Hypothesis and Regularity: Peirce on Explanation
David Boersema
Professor of Philosophy
Pacific University
Contact information:Department of Philosophy
Forest Grove OR 97116
Phone: 503 352 2150
Fax: 503 352 2242
Email:
Word count: Abstract – 51 words; Paper – 3040 words
Abstract
The purpose of this paper is to explicate Peirce’s views on explanation and to suggest briefly how they relate to contemporary models of explanation. In particular, I note why Peirce’s views run counter to the Causal model of explanation (and I take his concerns as appropriate criticisms of the Causal model).
Hypothesis and Regularity: Peirce on Explanation
Very soon after its enunciation by Hempel and Oppenheim (1948), the Covering-Law model of explanation was criticized on a number of grounds. Some criticisms focused on the internal inadequacy of the model. For example, Bromberger (1966) proposed his well-known flagpole counter-example. Other counter-examples quickly followed (cf., Kitcher and Salmon (1989)).
Besides criticisms that focused on the internal inadequacy of this model, other criticisms focused on its overall inadequacy. For example, Scriven (1962) early on argued that viewing explanation as an argument structure ignores the pragmatics of actual explanations. By the mid-1960s Gorovitz (1965) suggested that this model omitted the importance of causation in accounting for explanation. The upshot of numerous concerns about Hempel’s model as well as its underlying logical empiricist grounding was that by the beginning of the 1970s other models of explanation were proposed, including the Causal model (associated with Salmon), the Unification model (associated with Friedman), the Pragmatic model (associated with van Fraasen), and the Information model (associated with Hanna). The past several decades have witnessed an ongoing debated among adherents of these various models.
At the same time, there has been (for some folks, at least!) a resurgent interest in pragmatism in general and with Peirce in particular. In this paper I want to explicate Peirce’s views regarding explanation and make some remarks about how they relate to contemporary models, particularly in its opposition to the causal model.
Peirce on explanation
As several commentators (e.g., Hookway (1985, 2000), Potter (1967) and Reilly (1970)) have noted, Peirce offered a number of accounts, even definitions, of explanation throughout his writings. One account that is often cited by Peirce scholars appears in his “Reply to Necessitarians,” where he states: “Explanation, however, properly speaking, is the replacement of a complex predicate, or one which seems improbable or extraordinary, by a simple predicate from which the complex predicate follows on known principles.” (CP, 6.612) Elsewhere he remarks that “every scientific explanation of a natural phenomenon is a hypothesis that there is something in nature to which the human reason is analogous” (1.316), as well as the claim that “An explanation is a syllogism of which the major premiss, or rule, is a known law or rule of nature, or other general truth; the minor premiss, or case, is the hypothesis or retroductive conclusion, and the conclusion, or result, is the observed (or otherwise established) fact.” (1.89)
These several quotations highlight a number of features of Peirce’s conception of explanation: (1) there is a structure to what constitutes an explanation (explanations are syllogisms), (2) explanations have an irreducible epistemic aspect (they involve what seems improbable or extraordinary), (3) they have an irreducible axiological aspect (they are hypotheses that there is something in nature to which human reason is analogous), and (4) explanations involve categorization(s) (they involve predicates, some to be replaced by others),. I will flesh out each of these features.
In his discussion of Peirce on explanation, Hookway states that Peirce’s view “is similar to that of the logical empiricists: we explain an event by invoking a law, in the light of which the event was likely to happen” (1985, p. 266). Given comments such as the following from “Some Consequences of Four Incapacities,” this interpretation of Peirce is quite understandable: “…explanation consists in bringing things under general laws or under natural classes” (5.289). While it is true that Peirce does invoke law in enunciating the structure of explanation, and while Peirce outlines a structure to explanation that is reminiscent of Hempel’s model, there are several aspects of Peirce’s view that set it apart from the logical empiricists. Before detailing those distinguishing aspects, I will say more about Peirce’s notion that explanations have a (syllogistic) structure. Peirce declares several times, besides the quotation given above (1.89) that an explanation is a syllogism. For example, in his analysis of “Induction and Hypothesis” he claims: “We have seen that Inductions and Hypotheses are inferences from the conclusion and one premiss of a statistical syllogism to the other premiss. In the case of hypothesis, this syllogism is called the explanation” (2.716). Framing the structure of an explanation as a hypothetical (or abductive) inference is fundamental for Peirce as it does, pace Hookway, require invoking law. Two aspects of law are particularly important: (1) that isolated facts never call for explanation and (2) explanation invokes law in the sense of drawing a necessary connection between the explanandum and explanans. In rejecting the explanation of isolated facts, Peirce provides the following example (involving the question why it is so difficult to walk on ice):
It is…not the simple fact that ice is hard to walk on which creates the demand for an explanation; it is, on the contrary, a puzzling nexus of facts. Tell a man who never saw ice that frozen water is very hard to walk on, and he may ask whether the feet stick in it, or put other questions in order to figure to himself what you mean; but as long as the fact is apprehended by him as a simple one, he will no more ask why it should be so than a common man asks why lead should be heavy. The fact is entirely sufficient as long as it is simple and isolated. It is when the difficulty of walking on ice is compared with the extraordinary distance that a ball can be bowled upon it, or with such other facts as would naturally lead one to expect that ice would be particularly easy to walk on, that a scientific explanation is sought…An isolated fact is precisely what a demand for an explanation proper never refers to; it always applies to some fact connected with other facts which seem to render it improbable. (7.200)
In the drawing of a necessary connection between the explanandum and the explanans, Peirce is equally explicit:
…[nominalistic explanations] merely restate the fact to be explained under another aspect; or, if they add anything to it, add only something from which no definite consequences can be deduced. A scientific explanation ought to consist in the assertion of some positive matter of fact, other than the fact to be explained, but from which this fact necessarily follows; and if the explanation be hypothetical, the proof of it lies in the experiential verification of predictions deduced from it as necessary consequences. (6.273)
These aspects of law (or connectedness) that speak to the structure of explanation also speak to the second feature of Peirce’s conception of explanation, viz., explanations have an irreducible epistemic aspect. Only by seeing a fact (e.g., difficulty of walking on ice) as an instance of some contrast class of facts can the notion of explanation be relevant. As various commentators have noted, this relates to what Peirce took as the sort of things that call for explanation. Potter (1967), for example, notes that Peirce thinks that sheer irregularities (such as the natural placement of trees in a forest or the shapes of grains of sand) do not call for explanation, since we aren’t surprised by such irregularities and purely formal regularities (such as laws of probability) do not call for explanation, since they are part of the a priori conditions of our knowing randomness at all. On the other hand, empirically observed regularities in nature do call for explanation since they are unexpected and breaches in empirically observed regularities call for explanation since they, too, are unexpected. As Peirce puts it: “…a uniformity, or law, is par excellence, the thing that requires explanation” (6.612). Likewise, in “A Guess at the Riddle,” he claims: “Indeterminacy, then, or pure firstness, and haecceity, or pure secondness, are facts not calling for and not capable of explanation. Indeterminacy affords us nothing to ask a question about; haecceity is the ultima ratio, the brutal fact that will not be questioned. But every fact of a general or orderly nature calls for an explanation…” (1.405). Quite simply, what matters is what is expected – given background assumptions (and theory) - such that some explanandum is placed in the context of a contrast class that calls for explanation. Surprise, even the strength of surprise, of course, is not the measure of a logical need for explanation, for Peirce (see 7.190). It is, as he puts it, “merely the instinctive indication” of a logical situation. Nevertheless, prior expectation or probabilities (provided by experience or explicit theory) structure what is taken as needful and possible as explanation.
Along with the central element of surprise/expectation that is integral to the epistemic aspect of explanation is the notion of interest. Says Peirce: “…we pay no attention to irregular relationships, as having no interest for us” (1.406). The notion of explanation as interest-laden (or, perhaps, interest-guided) bridges the focus on the epistemic aspects of explanation and the axiological aspects. Explanation is part of inquiry, which, for Peirce, is never disinterested. Just as he rejects Cartesian doubt as a serious undertaking, so, too, he rejects explanation disconnected from genuine puzzlement as a serious undertaking (or desideratum). What is serious is the irreducible element of purpose. With respect to scientific explanation, and, so, the purpose(s) of science, knowledge and predictability are the desired end:
In order to define the circumstances under which a scientific explanation is really needed, the best way is to ask in what way explanation subserves the purpose of science…Now what an explanation of a phenomenon does is to supply a proposition which, if it had been known to be true before the phenomenon presented itself, would have rendered that phenomenon predictable, if not with certainty, at least as something very likely to occur.
It may, however, be objected that if we are to go back to the ultimate motive for explanation, I should have asked what the danger is to which error would expose us…restricting myself, as I do, to scientific reasoning, I need not go behind the recognized purpose of science, which stops at knowledge. (7.192, 201)
The emphasis noted earlier on law and necessary connection as integral to explanation point to the fourth aspect of Peirce’s position, that explanations involve categorizations and the replacement of complex predicates with simpler ones. “Complex” and “simple” are, of course, relative terms and what Peirce means by them in this context is broader or narrower reference classes. For example (though this is not one that Peirce gives), if Smith is 30 years old, but dies within five years, we would be surprised and ask for an explanation. If we come to discover that Smith was a heavy smoker, we now think Smith’s death in such a short amount of time is explained. This is because we now place Smith into a narrower reference class (heavy smokers) rather than the broader reference class (30 year olds) and our respective expectations about longevity are different. A constraint on the breadth or narrowness of the reference class is the element of relevance, though Peirce never states this explicitly. For example (again, not his), if the proffered explanation for why Smith got pregnant is because Smith failed to take birth control pills, that would function less felicitously if we learn that Smith is male than merely if we learn that Smith is a non-pill-taker.
What can be culled, then, about Peirce’s view on explanation? As a topic within the broad category of inquiry (and this is important for him; it is a topic within this broader category), it relates metaphysical issues such as his commitment to Law (or Uniformity or Regularity) as a third fundamental kind of category (as opposed to quality and individuality). It also relates to epistemological issues such as expectation and predictability, and to axiological issues such as purpose and interests. But for present concerns, how does Peirce’s view on explanation relate to contemporary models?
Peirce and contemporary models
How do Peirce’s views relate to various contemporary models of explanation? From Peirce’s perspective, Hempel is quite right to insist that when we ask for an explanation of some phenomenon, we want to relate it to other facts that we connect to that phenomenon and we do this relating via placing it in the context of laws. Likewise, this is the demand of the Unification model, though they drop Hempel’s logical empiricist presumptions about explanations being an argument. Like the Covering-Law and Unification models, Peirce insists there is a logical structure to explanation, that it involves the connection of some phenomenon to be explained with law and regularization. This speaks not only to the importance of law, but also to categorization and the partitioning of phenomena in ways that demonstrate and reveal regularization (albeit interest-guided regularization, not necessarily an element of the Covering-Law or Unification models). Clearly, as the Causal model insists, many explanations are calls for establishing causal connections. Sometimes demonstrating a causal connection is sufficient to resolve the problem that we face by the appearance of that particular phenomenon we want explained. Other times correctly indicating a causal account is not sufficient. Simply showing that a caused b might well not resolve the problem or puzzle at hand. Indeed, demarcating just what parts of the causal nexus are relevant and, so, part of the cause (vs. a cause), is in part a matter of discovering facts about the world but also is in part a matter of what factors are the ones we are interested in. Like Salmon says, we want statistical relevance, but what will count as being statistically relevant will be, again, in part a matter of facts about the world (i.e., what things really do cause other things to happen) and in part a matter of what problem we are interested in resolving. As Gorovitz argued, it’s the differentiating factor among that causal nexus, relative to the problem we care about, that is what we want in the explanans. These concerns about the Causal model – that it leaves out epistemic and axiological elements of explanation – are ones that Peirce shares. So Peirce states:
…I cannot admit that an explanation is a description of the fact explained…In all cases, it is other facts, usually hypothetical, which constitute the explanation; and the process of explaining is a process by which from those other facts the fact to be explained is shown to follow as a consequence, by virtue of a general principle, or otherwise. (6.612)
…if we should find that this object which seemed white, in the first place was white, and then that it was a crow, and finally that all known crows were black, then the fact of this seeming and really being white would require explanation. It might be an albino, or it might be some new species or variety of crow. But perhaps it will be insisted that this thing’s appearing white does call for an explanation; - that we want to know the cause of its being white. To this I reply that it has always been agreed that the tendency of the understanding was merely toward synthesis, or unification. Now no fact could possibly be more unified and simply than the fact that this is white, taken in itself. It would seem therefore, that, if we consider this fact isolated from all others, it completely accomplishes the tendency of reason. To find a cause for the whiteness would only be to complicate our conception of the matter…. (7.198)
Underlying his concern that attempts to explain by enunciating causes are insufficient (because they omit the ineliminable epistemic and axiological aspects), is Peirce’s conception of causation, which includes three important aspects: (1) facts vs. events, (2) law, or regularity, and (3) purpose vs. final cause.
Included in his rejection of nominalism, Peirce remarks that “it is the ineluctable blunder of a nominalist, as Boethius was, to talk of the cause of an event. But it is not an existential event that has a cause. It is the fact, which is the reference of the event to a general relation, that has a cause.” (CP, 6.93) Of course, as we commonly use the expression “every event has a cause,” Peirce would agree, but only if this is understood to mean that any given event is a selected, or, abstracted, portion of the world (i.e., a fact). This is not simply the claim that we can characterize various events under different descriptions. Rather, as Manley Thompson (1953) put it, for Peirce ‘cause’ is to be understood “as meaning a statement of the conditions which must obtain whenever a particular sort of event occurs – the conditions which make up the property of being an event of such and such a kind.” (1953, p. 94, emphasis added) And this points to the second aspect of causation, viz., law, or regularity.
Using Aristotelian terms, Peirce frequently discusses efficient cause and final cause. Both, he claims, are necessary for characterizing causation:
Efficient causation is that kind of causation whereby the parts compose the whole; final causation is that kind of causation whereby the whole calls out its parts. Final causation without efficient causation is helpless; mere calling for parts is what a Hotspur, or any man, may do; but they will not come without efficient causation. Efficient causation without final causation, however, is worse than helpless, by far; it is mere chaos, and chaos is not even so much as chaos, without final causation; it is blank nothing. (CP, 1.220)