Operationism again: What Did Bridgman
Say? What Did Bridgman Need?
[Reply to Grace, R. C. (2001). On the failure of operationism. Theory and Psychology, 11, 5-33.]
Christopher D. Green
Department of Psychology,
Grace's (2000) critical analysis of the use operationism in psychology is founded on two basic claims. First, he argues that the main difficulty with mid-century psychologists' interpretation of Bridgman was that they believed a single operation to be sufficient to define a new concept. Second, he claims that the main thrust of recent criticisms of operationism is that operational definitions should be used for existing concepts only and not to create new ones. In his view, both the earlier difficulty and the recent criticism are effectively quashed by the adoption of "convergent operationism," first proposed in the 1950s. In the present commentary, I try to show that neither of the basic claims is quite correct, nor does "convergent operationism" do much to resolve the matter. The primary difficulty with operationism, in contrast to Grace's analysis of the situation, was that it was proposed as a solution to a metaphysical problem which it was not even remotely adequate to meet, and, in addition, that the presumed metaphysical problem it was intended to solve was, in the first place, a mistaken diagnosis of the difficulties that faced science in the 1920s, when operationism was first proposed.
It is testament to the power of Percy Bridgman's "operational attitude" that we are still debating its merits three-quarters of a century later. The idea was a simple one, aimed at solving a complex problem, and simple solutions to complex problems are always welcome. Unfortunately, like all too many simple solutions to complex problems, it turned out to not really solve the problem at all. Indeed we still haven't solved the problem, although thanks to the work of Bridgman, among others, we are a fair bit more sophisticated in our thinking about it than we were at the time of the relativistic and quantum revolutions in physics. What Bridgman's solution indirectly provided psychologists with, however, was a useful "rule of thumb" for experimental practice; one that would help guard against some of the "metaphysical" excesses of early psychological researchers, or so it was thought. However useful this "rule of thumb" may have been, it nonetheless fell far short of dispelling the quandary that bedeviled Bridgman and other scientists in the 1920s. I will work through some of these issues in attempting to respond to Grace's effort to reappraise Bridgman and what he once called his "Frankenstein monster": the operational definition.
Grace (2001) argues that the main difficulty with mid-century
psychologists' interpretation of Bridgman was that they believed a single
operation to be sufficient to define a new concept. According to Grace, however, Bridgman knew
that this was not sufficient, as evinced by the examples of multiple operations
he employed in his own writings.
Psychologists' misapprehension of the situation was dispelled, Grace
continues, with the arrival in the 1950s of "convergent
operationism," a collection of techniques that mandated the taking of
multiple measures of the variables under study, especially when such variables
are not "directly" observable.
Grace pays particular attention to an article by Garner, Hake, and Erikson (1956), but he also mentions some better-known
papers on related matters by Cronbach and Meehl (1955), and
Grace also argues that the main thrust of recent criticisms of operationism -- such as those by Leahey (1980), Koch (1992), and Green (1992) -- is that "operational definitions should be used for existing concepts only and not to create new ones" (ms. p. 4, see also ms. p. 32), but that this claim fails to undermine operationism because the "convergent operationism" of the 1950s resolves the problems that were inherent in attempting to define new concepts with single operations.
I think Grace has misread somewhat the thrust of my own position (Green, 1992). I had very little to say about whether operationists should use multiple operations to define their concepts, namely because I think it matters relatively little for the issue that drove Bridgman to propose his "operational attitude" in the first place (about which more below). What I repeatedly emphasized in my paper (see, e.g., p. 300) was a criticism first made public the year after Bridgman's book was published by L. J. Russell (1928), and which I think remains decisive. It is, as far as I can see, a matter with which Grace's paper never truly comes to grips. If operations literally define concepts (i.e., the concept is nothing but the operation by which it is defined), then each operation defines a new concept. For instance, length-as-measured-by-a-ruler, length-as-measured-by-triangulation, and length-as-measured-by-radar become entirely distinct concepts -- in every way as independent as are, say, length and mass -- or so it is implied under a relatively strict reading of Bridgman's "operational attitude." It is important to note that this is not, as is often claimed by Bridgman's defenders, simply a matter of illicitly raising a modest methodological "rule of thumb" to the level of a metaphysical principal, and then showing that it fails in a realm far more extravagant than that for which it was intended. It was, indeed, first noted by Bridgman himself (1927, p. 10), but he believed it to be something he could live with. Russell (1928), however, showed it to bear sufficiently ridiculous implications as to reduce the principle to philosophical rubble; viz., if length measured by a ruler and by triangulation are different concepts, then so are lengths measured by two different rulers, or by two different people, or even by the same person and ruler but at two different times. The only conceivable way one can make sense of the idea of multiple measurements employing one and the same concept -- and this is an absolutely crucial point that seems to elude many defenders of Bridgman -- is to regard the concept as somehow "apart from" the operations employing it. (Of course, exactly how one should gloss "apart from" continues to be a matter of philosophical debate). That is, length is just length (whatever it might be), and it is merely recorded, not literally defined into existence, by the measurement operation.
It is this, of course, that undermines Grace's contention that the problems of operationism were solved by "convergent operationism." If there is no concept "apart from" the operations used to measure its presence, then there is, to put it in its bluntest form, nothing for the operations to converge upon. Mightn't the operations merely converge upon each other? Of course, but without an underlying concept relating them, this would have no more significance than both a thermometer in the mouth and a tape measure around the belly both reading '37'. This argument has gone around and around many times over the last several decades. My suspicion is that Grace would respond that: (1) that Bridgman did not ever intend to produce a "dogmatic philosophical system," (2) that he did not mean term "logic" in his title to be taken so seriously, and (3) that he did not mean the term "synonymous" to be taken so literally in his most quoted principle (1927, p. 5) -- "the concept is synonymous with the corresponding set of operations" (all cited in Grace, ms. p. 11).
These points, though interesting from a biographical perspective, are ultimately, I think, beside the philosophical point. As great a physicist as Bridgman was, he was not a trained philosopher. Perhaps partly as a result, he displayed considerable inconsistency in outlining exactly what problem he was trying to solve, and exactly how he proposed to solve it. As such, pouring through Bridgman quotations to find out what he "really" intended is a fairly futile exercise. I happily concede that he often wrote, especially when his position was under attack, as though he did not want his position to be regarded as an "-ism." From a philosophical perspective, however, it seems relatively clear that what Bridgman needed in order to solve the philosophical problem before him was, in fact, a fairly grand metaphysical principle (the fact that it was, in a certain sense, anti-metaphysical, notwithstanding), and that to enunciate such a principle, the literal use of terms such as "logic" and "synonymous" was perfectly appropriate.
This brings me, at last, to what I think is a second difficulty with Grace's position. Namely, Grace does not seem to adequately take into account the philosophical and scientific context in which Bridgman was working. Newtonian physics had been universally regarded as the foundation of scientific thought for about two centuries (although published in 1687, the Principia was by no means immediately accepted as authoritative). The rest of physics had been built upon its basis. Indeed, it had eventually reached out to profoundly influence other natural sciences as well. The whole enterprise had been so apparently successful that in 1900, Lord Kelvin had the confidence to proclaim that the work of physics was very nearly completed. Then came along the theory of relativity and quantum theory, seemingly undermining the entire edifice. The immediate assumption was that scientists had made some terrible, grave error, and that the appropriate response was -- in words of Bridgman's that Grace himself cites (ms. p. 6) -- "to understand so thoroughly the character of our permanent mental relations to nature that another change in our attitude, such as that due to Einstein, shall be forever impossible." Bridgman was not alone in this desire. In a similar vein, Stevens, the psychologist who seems to have best understood Bridgman's original motivation, wrote that the primary aim should be bring about "the revolution that will put an end to the possibility of revolutions" (also cited by Grace, ms. p. 13).
The problem, as they saw it, was that "metaphysical" (i.e., unobservable) concepts had been illicitly allowed into scientific discourse, and that only by ruthlessly eliminating them would science finally be able to march steadily, progressively into the future. By contrast, the last century of science has shown that this was not really the error at all. Science, by contrast, appears to require theoretical concepts that are not fully defined in advance. As even the eminently conservative and respectable philosopher of science Carl Hempel (1952, p. 29) came to believe, scientific concepts must have an "openness of meaning" so that science can discover how the world works rather than decreeing it a priori. Instead, Bridgman's real problem, and that of his fellow scientists, was to believe that science can successfully develop in advance a definitive, eternal set of fundamental concepts that will never be subject to overthrow by future discovery or future thought. In short, the problem was misidentified by most of the scientists and philosophers of the day: "metaphysics" was not the main source of difficulty; their own fear of scientific revolution was.
Nevertheless, they acted aggressively to dispel what they took to be the problem. Given that Bridgman's aim was to eliminate the possibility of future scientific revolution, he aimed to generate a fundamental vocabulary of science that could never be overturned. Perhaps not unreasonably, he presumed that any operation he could perform could also be performed, at least in principle, by any future scientist as well. Thus, a scientific vocabulary based on such operations could never be overturned because its terms would always be available for "inspection," so to speak, by anyone who cared to do so. The idea, though seemingly straightforward, was shown to fail almost immediately by Russell (1928) among others. Even the logical positivists could see that it was unworkable as a solution to the problem of establishing a foundational scientific vocabulary. Their own related proposal, verificationism, failed soon after. Initially verificationism was the claim that the meaning of a term is just the means of its verification -- e.g., "Sugar is soluble in water" means simply that "If sugar is placed in water it will dissolve." The relation to operationism is obvious. After a number of unsuccessful attempts to resolve what had at first seemed to be minor problems with the logical formulation of the principle, however, verificationism (and the whole attempt to fully define scientific terms in advance) was pretty well scrapped in the mid-1930s in favor of Carnap's efforts at "partial reduction." As Grace concedes (ms. p. 10), in light of these events, Bridgman "modified his view of operational analysis considerably over the years" ("softened" was the term I used in Green, 1992, p. 296), but these changes served to remove it even further from the possibility of attaining its original goal -- to block the possibility of future scientific revolution by finally establishing an irrevocable fundamental scientific vocabulary.
Psychologists, however, seem to have been virtually unaware of these
philosophical developments, or if they were aware of them, they did not in the
main recognize them as having implications for their own work. Indeed, they did not, for the most part, seem
to fully realize that operationism had been proposed as a bulwark against the
possibility of future scientific revolution.
What they were looking for was much more modest -- they just wanted some
basic methodological principles that would serve to legitimize their
still-fledgling experimental practices.
Operationism -- at least as it was re-interpreted by Tolman -- seemed
well-suited to do so. But
Tolman, as I have previously argued (Green, 1992), effectively turned
operationism inside out. Instead
of replacing "metaphysical" terms such as "desire" and
"purpose," Tolman used it to legitimize them by giving them
operational definitions. One might be
tempted to argue that Tolman succeeded in eliminating the "metaphysical
portion" of these terms with his definitions, but it seems (at least to
me) that Tolman continued to be a mental realist
despite his apparent adherence to behavioristic principles -- his operations
did not literally define his
concepts; they were measures of the putative symptoms of the presence of what
the concept referred to (e.g., high blood pressure doesn't define anger, it merely (fallibly) indicates its presence). By contrast, the strict operationist does not
continue to be a realist about abstract length.
S/he defines length by the
measurement operation. As Bridgman put
the matter with respect to
By the time of the articles advocating "convergent operationism" touted by Grace, psychology was so far down the path begun by Tolman's reinterpretation of operationism that comparisons to Bridgman's original intent are difficult even to evaluate. Now, if all Grace wants to claim is that it is a good idea for experimentalists to obtain several different measures of the variables they are studying, especially when they are not directly observable, and that Bridgman, qua good experimental scientist, would have concurred, then he and I could not be in fuller agreement. But this is a far cry from the issues that drove Bridgman, qua philosopher of science, to propose operationism initially. Not only do the techniques of "convergent operationism" fail to establish a fundamental scientific vocabulary forever immune to revolution, they are indeed inconsistent with any reading of Bridgman's original "operational attitude" that is strict enough for it to have ever been considered as a candidate "solution" to the "problem" of scientific revolution. As was pointed out by Russell so many decades ago, it is not clear that there can even be multiple measures of a given variable if the variable is itself literally defined by the measurement operation. In the final analysis, this problem continues to undermine Bridgman's "operational attitude" as effectively now as it did near the beginning of the last century.
Percy W. (1927). The
logic of modern physics.
Campbell, D. T. & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, 81-105.
Cronbach, L. J. & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52, 281-302.
Garner, W. R. Hake, H. W., & Erikson, C. W. (1956). Operationism and the concept of perception. Psychological Review, 63, 149-159.
Grace, R. C. (2001). On the failure of operationism. Theory and Psychology, 11, 5-33.
Green, C. D. (1992). Of immortal mythological beasts: Operationism in psychology. Theory and Psychology, 2, 291-320.
Hempel, C. G.
(1952). Fundamentals of
concept formation in empirical science.
Koch, S. (1992). Psychology's Bridgman vs. Bridgman's Bridgman: An essay in reconstruction. Theory and Psychology, 2, 261-290.
Leahey, T. H. (1980). The myth of operationism. Journal of Mind and Behavior, 1, 127-143.
Logan, G. D., Coles, M. G. H., & Kramer, A. F.
In A. F. Kramer, M. G. H. Coles, & G. D. Logan (Eds.), Converging operations in the study of visual selective attention
Russell, L. J. (1928). [Review of The logic of modern physics]. Mind, 42, 355-361.