The ideas in this paper are based on an analysis of statistical explanation that uses the information transmitted by a theory. Consider a theory that specifies a probability distribution on the events of some domain, where for purposes of analysis we divide the variables that describe the domain into two sets: [M], a set of variables whose values are to be explained, and [S], a set of variables whose values are used to explain the values of the variables in [M]. The information transmitted by the theory is

(1)

where H(S) and H(M) are the uncertainties of the events in [S] and [M], respectively, and H(S x M) is the uncertainty of the joint events.

To illustrate this idea, consider the problem of explaining medical symptoms such as fever, coughing, skin rash and abdominal pain.

Since Professor Greeno's paper is, to my mind, both philosophically interesting and somewhat technical, it seems to me that the most useful function I can serve as a commentator is to try to provide some of the background necessary to understand the paper, and to attempt to put his discussion in some philosophical perspective. My main aim will be to relate his discussion to the issues that have been at the forefront of the philosophical discussion of scientific explanation in recent years.

It would be hard to dispute the claim that Professor Carl HempePs work has dominated the discussion of the nature of scientific explanation for the last twenty-two years - since the publication of the classic article with Paul Oppenheim on the logic of explanation.

The idea of using information-theoretical concepts to quantify the notion of explanatory power of theories is an attractive one. Where the theory is adequately represented by a probability measure p on Boolean algebra A of propositions, it is natural to think of measuring explanatory power by some function of H(A), the entropy or uncertainty of p. But recently, Greeno and others have been exploring measures of explanatory power which are suggested by more detailed representations of theories - representations which are themselves suggested by communication-theoretical analogies.

To avoid inessential complications, suppose that the Boolean algebra A is finite, and let ‘a’ range over the atoms of A, viz., the strongest propositions in A which are not logically false. (Any two atoms are logically incompatible, and any proposition in A is expressible as a disjunction of atoms.) Then we have

where we set p(a) logp(a)=0 in case p(a)=0.

The last two decades have witnessed a basic change in the philosopher's working model for capacities. The change has been from a stimulus-response model to a fine-structure model. The stimulus-response model was an attempt to understand capacity propositions by means of conditional propositions in which the antecedents expressed operations in certain circumstances and the consequents expressed realizations of the capacities in question. No attempt is made on this model to relate such conditional propositions to components of the entities with the capacities. An entity with a capacity is just a black box.

Now the conditional proposition to which a capacity proposition is related on the stimulus-response model is a modal conditional. It expresses the idea that the behavior characteristic of the capacity must follow upon those operations in those circumstances for an entity like one with the capacity.

My task in writing a paper to accompany Milton Fisk's complex and provocative one is unusually difficult. His argument is so condensed and at such a high level of abstraction that I could easily devote my entire discussion to a single paragraph of his, to his very first one for example. Instead of doing this, I thought it would be more helpful to say something about the historical background from which his essay derives, and then go on to raise some difficulties regarding his central claim that nature is something over and above part and property. I will end with some suggestions on my own part as to how the categories of capacity and nature might best be related.

But first it is worth asking: to whom is this paper addressed? Its language and general approach are not those to which readers of the philosophy of science are accustomed.

Fisk begins his discussion by asserting, that the FS model attempts to complete the SR model by accounting for the necessity with which, according to the SR model, a capacity will be exercised when certain operations are performed. He argues that the FS model ultimately fails in this task, and he seeks to improve upon the FS approach by grounding the relevant necessity in an entity's nature - a component that defenders of the FS model ignore or misinterpret. For the most part my comments will concern the general strategy of his paper rather than details of the arguments he offers.

Since the major focus of Fisk's paper is on the necessity with which, according to his description of the SR model, a capacity is exercised under certain conditions, it is worth emphasizing that this idea is pretty implausible.

“Philosophy of science without history of science is empty; history of science without philosophy of science is blind”. Taking its cue from this paraphrase of Kant's famous dictum, this paper intends to explain how the historiography of science should learn from the philosophy of science and vice versa. It will be argued that (a) philosophy of science provides normative methodologies in terms of which the historian reconstructs ‘internal history’ and thereby provides a rational explanation of the growth of objective knowledge; (b) two competing methodologies can be evaluated with the help of (normatively interpreted) history; (c) any rational reconstruction of history needs to be supplemented by an empirical (socio-psychological) ‘external history’.

The vital demarcation between normative-internal and empirical-external is different for each methodology. Jointly, internal and external historiographical theories determine to a very large extent the choice of problems for the historian.

The invitation which has brought me here to comment on Professor Lakatos’ paper has given me much pleasure, for I have long been an admirer of his work, particularly of his early four-part paper, ‘Proofs and Refutations’. That does not mean, of course, that we have often agreed, but I have enjoyed the arguments that resulted and looked forward to this one. My pleasure, furthermore, was considerably enhanced when I discovered that Lakatos was going to be able to confound all precedent, his own and others, by getting this paper to me well in advance. It is a privilege few commentators are given, and I am correspondingly grateful.

All that I could have said before opening Lakatos’ manuscript — in fact, I did so in letters to both Lakatos and Roger Buck. Reading it has only increased my satisfaction, but in an unanticipated way.

At the risk of being ostracized (if not annihilated) by the community of Popperians present, I wish to remark that Professor Lakatos is - and, I think - cannot help being, a second-level inductivist. If Professor Kuhn has pointed out (most eruditely) that science quite frequently is in a rut, and occasionally gets out of it (and into a new one), then Professor Lakatos appraises problem and theory shifts, and methodological innovations in the sciences, in the light of his criteria of ‘progress’ or ‘degeneration’. There can be little doubt that he wishes to serve (at least) in a critical and/or advisory capacity to scientists. But he can do that only if he ‘places his bets’, i.e., conjectures as to the fruitfulness of a method, and along with it of a theory engendered or supported by such a method along the lines of success or failure, whichever may be plausibly indicated.

How are we to decide between the different methodologies that have been proposed for science? In particular, how are we to decide between inductivism, conventionalism, falsificationism, and research programism? Lakatos says that we should use history of science to help us decide. He proposes, or at any rate seems to propose, the following criterion: Given competing methodologies we should prefer that methodology according to which more of the actual history of science is internal and rational, and more of scientists’ own judgments about science are correct. This criterion presupposes the following proposition, which Lakatos also asserts: The different methodologies lead to determinable and different dividing lines between internal (rational) history of science and external (empirical) history of science.

This paper is intended to be a small contribution to a future comprehensive Theory of Scientific Growth. I take it that such a theory would give an idealized description of the repeating patterns of growth found within the history of science and show how these developmental patterns are different from those found in the case of theories such as witchcraft, on the one hand, and the patterns found in the growth of the ‘practical arts’, such as pottery-making, on the other. The theory would go on to explain why such patterns might be expected to produce scientific knowledge by pointing out the critical forces which are operating and the rationality of the responses to them. In short, an adequate philosophical theory in this area should not only give the kinetics of scientific growth, but also the dynamics of that process.

Of the many problems involved in discussions of acceptability and justification I wish to consider in this paper what types of sentences can be used in order to justify a given claim or theory in science, what types of sentence can be appealed to in testing (showing to be acceptable or rejecting) a theory or claim. Rather than rehashing old critcisms of previous work, or constructing new counters to works already known to be inadequate, I propose to get on with the problem at hand. Necessarily this ‘getting on’ is in large part a taxonomic and programmatic venture. I wish to suggest, briefly and by examples, that in fact there have been, and are, at least four different types of bases to which appeal is made in testing or justifying a theory or claim. These four types are:

• (a) Observation reports.

• (b) Accepted Philosophical principles.

The line which separates theory from observation, once drawn with some assurance by students of the structure of science, has more recently come under a great deal of criticism. Some critics allow that there is a distinction to be made between theoretical claim and observational record, but maintain that the distinction merely indicates the character of the directions toward the extremes on a continuous spectrum of scientific statements and cannot usefully be taken as a fiduciary mark for the bifurcation of the language of science. Others, among whom is numbered the object of my attention today, Paul Feyerabend, will not admit the legitimacy of the distinction at all in the form in which it has customarily been cast. Under the distinction in this customary form, Feyerabend, it would seem, would want to judge all statements to be theoretical.

Although there continue to be many interpretations of the nature of science, not simply thought up from afar by philosophers, but in the highest ranks of scientific circles, there are only a few that are significantly different from one another and influential (Wisdom, 1971). I shall confine myself to instrumentalism, conventionalism, and induction. (I shall omit even operationalism as being easily subsumable under conventionalism.) The main thesis will be that they constitute different ways of regarding knowledge as rooted in pure observation, and that the philosophy of observationalism they presuppose is false. The falsifiability interpretation does not view knowledge in this way and is not discussed here.

Whatever the shortcomings of these three positions may be, however, firmly held or rejected they may be according as one assesses existing criticisms, these criticisms mainly lead to a judicial appraisal without pointing to anything new.

Any study of the philosophical literature dealing with the cluster of topics generally identified as ‘functional description’, ‘functional analysis’, and ‘teleological explanation’ naturally raises the problem of confirming, disconfirming, or at least relating the alternative logical models proposed by philosophers to the actual usage of biologists. A close examination of current biological literature reveals that acceptance or rejection of what philosophers or sociologists might call a ‘functionalist’ perspective or approach is not significant for the division of biologists into schools or factions. It is certainly not the case that a functionalist perspective distinguishes the wholistic or organismic faction from the molecular biologists. In fact, molecular biologists shift back and forth from ‘functional’ to simple causal perspectives with considerable indifference. This is particularly true of the literature based upon the so-called operon hypothesis developed by the Nobel laureates Jacob and Monod.

The concept of ‘feedback’, originally a technical term of cybernetics, has been so widely applied that it has become part of our everyday vocabulary. Though both scientists and philosophers have considered in detail a number of conceptual issues suggested by the perspectives of cybernetics, this basic concept has not yet received the critical attention it deserves.

Thus, while Taylor [1-3] Scheffler [4-5], and others have criticized Rosenblueth, Wiener, and Bigelow [6-8] for trying to define ‘teleological’ behavior in terms of negative feedback, their objections are directed primarily against the use of ‘naive’ or ‘black box’ behavioristic assumptions and the choice of the term ‘teleological’ to describe the phenomena in question. The concept of feedback is largely accepted by both factions on more or less intuitive grounds.

With the realization that the grounds upon which an hypothesis comes to be formulated can be considered separately from the grounds upon which it is accepted, it has become popular among some philosophers and scientists to claim that although it may be of psychological interest to understand the genesis of hypotheses, there can be no logic of search or discovery. While it is unclear exactly what is meant by the claim that there can be no logic of search, it is clear that this opinion, coupled with anecdotes of Poincaré's sudden solution of a mathematical problem while stepping on a Madrid streetcar, and Kekulé's vision of a snake biting its tail, have left the aura that the generation of an hypothesis is as mysterious as a Gestalt shift in perception of a figure.