanation must invoke laws, whether an explanation must cite causes. And there are many other disputes about the form explanations must take. To my knowledge, however, there is no dispute over the claim that explanations must consist of descriptive statements. They do not contain normative claims. They do not consist of questions (except perhaps as elliptical expressions of statements). They do not consist of imperative statements. A true explanation consists only of true descriptive propositions. Consequently, we cannot identify explanations as a class by syntactic means similar to the way we distinguish normative claims, questions, or orders.
Explanation is not merely true description, however. That is, explanation is description of a particular sort. So we cannot identify explanations as a class just by the syntactic means we use to distinguish descriptive statements as a class together with the semantic distinction of truth.
A prima facie determination that a set of propositions is explanatory for some event must rely on intuitions about the potential for those statements to produce a sense of understanding of why the event occurred. The determination cannot be made merely on the basis of topical relevance. For any event, there are many sets of topically relevant and true descriptive claims that are not explanatory. Using any criterion based on causal relevance, considerations of statistical relevance, unification, or any similarly controversial features will clearly beg the question against some theories of explanation. We require a more nearly neutral means of identifying explanations prior to theorizing, and this can come only from the subjective and, to some extent, relative sense of understanding.
The ineliminable function of a sense of understanding in theorizing about the nature of explanation poses the threat of a dangerous sort of subjectivity infecting our thinking about explanation. Subjectivity is indeed dangerous here, given the central place of inference to the best explanation in all scientific theorizing. The fear is that our distinguishing explanatory from non-explanatory hypotheses (and our assigning considerably more importance to explanatory hypotheses in science) might be based ultimately in a species-specific peculiarity in the way we humans happen to process information, rather than in any objective and natural features of the world. This fear is not unfounded, for peculiar conceptions of the nature of explanation have led people to accept bizarre views of the world. Some metaphysical theories, regarded as untenable even by strong defenders of metaphysics generally, have been defended on the basis of highly idiosyncratic views about what is required for adequate explanation of some aspect of experience.
This fear can be allayed, however. The subjective sense of understanding is necessary as an initial guide to the nature of explanation, but it is by no means the final word. There is a complex system of social correction of notions of explanation. An initial notion of explanation leads to and supports theories about the world by way of inference to the best explanation. But theories about the world in turn refine our notions of explanation. We expect that this process of refinement and correction leads us closer to an objectively accurate picture of the world and of the natural relationship of explanation. Filling in the details of how this process of refinement and correction works requires much work and obviously extends far beyond the scope of this paper. We must nonetheless fill in these details in order to secure objective validity for an inferential process that necessarily originates in relatively unconstrained subjective impressions but also serves to ground our theoretical advances in a scientific understanding of the world.
Achinstein, Peter. 1983. The Nature of Explanation. New York and Oxford: Oxford University Press.
Bromberger, Sylvain. 1992. On What We Know We Don't Know. Chicago: University of Chicago Press.
Greeno, James G. 1970. "Evaluation of Statistical Hypotheses Using Information Transmitted." Philosophy of Science 37: 279-93. Reprinted in Salmon et al. (1971).
Hempel, Carl G. 1965. "Aspects of Scientific Explanation." In Aspects of Scientific Explanation and Other Essays, 331-496. New York: The Free Press.
Hempel, Carl G., and Paul Oppenheim. 1948. "Studies in the Logic of Explanation." Philosophy of Science 15: 135-175. Reprinted with a Postscript in Aspects of Scientific Explanation and Other Essays, 245-295. New York: The Free Press, 1965.
Jeffrey, Richard C. 1969. "Statistical Explanation vs. Statistical Inference." In Essays in Honor of Carl G. Hempel, ed. Nicholas Rescher, 104-13. Dordrecht: D. Reidel. Reprinted in Salmon et al. (1971).
Kitcher, Philip. 1976. "Explanation, Conjunction, and Unification." Journal of Philosophy 73: 207-212.
________. 1981. "Explanatory Unification." Philosophy of Science 48: 507-531. Reprinted in Theories of Explanation, ed. Joseph C. Pitt, 167-187. New York and Oxford: Oxford University Press, 1988.
________. 1989. "Explanatory Unification and the Causal Structure of the World." In Scientific Explanation, ed. Philip Kitcher and Wesley C. Salmon, 410-505. Minneapolis: University of Minnesota Press.
Railton, Peter. 1978. "A Deductive-Nomological Model of Probabilistic Explanation." Philosophy of Science 45: 206-226. Reprinted in Theories of Explanation, ed. Joseph Pitt, 119-135. New York and Oxford: Oxford University Press, 1988.
________. 1981. "Probability, Explanation, and Information." Synthese 48: 233-256. Reprinted in Explanation, ed. David Hillel-Ruben, 160-181. New York and Oxford: Oxford University Press, 1993.
________. 1989. "Explanation and Metaphysical Controversy." In Scientific Explanation, ed. Philip Kitcher and Wesley C. Salmon, 220-252. Minneapolis: University of Minnesota Press.
Salmon, Wesley C. 1970. "Statistical Explanation." In The Nature and Function of Scientific Theories, ed. Robert G. Colondy, 173-231. Pittsburgh: University of Pittsburgh Press. Reprinted in Salmon et al. (1971).
Salmon, Wesley C. 1984. Scientific Explanation and the Causal Structure of the World. Princeton: Princeton University Press.
Salmon, Wesley C., with Richard C. Jeffrey and James G. Greeno. 1971. Statistical Explanation and Statistical Relevance. Pittsburgh: University of Pittsburgh Press.
van Fraassen, Bas. 1980. The Scientific Image. Oxford: Clarendon Press.