I am concerned in this essay with a body of attitudes and themes which together constitute the core of contemporary empiricism in the philosophy of science. I am also concerned about them, and what I consider to have been their wholly detrimental influence on this field of study.
As a result, I will not trouble much with the specific differences among the various strains within empiricism—Logical Empiricism, Instrumentalism, Fictionalism and the like—but will instead concentrate on the framework of ideas within which these viewpoints were generated. To this end, I begin by making as fair a sketch as I can of the genesis and development of these ideas within their own context before passing to detailed criticism, evaluation and polemic. This in turn will engender some general conclusions about the implicit epistemology and ontology of science, as well as a reappraisal of the problems the philosophy of science should be studying.
Contemporary empiricism, in common with its classical predecessors, holds that not only is there no knowledge without sensuous experience a claim which only Mystics deny—but moreover that (true) knowledge is nothing more than what can be given immediately to the senses and what can be (validly) inferred from this sensuous core according to sound canons of reasoning.
This contention has drawn much support—unwarrantedly, it will be argued here—from the success of the experimental method in the physical sciences. For both the discovery of general laws of nature following detailed experimental investigations, and the procedure of testing theories by trying out their predictions with experiments, are easily construed as confirming that the basis of science lies in observation. And the triumph of scientific theories framed in the language of Bacon and Newton over their 'metaphysical' rivals of Cartesian or Leibnitzian persuasion has served to entrench this empiricist language all the more deeply into modern scientific writing, and to further the identification of scientific knowledge with the empiricist ideal.
The chief contrast between empiricism in the twentieth century and its earlier forms is clear at a glance—its central emphasis on logic. This is not surprising considering how much in vogue logical analysis was following the accomplishments of Russell and Whitehead in the foundations of mathematics, providing, as it seemed to, a natural framework for the positivist sentiments [125/126] that had been provoked by the Metaphysical Idealism of Bradley and the other Hegelians.
Yet despite this immaculate philosophical pedigree, much of the initial impetus for an empiricist philosophy of science came from within science itself. For a physical science insecure in the wake of the spectacular overthrows of long‑cherished assumptions by the new theories of Relativity and Quantum Mechanics, and for a social science embarrassed by its lack of comparable progress, it was a natural reaction to turn to a more thorough‑going empiricism than that in which their language was already half‑imbued in search of a safe and infallible foundation.
For these reasons the initial efforts of the logical empiricist programme were pushed strongly towards establishing a clear criterion of significance which could be employed to weed out the undesirable 'metaphysical' and speculative elements of science from the otherwise sound empirical foundation. From this pure core of empirical science, all valid scientific knowledge could then be derived.
This, it should be noted, is very much the intent of the Machian philosophy on which the young Germans and Austrians who largely initiated the modern empiricist philosophy of science were all weaned. Only now, if the pure core of positive science could be established through scientific observation, then all other valid knowledge could be deduced infallibly from it with the new logic. Thus all valid and meaningful knowledge could only derive from these two constituent parts, empirical and logical, so that with the reduction of mathematics to logic taken for granted, a criterion of significance was readily come by. Hempel has expressed this as: "A sentence makes a cognitively significant assertion, and can thus be said to be either true or false, only if it is
either 1) analytic or self‑contradictory
or 2) capable, at least in principle, of experiential test." [1]
The second criterion has been stated deliberately vaguely in order to cover the many formulations it has undergone. Its first crude manifestation, at the hands of the Logical Positivists, was as the Verifiability Criterion of Meaning. Roughly, this went: "A sentence has empirical meaning only if it can be empirically verified." When tailored to suit the new logicist programme it ran: "A sentence has empirical meaning if and only if it is not analytic and follows from some finite and logically consistent class of observation sentences.” [2]
In this form, the criterion was designed very much as an antimetaphysical weapon. However it was fearfully blunt, and as it stood would take out not only metaphysics but scientific method itself; in which, as Popper pointed out [3], a theory is first advanced and then tested against experience, and cannot 'follow' from observations.
However, Popper's criticisms were construed by the positivists not as pointing to a gross inadequacy in their account of scientific method—for they [126/127] were not attempting to reproduce science as it stood, along with all its speculative elements—but, significantly, as a purely logical objection: that, since no finite class of instances will serve to confirm a universal statement, the requirement of complete verifiability disqualifies as meaningless all statements expressing universal laws. It also disqualifies the negations of existential statements which themselves are meaningful and allows disjunctions of meaningful and meaningless statements.
This permitted them to dismiss Popper's doctrine of Falsifiability [4] on entirely analogous grounds: that it excluded existential statements from science, and thus denials of admissible universal statements too, whilst admitting conjunctions of meaningful and meaningless statements. [5]
Ayer attempted to bypass these difficulties by requiring for an empirically significant statement only that "some observation‑statement can be deduced from it in conjunction with certain other premises, without being deducible from those other premises alone." [6] However it was soon seen that such a criterion was far too liberal, and the problems with this formulation of the criterion proved intractable.
Consequently a new tack was adopted. Under the rising influence of 'philosophy of language', the failure of the verifiability criterion was blamed on the "liberal grammatical rules" of the natural languages in which the statements were expressed, rules which allowed sentences such as Bradley's infamous "The Absolute is perfect" which "make no assertion whatsoever". [7] As a remedy to this lamentable state of affairs it was now proposed that a sentence would have no meaning unless it was translatable into a language whose vocabulary and rules were empirically or logically interpretable in accord with the empiricist criterion of meaning. The effect of this new condition was to switch the emphasis away from statements and onto the terms that constitute the vocabulary of the empiricist language.
However this emphasis on terms was not new to the philosophy of science, for two attempts had already been initiated in the sciences themselves to 'purify' empirical knowledge by divesting it of all terms designating metaphysical concepts such as 'atom', 'gene', 'potential' and 'motive'—the so‑called 'theoretical terms'.
One such approach, of which Bridgman was the main protagonist, was that of the Operationists. Like the positivists, Bridgman was inspired by the work of Einstein [8]; and taking his cue from the concepts of length, simultaneity and mass necessitated by the Special Theory of Relativity, and the different experimental methods needed to measure them, he claimed that to know the meaning of any physical quantity was to know how to measure it, so that "the concept is synonymous with the corresponding set of operations". [9] Thus, like verificationism, operationism amounted to a formula for weeding out meaningless hypotheses and concepts, since if no experimental operations [127/128] could be devised to decide them and give them a value then they had no meaning. [10]
However, this hardly constituted a sophisticated theory of meaning. For, in order to allow room for the significance of mathematical and logical concepts, Bridgman was forced to extend his concept of operation to 'mental' and 'paper and pencil' operations [11], which rendered his criterion undiscriminating and consequently vacuous. Yet even if it is construed only as a methodological principle in science, operationism is beset with considerable difficulties.
For example, there is the problem that if each different operation procedure yields a different concept [12], there would be no way of identifying, for instance, the sundry masses of a body as measured by its acceleration under forces of being pushed, pulled, swung. attracted by magnets, blown in the wind etc. Acknowledging this sort of difficulty, Bridgman allows that such concepts can be tentatively identified on the grounds that no numerical difference between their experimental values has been detected. However, as L.J. Russell objected, if numerical identity can be a criterion for the identification of concepts, then the concept clearly is not synonymous with its corresponding operations. [13]
The reason that the identification of 'different' concepts could be only tentative lies in Bridgman's strictly empiricist motivation. "Concepts can be defined only in the range of actual experiment, and are undefined and meaningless in regions as yet untouched by experiment" [14], so that the identification of concepts, even if confirmed in some domain, is "loose" and always subject to further experimental verification.
Hempel objected to this that exactly the same "inductive risk" applies to the assumption implicit in the operational definition that the same operations will continue to produce the same results in the measuring instruments [15]: for a consistent operationism would then deny this assumption too, since concepts are defined in terms of actual, not potential, experiments. But in this event, the reduction of scientific knowledge to a fragmentary, inexplicable chaos would be complete.
In a similar vein the Behaviourist psychologists, most notably Hull [16] and Skinner [17], claimed that no hypothetical concept such as 'intelligence', 'freedom' or 'pride' has any meaning beyond the particular observable traits of behaviour associated with it, thus restricting themselves to finding regularities in the observable manifestations of behaviour alone, with no reference to any hypothetical inner processes. In fact, they went beyond this, and claimed that if a theoretical term does serve its purpose, and establishes a definite connection among observable phenomena, it is then dispensable, since any such connection will be replaceable by a relation directly linking observational antecedents to observational consequents. [128/129]
Curiously, though, this approach proved quite barren as a methodological procedure. The Behaviourists never managed to formulate any theory of behaviour which did not refer at least tacitly to motivations as distinct from observable traits, and the closer they came the more vacuous and unexplanatory were their theories.
Nevertheless, this argument for the dispensability of theoretical terms aroused much interest among empiricist philosophers of science, and the Craigian and Ramseyan methods of elimination were extensively discussed. It was left to the trusty Hempel, however, to argue that even if these terms could be replaced by means of such logical procedures, there was little point in doing so if theoretical concepts were methodologically and heuristically indispensable [18]. With no theory to unify them, empirical facts and inductive generalisations from them would remain diverse and unconnected: no new classes of facts could be subsumed under existing generalisations, no links could be forged between different domains of facts, nor even suggested, with the result that no scientific progress would be possible. And to allow concepts and hypotheses to perform these roles only to replace them afterwards would hardly count as a demonstration of their dispensability.
By itself, however, this argument could have solved the "Theoretician's Dilemma" only at the expense of accentuating further a deeper dilemma which had been straining the heart of the logico‑empiricist movement since its inception: whether its goal was the reconstruction of science or a more precise reproduction of its essence. The aspiration to hold scientific knowledge and method as a paradigm and the programme of reconstructing science on a strict logical and empirical basis were proving to be much more in conflict than the originators of the logico‑empiricist movement had originally supposed. If theoretical concepts are not directly definable in terms of observational ones (i.e. translatable into the empiricist language), then according to the empiricist criterion they are meaningless and should be eliminated. But how could this be reconciled with their obstinate indispensability?
Thankfully for the empiricists a way out of this predicament has been provided by the work of Carnap. This arose out of his criticism of operationism, which he had expressed in his characteristically careful and formal way. [19] No explicit definition of theoretical constructs in terms of their observational manifestations, he showed, would do. For if, for instance, a property Q were introduced in terms of the observational effects E displayed under some given conditions C by
Qx ≡ (Cx ⊃ Ex)
then x will always have the property Q if it does not satisfy the test conditions C (since Cx ⊃ Ex is only false if Cx is true and Ex false).
This flaw can only be avoided if the definition of a theoretical term in terms of the observational language is no longer required to be an explicit [129/130] one. [20] Carnap proposed instead conditional definitions or "bilateral reduction sentences" of the form
Cx ⊃ (Qx ≡ Ex)
Now the meaning of Q is given only when x in fact satisfies the conditions C—otherwise it is unspecified.
The importance of this suggestion of Carnap's did not lie in the notion of reduction sentences themselves, which were found to be too restrictive, but rather in that it gave rise to the notion of 'partial interpretation' of theoretical terms in terms of observational ones—the idea that the meaning of such terms was given by the open‑ended combination of a number of rules which related them to observational terms but did not explicitly define them. [21]
Inspired by the Polish logicians' semantics of mathematics to abandon his earlier purely syntactical account of science, Carnap developed this idea further [22], proposing that theories by themselves could be regarded as uninterpreted formal calculi, which could then receive indirect interpretation through rules linking them to observation sentences. More precisely, a theory is regarded as a set of theoretical postulates T expressed in a formal language whose non‑logical terms, the theoretical terms, are by themselves meaningless. The theory as a whole and the sentences and terms it comprises then receive what meaning they possess by means of 'correspondence rules' C, sentences connecting such terms with observational terms—that is, those terms denoting things, events, properties and relations which are directly observable, and whose meaning is therefore considered self‑evident.
This rendering of the meaning of scientific terms was seen as doubly advantageous over those requiring direct definition of theoretical terms in terms of observational ones in that it not only allowed for the construction of theories independently of the observational consequences they were found to possess, but also allowed for changes of meaning of theoretical terms when new relations among theoretical postulates were discovered, each in accordance with scientific practice. So it, seemed that, in so far as each was justified, the demands of both reconstruction and reproduction had been met. And it is around this notion of partial interpretation that the ranks of logical empiricism have closed.
So convincingly have they consolidated, in fact, that the philosophical position incorporated in this partial interpretation thesis and its concomitant methodological, epistemological and ontological implications has come to be known as 'The Received View’, and is treated with according respect. It is my view that this respect is quite unmerited, and having now laid out the main themes and perspectives which comprise this position, I intend in the criticisms that follow to substantiate my irreverence for it.
Up to this point we have been content to follow the generation of this viewpoint as it develops under criticisms which were mainly internal to it and [130/131] of a largely formal nature. But how well does it serve in its intended function. as a workable interpretation of science?
We have seen, for instance, how the attempt to purge science of the supposedly unfounded theoretical terms gave way to a sort of apartheid, with theoretical and observational kept strictly segregated, fraternising only under the dominion of correspondence rules. But is such a strict dichotomisation possible in a working science? Can the terms occuring in real science be unambiguously demarcated in this manner, or do we rather find that the same term can be theoretical in one context, yet can appear in an observational statement in another?
The response of the empiricists to such suggestions was a remarkably yielding one. Carnap appealed to the "customary and useful" division of the language of science [23], Hempel claimed that no precise criterion was needed [24], whilst Nagel agreed, noting that he could tell the back of his head from the front without needing a clear demarcation between them. [25]
The implication of these remarks is that there is a continuum of which the strictly theoretical and strictly observational are only conventional extremes, like hot and cold. But can empiricism accommodate such a relativity of observational and theoretical? Not, it would seem, whilst it accords ontic status to the former alone. For according to the empiricist ontology, an unobservable theoretical entity can only be said to exist in the same sense as a contour, being merely a "fertile explanatory concept” [26], whose only meaning is attained through connection with observable phenomena, and whose only purpose is to help effect "an economical systematisation . . . of a large class of particular facts and generalisations". [27] The reality of observable things and events, and their properties and relations, is on the other hand self‑evident.
Now clearly as long as one adheres to this ontology, to maintain that there is no rigid dividing line between observational and theoretical is to allow that a mere convenient fiction may yet be a real object. This alone is Idealism enough, but the obvious next step is to count all observables as explanatory fictions for organising sensory phenomena, and then the observational/theoretical distinction becomes redundant anyway. [28]
Perhaps recognising this tendency in modern empiricism towards idealism, others have sought, in the tradition of Locke and the later Russell, to allow for the possibility of the existence of entities that are not directly observable. but which underlie and are responsible for the observed phenomenal effects. Suppe, for instance, has claimed that not only is such an ontological commitment compatible with the partial interpretation thesis, but is actually entailed by it—“the commitment being that entities exist which have the observable manifestations specified by TC (theoretical postulates together with correspondence rules)". [29]
Plausible is it seems however, this conclusion proves more than it little [131/132] awkward, since the knowledge claim for the existence of such entities is neither empirical nor analytic. Consequently, if the meaning of scientific language is such that it contains this ontological commitment, then this contradicts the empiricist criterion of meaning on which the partial interpretation thesis is grounded.
Far from being troubled by their confusion on such consequential matters though, the chief protagonists of contemporary empiricism have with extraordinary disingenuity declared all such ontological controversies meaningless. In this they have fallen back not only on the empiricist criterion of meaning as if it were an a priori truth, but also on Carnap's doctrine of the conventionality of language. According to this remarkable doctrine, the fact that physics, for instance, is formulated in a language which ostensibly refers to objectively existing things and processes is no vindication of realism. It is merely an expedient adopted by working scientists for their convenience. It might equally well be possible to formulate a 'language' for physics which satisfied the criteria of a phenomenologist or an instrumentalist. But all such formulations in the "material idiom" are essentially ambiguous, and in order to avoid being misled by their apparent ontological connotations one should translate them into an idiom where their meaning is transparent. [30]
Hempel has taken up this theme in his treatment of the 'problem' of theoretical terms. He concludes that although these terms are pragmatically and methodologically essential, any claims for the independent existence of their references are no more valid than the claims of a strict instrumentalist. [31] Similarly, Nagel has said "The opposition between these views (Instrumentalism and Realism) is a conflict over preferred modes of speech". [32]
To illustrate this doctrine with an example, a palaeontologist may conceive of himself as discovering information about creatures that existed millions of years before from their fossil remains. According to the Received View, it is his privilege to speak this way, but he should understand that such a claim is entirely equivalent to saying that the concept of the past existence of these creatures is necessary for the ordering of the empirical evidence he has at hand, but this does not necessarily entail the assumption that they really existed.
Here we can see the extraordinary arrogance and presumption of this position! The working assumption that has made the entire field of modern palaeontology what it is, and has been responsible for all its achievements and advances, is not even necessary! (For clearly, if the creature's existence is just a concept, it, being a mental entity, can hardly have caused the production of material fossils.) And according to what new evidence does this pivotal assumption seem so shaky? An armchair theory of meaning (with an armchair far removed from any science library) and a philosophy of language on which it is kinder not to comment. [33] [132/133]
This gives us some answer to the question of the adequacy of the Received View as a workable interpretation of science. But it must now be examined in closer detail.
In a work generally regarded as the standard analysis of the nature of scientific knowledge, Nagel defends the orthodox theoretical/observational dichotomisation in terms of the distinction between theory and experimental law—a distinction for which he claims "three well‑marked differentiating features". [34]
The first depends on the claim that each 'non‑logical constant' in an experimental law (but in general not each in a theoretical statement) "has a meaning that is fixed by an overt laboratory or observation procedure" [35]—that is, each non‑logical descriptive term is either observational or operationally defined.
But if this criterion is taken seriously, an experimental law in, say, Elementary Particle Physics could contain no reference to energy‑momentum distributions, magnetic flux density intrinsic spin, baryon charge or any of the other physical phenomena that such laws do in fact refer to, but would appear instead as an unseemly mess of references to switches being turned on, needles moving, traces of vapour in little boxes, and so on. In other words, to remove 'theoretical concepts' from an experimental law at this level of scientific enquiry is to void the law of all significance whatsoever—a telling indication that the notion of observation giving meaning‑content to a theory is perverse if not inverse.
Nagel's second differentiating criterion is intimately bound up with the whole empiricist conception of scientific methodology, which is derivative of the model attributed to Bacon and Newton and taken up in earnest by Hume. The scientist, in his study of some particular phenomena, observes some regularity in their behaviour: a "constant conjunction", such as Boyle's law of the relation between pressure and volume in a gas at constant temperature. This constitutes an experimental law; it is founded (or at any rate. foundable) on inductive reasoning, and its phenomenal character lends it immunity against changes in theoretical interpretation. The theory, on the other hand, (which made no appearance in Hume's account), can not be a generalisation from observed data. It is a hypothetico‑deductive system whose function is to correlate and explain some set of experimental laws and facts.
Now just how it is that a theory can explain a set of experimental laws and empirical facts, when according to Nagel's own account its terms have no meaning beyond that conferred on them by their connection with precisely those laws and facts they correlate, is a problem that Nagel does not seem to appreciate—but we will return to this shortly. [36] Meanwhile, just how accurate is this picture of scientific method that Nagel has given? [133/134]
Clearly, scientists do not merely sit passively and watch for any regularity that might crop up. They will have preconceived ideas as to which traits of behaviour are important, as is the case with the laws of both Balmer and Wien which Nagel cites. Both are dependent on a substantial body of theoretical knowledge, and could not have been formulated without the use of theoretical concepts such as 'wave length' and 'radiation spectrum'.
Furthermore, experimental laws are hardly the immutable cornerstones around which theories rise and fall, as Nagel supposes. Balmer's law, to use Nagel's own example again, was accounted for precisely by Bohr's theory of the atom: both were soon found to be approximate and oversimplified. Progressively deeper understanding and more penetrating insights into the processes involved were attained by the successive theories of Sommerfeld and Wilson, and of Dirac, and then by the more advanced quantum field theories. Each of these gave rise to finer distinctions and more detailed predictions—and these were checked by correspondingly subtler empirical techniques and arguments—not by reference to Balmer's law.
Thus, in contradiction to Nagel's account, theory and empirical observation are in constant interplay, every advance in theoretical understanding demanding a corresponding increase in subtlety of empirical technique to test it. Any experimental law is expressed in terms of concepts which are part of some theoretical scheme—even if this scheme is not rigorous enough to be termed a theory [37]. The fact that many such theoretical concepts (such as 'weight', 'pressure' and 'temperature') have become part of everyday, and therefore observational, language is no licence to forget that they are a product of human thought and accordingly subject to thought's limitations—not God‑given, unalterable and true. An empirical law is as corrigible as the theories and assumptions it rests on and those it supports, and is therefore never completely true and rarely completely false.
This is the case even for those empirical laws obtained through inductive generalisation from observed data. [38] No induction can be performed outside a theoretical framework, with no presuppositions or preconceptions about how the phenomena are most usefully to be viewed and which aspects of them are the most pertinent. Moreover, even a casual perusal of the history of science shows that it is characteristic of any inductive generalisation that the 'data' it correlates will be found to be too imprecisely defined and that the correlation achieved will be oversimplified and valid only in a restricted domain. It is because of these failings that new theories are constructed, in order to suggest a better explanatory framework in which to describe what is happening. Not only experimental laws and inductive generalisations are explained with the help of a successful theory, but also their limitations and inadequacies.
We are now in a position to see why it is that theories have "relatively more inclusive explanatory power" and "greater generality" than inductive [134/135] generalizations [39] — Nagel's third criterion for distinguishing theories from experimental laws—a fact which, as we noted earlier, is inexplicable on the empiricist account of meaning. But in reaching this position we have seen all the epistemological significance of the distinction between observational and theoretical undermined.
So much for Nagel's defence of the orthodox dichotomisation of science. What further conclusions about the nature of scientific knowledge can be drawn from the failure of the accepted account?
First of all it will be apparent that, in their earnest but self‑indulgent scramble to perfect and protect the empiricist position, its proponents have almost completely failed to come to grips with the science they are supposedly interpreting. Their miserable picture of scientific knowledge as a painstaking accumulation of observations and empirical generalisations, with theory's role as mere correlator and predictor, not only does scant justice to the richness and vitality of scientific enquiry; it also leads to a view of scientific knowledge as somehow opaque and alien, and of an essentially different nature from non‑scientific knowledge. As we have seen, for the empiricists it becomes a "remarkable fact that scientific advances are accomplished by means of laws referring to hypothetical or theoretical entities [40], and a “perplexing problem" why such entities or the theoretical terms denoting them should be introduced at all. [41] But it is hardly surprising that such eminently natural strategies should seem 'remarkable' or 'perplexing' if one is perverse enough to regard a theory as an essentially meaningless formal calculus, which is interpreted in terms of clumsy laboratory operations and phenomena described in the unsophisticated concepts of a 'public observation language'.
The whole point of a theory is to explain something. [42] In order to explain, say, certain observations, it must present a conceptual scheme which provides a superior understanding to that supplied by the original 'observational' concepts. This will scarcely be possible if it is insisted that the new conceptual scheme can only be understood in terms of the old. [43] Thus the empiricist account of meaning, as hinted before, really is inverse! Contrary to Carnap's purely formal interpretation [44], it is logically necessary for a theory to be meaningful in its own right in order for it to provide the new insight that is its very function. And since, as we have seen, it is only within a theoretical context that observations can be made and interpreted, the meaning of all scientific terms and the observational statements, procedures and laws expressed in terms of them must derive from the theoretical contexts in which they function. [45]
The partial interpretation thesis, then, along with its accompanying clutter of pseudo-dichotomies and correspondence rules, must be entirely rejected. Its cloistered upbringing away from the realities of science has left it unable to cope when confronted with them, and unable too to recognize [135/136] the inadequacy of its own suppositions. Let us take a last searching look at these.
Contemporary empiricism in science, as we have seen, began as an attempt to purify knowledge. To this end it was felt necessary to exclude all those elements that were not amenable to the test of experience—hence the criterion of empirical meaning quoted near the beginning of this essay: "cognitively significant . . . only if capable . . . of experiential test". But even this first innocent step is a compound stumble!
There is, to begin with, the problem that, since nothing can be tested if its meaning is not in large part already understood, meaning must be prior to and distinct from testing, as is evident from the arguments above. Thus in defining meaning in terms of testability the cart was set before the horse at the outset of the empiricist programme. [46]
Nevertheless it is important to try to understand why this inversion was not seen as objectionable. The positivists had conceived of their task as the reconstruction of science—what appeared to be meaningful in the context of working science was therefore not important. (Their subsequent slide into reproduction obscured these artificial foundations.) [47] The point was that if some item of knowledge did not have consequences that could be checked by observation or experiment than there could be no evidence for (or against) it. Observational consequences were therefore deemed as epistemologically of prior importance to hypothesis and theory. [48]
Some telling reasons why they are not have already been discussed at length above; but there is a further confusion here concerning evidence and testing. For, requiring of a theory that it must be tested through its observational consequences not only ignores the fact that a theory does not entail its own consequences—these must be conjectured with the help of further scientific knowledge, which will itself be dependent on theory and therefore corrigible; it also presupposes that confirmations (and falsifications) of the theory's descriptions and predictions provide evidence directly for (or against) it that the weight to be attached to such results does not need to be theoretically evaluated. This is plainly false. Thales may have based his claim that the earth floats on water on observations—water gushing up through springs, movement (tremors etc.) of the earth's crust and so forth. He could have 'confirmed' his hypothesis through experiment by digging down a few feet into the earth (anywhere and at any time). No-one operating in the context of today's knowledge would want to claim that the evidence for his hypothesis was anything like conclusive: but now, as then, the only way to challenge his hypothesis is by evaluating its content and the evidence for it—that is, by showing how the same observations can be interpreted differently, and by confronting it with other accepted knowledge and evaluating consequent contradictions. [49] [136/137]
This brings me now to my final main point in this discussion: namely, that there can be no philosophy of science worthy of the name that does not deal with the content of science. I have stressed throughout this essay the artificial and self‑perpetuating character of the modern empiricist position and its isolation from the science it purports to describe. I have intimated that it is this dearth of contact that accounts for its lack of realism not only epistemologically but also in the nature of the problems with which it has concerned itself. [50] But not even close contact with real scientific problem situations is enough if philosophy of science still concerns itself only with problems of form the structure of scientific 'revolutions' as distinct from the content of the revolutionary ideas constituting them, rules for the acceptance or rejection of theories irrespective of their promise or value in the epistemological‑cum‑social context in which they occur, formal criteria of demarcation for science, and so forth.
This is not to deny, of course, that the study of form and the use of formal methods, when not undertaken in isolation from consideration of the content of scientific knowledge, is extremely useful and important. But philosophy of science as it is generally practised seems to have lost sight of the wealth of genuine problems already within science that should be the focus of its attention, so involved has it become in creating its own artificial subject matter by a snowball of internal criticism increasingly abstracted and removed from science itself.
To provide a selective sample of such problems, both specific and general. there is: the problem of what it is that accounts for the success of the theory of Quantum Electrodynamics despite its obvious mathematical and philosophical inconsistencies; the problem of whether or not the concept of a 'virtual particle' is coherent [51]; the difficulties with the present 'bootstrap' accounts of the elementary forces in nature [52]; whether the corpuscular and field‑theoretic metaphysics underlying physical science are mutually compatible or whether either could be separately sufficient [53]; whether any sense can be made of the concept of truth—especially as a meaningful desideratum for theories—in the context of a fallibilistic epistemology [54]; to what extent some of the more esoteric and mathematical scientific language is representational as opposed to directly referential, and how to interpret it in a realistic and more perspicuous fashion [55]; whether any sense can be made of the growing content of scientific knowledge in terms of a more enduring element than observation, theory or research programme, the relation and interaction between epistemological and socio‑political evaluation of theories and ideas in science—what actual values are inherent in the activity of scientific research; and many others. [56]
And not only the internal philosophical problems of science have been obscured by this neglect of its content. The wider social implications of the [137/138] new advances and ideas burgeoning on all sides have been left at the mercy of a few science fiction writers, a growing band of anti‑scientific philosophers and the compilers of sensationalised news reports.
It is both strange and regrettable that those most qualified to contribute to the philosophical enrichment of the sciences and the evaluation of the implications of scientific knowledge for society—scientifically trained philosophers—should have so completely abdicated their responsibility. It is the contention of this essay that the fault lies very largely with the tradition of artificial problems and poor philosophy with which philosophers of science have had to contend.
I very much hope that the critical evaluation of the empiricist backbone of this tradition which I have presented here does something to remedy the situation.
NOTES
1 C. G. Hempel, "Problems and Changes in the Empiricist Criterion of Meaning", in: L. Linsky, Semantics and the Philosophy of Language, University of Illinois Press, Urbana 1951, p. 163.
2 Ibid., p. 167.
3 K. R. Popper, Logik der Forschung, Springer, Wien 1935.
4 Roughly stated, this said that a hypothesis was not scientific unless capable of empirical falsification.
5 It should be noted that these objections only carry in this form if Popper's doctrine is first tortured into a thesis concerning the meaningfulness of statements as opposed to the scientificality of hypotheses, as Hempel distorts it in the work cited.
6 See A. J. Ayer, Language, Truth and Logic, Penguin, Middlesex 1946, especially p. 52 and p. 15. nb: The terms 'sentence' and 'statement' are used interchangeably, depending on the user's logical persuation.
7 C. G. Hempel, op. cit.,
8 See P. W. Bridgman, The Logic of Modern Physics, MacMillan, New York 1927, pp. 1‑2. G. Schlesinger (in his Method in the Physical Sciences, Routledge and Kegan Paul, London 1963) quotes a passage from Einstein's The Theory of Relativity (1920, p. 22) to the effect that unless an experimental criterion for simultaneity is given, the concept is meaningless. He also refers to passages in which C. G. Darwin, Frank and Dingle in addition to Bridgman, all claim inspiration from Einstein for their extreme empiricism.
9 P. W. Bridgman, op. cit., p. 5.
10 Ibid., See pp. 28‑37, for instance.
11 Ibid., p. 5.
12 Ibid., p. 10.
13 L. J. Russel, "The Logic of Modern Physics (Review Article)", Mind 37, 1928, pp. 355‑361.
14 P. W. Bridgman, op. cit., p. 7.
15 C. G. Hempel, "Operationism, Observation and Theoretical Terms", in: (eds.) A. Danto, S. Morgenbesser, Philosophy of Science, Readings, Meridan, New York 1960, pp. 101‑120. There is, incidentally, no little irony to this criticism of Hempel's. For it is surely precisely this [138/139] reluctance to take risks that has been one of the chief motivations for the empiricist programme—why else the concern for stable and infallible foundations, the rejection of hypothesis, and the refusal to commit oneself to the physical existence of entities remote from formal human experience—unless it be the fear of being found wrong?
16 See C. L. Hull, "The Problem of Intervening Variables in Molar Behaviour Theory", Psychological Review, no. 50, 1943, especially p. 284.
17 See B. F. Skinner, Science and Human Behavior, MacMillan, New York 1953. especially p. 35.
18 See C. G. Hempel, "The Theoretician's Dilemma", Aspects of Scientific Explanation. Free Press, N. Y., Collier‑MacMillan, London 1965, pp. 173‑226.
19 See R. Carnap, "Testability and Meaning", Philosophy of Science III, 4th Oct '36, IV, 1st Jan '37, especially pp. 439-444.
20 —accepting the material conditional as unbetterable for capturing the intent of the definition.
21 With its implication that there is a part of the meaning of a theoretical term which is not given through its relation to experience (what H. Feigl, "Existential Hypotheses: Realistic vs. Phenomenalistic Interpretations", Philosophy of Science 17, 1 1st Jan '50, called the 'surplus meaning' of a term), the notion of partial interpretation represents a departure from strict empiricism. Consequently, die‑hard positivists like Goodman felt impelled by "philosophical conscience" to only admit terms which were explicitly definable.
22 See for instance R. Carnap, "The Methodological Character of Theoretical Concepts". in: (eds.) H. Feigl, M. Scriven, Minnesota Studies in the Philosophy of Science, vol. 1, University of Minnesota Press, Minneapolis 1956, especially pp. 45‑47.
23 Ibid., p. 38.
24 C. G. Hempel, "The Theoretician's Dilema", (eds.) S. Feigl, M. Scriven, Minnesota Studies in the Philosophy of Science, vol. 2, University of Minnesota Press, Minneapolis 1958, p. 42.
25 E. Nagel, The Structure of Science, Harcourt, Brace and World, New York 1961, p. 83.
26 S. Toulmin, The Philosophy of Science: an Introduction, Hutchinson's House, London 1953, especially chapter 4, 'Theories and Maps', pp. 84‑125. Or in Carnap's more pragmatic phrasing, a term denotes a real entity only if the partially interpreted theory in which it occurs "is accepted" can be "guide one's expectations" by enabling "predictions about future observable events" (to be derived) from observed events with the help of T (the theory) and C (the correspondence rules) — R. Carnap, "The Methodological. . .", op. cit., p. 45; or if, as Nagel puts it, "available empirical evidence is sufficient to establish the adequacy of the theory as a leading principle for an extensive domain of enquiry". (E. Nagel, op. cit., p. 151.)
27 C. G. Hempel, op. cit., p. 87.
28 This, in fact, is not so much the next step as the previous one. The severance from modern empiricism of the subjectivism of phenomenalism and extreme positivism is seen as one of its crowning achievements and crusading goals (See I. Scheffler, Science and Subjectivity, Bobbs‑Merrill, Indianapolis 1967); what is not seen is the ontological incoherency of this compromise position, which entails such consequences as real (because observable) properties such as mass and charge being carried by fictional (because unobservable) particles.
29 F. Suppe, "On Partial Interpretation", Journal of Philosophy '68, 3rd Feb. 1971.
30 This idiom is the "formal idiom" or realm of "pure syntax" in Carnap's earlier locubrations (e.g. The Logical Syntax of Language. The Humanities Press, New York 1951); its later manifestation as the 'ideal empiricist language' we have already encountered.
31 C. G. Hempel, "The Theoretician's. . .," op. cit.
32 E. Nagel, The Structure. . ., op. cit., p. 152.
33 For criticism see A. Schaff, Language and Cognition, McGraw‑Hill, New York 1973.
34 E. Nagel, op. cit., pp. 81‑90.
35 Ibid., pp. 83‑84.
36 See following pages of the present work and also footnote 42.
[139/140]
37 Compare Bunge's dictum that it is not so much that observations are 'theory‑laden’ as that theory and observation are meaning‑laden.
38 —and by no means all of them are. Many more are the result of theoretical prediction or manipulation—for example, in most observation laws.
39 E. Nagel, op. cit., p. 89.
40 C. G. Hempel, "The Theoretician's . . .”, op. cit., p. 41.
41 Ibid., p. 43.
42 Van Fraassen, in a recent paper ("The Pragmatics of Explanation", forthcoming), disputes precisely this claim. He points out (rightly) that theories do not explain by themselves, but rather are used by people to explain things. Thus it is not true that "explanatory power is a virtue of theories by themselves, or of their relation to the world”; rather, analogously to counterfactuals, explanation is dependent on context and on the explainer's evaluation of which kinds of factors are explanatorily relevant, and consequently on his interests. But I cannot see that it follows that "There are no explanations in science"—unless science means just scientific theory, as opposed to the whole enterprise—or that theories are not designed primarily to effect explanations.
43 Feyerabend makes the same point in his Against Method (NLB, London 1975), especially chapter 17. But the alternative he proposes (p. 256) to explaining in terms of the old concepts arguing with unexplained terms—seems to me obscurantist. Can't one explain by promoting arguments that extrapolate from concepts already understood? Surely the meaning of an argument amounts to more than the meaning of the terms used in it?
44 "There is no independent interpretation for L1 (the language in which the theory T is couched). The system T is itself an uninterpreted postulate system" — R. Carnap, "The Methodological . . . op. cit., p. 47.
45 These concepts are by no means exhausted by the explicit theories one is operating with; there are all the concepts and implicit theoretical schemes as well. Compare Bunge's notion of ‘proto‑theories' — M. Bunge, Scientific Research, vols. I, II, Springer‑Verlag, New York 1967; Foundations of Physics, Springer‑Verlag, New York 1967.
46 — as Bunge has stressed in many publications. For his most recent views on meaning see his Semantics I and II, Reidel, Dordrecht 1974.
47 Again, cf. Feyerabend, op. cit., fn. 141 of ch. 17, p. 280.
48 See C. G. Hempel, "Operationism, Observation and Theoretical Terms". op. cit.: "all significant scientific statements must have experiential import" — "the latter consists in testability by suitable data of direct observation"; "sentences which are entirely incapable of any test must be ruled out as meaningless "pseudo‑hypotheses".
49 This conception of knowledge is hardly an invitation to relativism. An old theory could no more appear in its original form in the context of contemporary knowledge than could an extinct animal appear with the same genes as before.
50 In addition to the problems touched on in the preceding discussion have been all their artificial offspring—'the paradox of confirmation', 'the New Riddle of Induction', the 'problem' of counterfactuals, 'the indeterminacy of reference', etc. For an indictment of the first three, (and similar exhortations for a new philosophy of science), See M. A. Bunge, "Problems and Games in the Current Philosophy of Natural Science", Proceedings of XIV International Congress of Philosophy, II, pp. 566‑574. For a critical of Quine's views on reference, see R. T. W. Arthur, A Disquisition on Properties: Part I, MA dissertation, McGill University, Montreal 1976, ch. 2.
51 See M. A. Bunge, "Virtual Processes and Virtual Particles: Real or Fictitious?", International Journal for Theoretical Physics, vol. 3. no. 6, 1970, pp. 507‑508.
52 See also references in note 53.
53 Cliff Hooker has done much work on this problem. See his 'Metaphysics of Science: "Atom [140/141] vs. Plena"' (1974) and his forthcoming "Science and Metaphysics". See also M. A Bunge Treatise of Basic Philosophy, Reidel, Dordrecht 1974, vols. 3 and 4 for an alternative construction of the metaphysics of modern science.
54 See Bunge's Semantics II for a theory of partial truth. His Semantics I contains a realist theory of meaning. (both op. cit.). For criticisms of his construal of reference, and a discussion of a realist metaphysics of science centred on this notion, see my "Reference as a Component of Meaning of Science", forthcoming in Philosophica, Gent 1977.
55 For a prolegomena to a theory of physical properties, see my A Disquisition. . ., op. cit. For a formal theory, see M. A. Bunge, A. Sangalli, "A Theory of Properties and Kinds", to appear in International Journal for General Systems.
56 For a complementary sample of problems see M. A. Bunge, "Problems and . . .”, op. cit., and his numerous publications on inter-theory relations, axiomatics, and the concepts of time, casuality law, ontic level, state, system, property, kind and thing.
SOURCE: Arthur, Richard. "The Empiricist Account of Scientific Knowledge—A Polemical Evaluation," Poznan Studies in the Philosophy of the Sciences and the Humanities, vol. 3, nos. 1-4, 1977 (Amsterdam: B.R. Gruner Publishing Co.), pp. 125-141. (Issue theme: Aspects of the Production of Scientific Knowledge, edited by J. Witt-Hansen.)
Salvaging Soviet Philosophy (1)
Positivism vs Life Philosophy (Lebensphilosophie) Study Guide
CONTACT Ralph Dumain
Uploaded 25 October 2007
Site ©1999-2022 Ralph Dumain