Who Has the Last Word? Radical Behaviorism, Science, and Verbal Behavior about Verbal Behavior

Abstract

A radical behaviorist epistemology recognizes the recursivity inherent in behavior analysis: as behaving organisms, we not only take behavior as our subject matter but we are also part of it. Such a naturalization of epistemology, however, is not without its critics. In this article, my aim is to assess some of the arguments that were directed against this approach by the American philosopher Thomas Nagel in his book The Last Word (1997). In particular, I address Nagel's arguments regarding (1) the shortcomings of naturalistic explanations of scientific knowledge and (2) the impossibility of circumventing a realistic, representational epistemology. Regarding (1), I argue that although Nagel is right in arguing that there is no neutral or external viewpoint from which we can understand scientific knowledge, the naturalistic explanation of such knowledge proposed by radical behaviorists is not only possible, but have important practical advantages, insofar as it allows the identification of the variables that control scientific behavior. Regarding (2), I argue that although behavior scientists will frequently talk and write in descriptive ways, the function of descriptive verbal behavior in science is not to represent reality but to coordinate our collective behavior in dealing with the environment. I conclude that instead of avoiding an evolutionary account of rationality, as Nagel suggests, we have every reason to further pursue it.

There is a recursivity that is characteristic of all behavioral sciences, including behavior analysis: we take behavior as our subject matter and we are also, as behaving organisms, part of this subject matter. Anything that applies to human behavior must also apply to the behavior of scientists. Behavior scientists, in particular, are inclined to understand scientific behavior (verbal or nonverbal), including their own, as the product of the same variables that affect any behavior (Normand, 2019; Skinner, 1956/1972a, 1945/1972b, 1957, 1963/1969, 1974; Smith, 1986; Zuriff, 1980).

From Skinner's point of view, this recursivity is a natural and desirable development of a science of behavior. If logical, mathematical and scientific verbal behavior could themselves be understood with the methods and concepts of behavior analysis, this could ultimately provide us with an "empirical epistemology" (Skinner, 1963/1969, p. 228), including an "empirical logic" (Skinner, 1945/Skinner, 1972b, p. 380; Skinner, 1957, pp. 430–431). This new, empirically based approach, with methods better suited for its subject matter, would enable significant advances in our understanding of the behavior involved in the production of scientific knowledge, then overcoming the traditional approaches offered by the philosophical disciplines of epistemology and logic.

The possibility of taking scientific behavior as a scientific subject matter in its own right has a long history, which can be traced at least to Hume (1777/Hume, 1996). The resulting approach is frequently called, after Quine (1969), a naturalization of epistemology, in the sense that scientific methods would be applied to the understanding of scientific knowledge itself. However, the very possibility of such recursivity and the status of its possible outcomes are not without its critics.

In this article, my aim is to assess some of the arguments that were directed against this approach by the American philosopher Thomas Nagel in his book The Last Word (Nagel, 1997). The book is entirely devoted to a critique of attempts to naturalize epistemology, which would try to undermine what the author regards as the authority of reason as the basis of scientific knowledge. In what follows, I will address Nagel's arguments regarding (1) the shortcomings of naturalistic explanations of scientific knowledge and (2) the impossibility of circumventing a realistic, representational epistemology. Although Nagel's arguments were not directly aimed towards behavior analysis, they are potentially applicable to it as a version of naturalized epistemology, and its consequences to radical behaviorism as a philosophy are worthy of proper assessment. What emerges from such an analysis is a sharp opposition between Nagel’s views on the nature and functions of the scientific enterprise and a radical behaviorist one. I will argue that the naturalistic explanation of scientific knowledge proposed by radical behaviorists is not only possible, but have important practical advantages, insofar as it allows for the identification of the variables that control scientific behavior; and that the function of descriptive verbal behavior in science is not to represent reality but to coordinate our collective behavior in dealing with the environment.

In spite of the fact that I use The Last Word as the source for my analysis of Nagel’s (1997) arguments, my aim goes beyond just reviewing the book. I try to present an in-depth analysis of the author’s arguments regarding logic, mathematics, and science. It must be noted that Nagel’s book also has a section on ethics, and although its content is also of interest to behavior scientists a proper assessment of it would require a different line of inquiry. Here I concentrate on the implications of Nagel’s arguments for radical behaviorism as a philosophy of science.

As in most nonbehavior analytic literature (most notably when it comes from a rationalist tradition such as Nagel’s), the arguments of the book are couched in mentalistic jargon, but this does not preclude its potential interest to behavior scientists. Moreover, I assume, with Burgos and Killeen (2019), that mentalism (or rationalism for that matter) is not a single, monolithic and easily definable view, but “a family of positions” (p. 244)—and thus I try to identify the arguments I am examining as precisely as possible.

Nagel's Objections to the Naturalization of Scientific Knowledge

Can We Have a Naturalistic Explanation of Science?

Nagel’s arguments in The Last Word (Nagel, 1997) are a contemporary stance of a rationalist tradition in the philosophy of science that dates back to Descartes and even Plato. This is recognized by Nagel himself. He states that “the conception of the authority of human reason that I want to defend is very like that of Descartes” (p. 18), and argues that “basically . . . Descartes’s cogito is correct” (p. 92).1

Nagel (1997) draws inspiration from the famous Cartesian insight that thinking cannot get outside of itself: when we think about thinking we are thinking. That being the case, for Nagel any attempt to understand thinking "from the outside" (e.g., in biological, psychological, sociological or anthropological terms) will necessarily be limited by the fact that this understanding itself must be carried out by rational means. By "thinking" or "reason" Nagel means mathematics and basic logic, including "the reasoning that goes into the generation and elimination of scientific hypotheses suggested by the available evidence" (p. 23). Nagel dubs himself a rationalist and a realist, and calls into question epistemological positions which he attributes to "subjectivists" and "relativists," most notably what he sees as their attempts to undermine reason as the final basis of epistemological judgment.

Nagel submits that an argument must be faced with another argument, following the commonly accepted rules of reason (logic and mathematics) —not with the observation that, after all, any argument is just an historical, behavioral, social, or contingent phenomenon. That doesn't mean, for Nagel, that reason has any inherent infallibility, for arguments can be refuted or displaced. But the limits of any rational discussion are to be found within the arguments themselves, including the data that support them—not in their explanation as natural and social phenomena:

Everything depends on the outcome of this peculiar contest over the last word. The subjectivist wants to give it to the recognition that justifications come to an end within our language and our practices. I want to give it to the justifications themselves, including some that are involved or implicated in that recognition, which is subordinate to them, just as the recognition that a notation is essential for thinking about arithmetic is subordinate to arithmetic itself. (p. 34)

If one thinks in logic, one cannot simultaneously regard those thoughts as mere psychological dispositions, however caused or however biologically grounded. ... One cannot embed all one's reasoning in a psychological theory, including the reasonings that have led to that psychological theory. The epistemological buck must stop somewhere. (p. 137; emphasis in original)

Thus, scientific statements are to be evaluated within the limits of reason, and any attempt to evaluate them "from the outside" would necessarily employ reason again. The attempt to explain or otherwise evaluate logic as just a behavioral, social practice would amount to "explain the more fundamental in terms of the less fundamental" (p. 39). Reason, in this sense, is considered by Nagel to be a basic type of procedure from which all other forms of knowledge depend, and these forms could not displace reason from its foundational role. Thus, Nagel takes issue with any suggestion that the "universal claims of reason" (p. 99) could be merely "relative," "subjective," "local," or "contingent." The universality of such claims would be beyond any possible doubt: "Nothing would permit us to attribute to anyone a disbelief in modus ponens, or in the proposition that 2 + 2 = 4" (p. 77). Deciding rationally if any statement is true would amount to "thinking about the evidence and the arguments and being open to consideration of whatever anyone brings up as relevant" (p. 89), and this would be an inescapable and fundamental scientific procedure, not subject to any further analysis that doesn't employ rationality itself.

Nagel suggests a hierarchy in which the rules of reason dominate any other propositions and are dominated by none, especially not by those that can be made regarding the rules of reason themselves as just historical, psychological or sociological facts. Thus, “simple logical thoughts" are "exempt from skepticism," for "there is no intellectual position we can occupy from which it is possible to scrutinize those thoughts without presupposing them" (p. 64). Any skepticism about basic logic would then be contradictory, because it would necessarily rely on logic itself. Again, it would also follow from this dominance of the rules of reason "the impossibility of any sort of relativistic, anthropological, or 'pragmatist' interpretation" which would try to get outside of these rules, and thus "the last word, with respect to such beliefs, belongs to the content of the thought itself rather than to anything that can be said about it" (p. 64).

Nagel concludes that what he calls "subjectivism" about human reason would be self-defeating, because its proposals would amount to nothing more than an alternative description of the world, including us and any relations we may have with it. Attempts to naturalize epistemology "from the outside" with the aid of historical, biological, psychological, or sociological data would not elevate us to a neutral meta-level of analysis; they would be only more empirical and rational hypotheses in competition with other descriptions trying to explain human knowledge. Confronted with alternative descriptions of the world, we would have no alternative other than to ask ourselves which one is more probably true, and to that end we would have to employ reason again.

Can we Avoid a "Representation of Reality" in Science?

From the previous arguments Nagel also draws the conclusion that a naturalistic conception of knowledge cannot encompass and explain a rival realistic conception from an "outside" point of view; rather, they are competing views of the world, including humans and their relations with other parts of it. These or any other epistemological conceptions would aspire to be impersonal and universal: "Reflection about anything leads us inexorably to certain thoughts in which "I" plays no part— thoughts that are completely free of first-person content" (p. 67; emphasis in original). These impersonal thoughts or statements would be our various attempts to construct shared descriptions of reality, with the general aim of "make sense of the world in which we find ourselves and of how it appears to us and others" (p. 81). The comparison and choice among different descriptions of the world would presuppose that there is a reality independent from our observations and from the language that tries to describe it:

We begin from the idea that there is some way the world is, and this, I believe, is an idea to which there is no intelligible alternative and which cannot be subordinated to or derived from anything else. . . . What we cannot avoid is the idea that something is the case, even if we don't know what it is. Doubts about the reliability or objectivity of our perceptions and judgments have to be based on revisions of our view of the world; they cannot escape it completely. (p. 81)

Thus, the idea of an objective reality would be inescapable. For example, for Nagel the very possibility of regularities described by scientific laws suggest that order is a characteristic of most natural phenomena, revealed to us in spite of our partial and localized experiences. These regularities would be "manifestations of something else, something which includes us but on which none of us has a privileged perspective" (p. 84). Nagel calls this something else a "mind-independent natural order," the existence of which would be disputed by subjectivists, for whom any attempts at scientific descriptions would be only "general features of our perspective or linguistic practice or point of view" (p. 86). Nagel, however, objects that this is in itself an "alternative world picture," which is "in direct competition with the objective judgments it is meant to displace" (p. 86). The existence of these alternative world views would imply, for example, a competition between

the hypothesis that objects attract one another with a force directly proportional to the product of their masses and inversely proportional to the square of the distance between them, versus the hypothesis that it is a property of objects only as they appear to us (or in our language-game) that they attract each other with a force . . . and so on. (p. 86)

Nagel does not suppose as a necessary truth that all aspects of the world are orderly, but he submits that "anything we can know about must be at least related in an orderly way to us, and an amazing amount has proved to be within our reach; given our achievements so far, it is reasonable to try to continue" (p. 92). The hypothesis that the world is orderly would be always open to refinement—but according to Nagel, "however we divide up the contributions of the external world and of our own perspective, the result is a conception of how the world is, ourselves included" (p. 97).

Nagel extends this line of argument to any explanations we may have about knowledge, science, logic, and mathematics: they would be attempts to represent how these processes actually work, even if always revisable. What Nagel calls the subjectivist account of knowledge would then be just an alternative attempt at such representation. However, Nagel considers this account as self-defeating, because it would always bring us back to the idea of an objective reality (described by the account) and to the need to rationally assess the account as more or less true, considering the available data. Whatever our explanations about anything are, we would be always confronted with the question, "What reason do we have to think the world is really like that?" (p. 95).

A Radical Behaviorist Perspective on Nagel's Critiques

The Possibility and Relevance of a Naturalistic Account of Science

As we have seen, Nagel fiercely criticizes naturalistic attempts to understand scientific knowledge, including behavioral and evolutionary approaches. In his book, however, he seems to oscillate between the recognition that such approaches are feasible and the sheer denial of its possibility. For example, although at one point he states that it is "natural" to pursue such understanding (p. 72), subsequently he regards it as "impossible" and observes that maybe there is "something wrong" with such an attempt (p. 76).

Naturalistic approaches of knowledge are of course possible in the simple sense that they exist, in behavior analysis and beyond (e.g., Daniels, Cole, & Wertsch, 2007; Geber, 1977/Geber, 2006; Schlinger, 1995). Since Skinner (e.g., 1945/Skinner, 1972b, Skinner, 1957, Skinner, 1974), our perspective on human knowledge is an evolutionary one. Living organisms had to "know" the world since they became a part of it if they were to survive and reproduce. An evolutionary history of knowledge has to begin with this kind of nonverbal knowledge and track the complex selective contingencies that lead to current scientific knowledge.2 As Skinner (1957, 1945/1972b, 1974, 1953/2014) repeatedly pointed out, scientific knowledge developed as a function of its practical consequences, not because humans previously decided about the best rational means for the achievement of truth. A complex set of phylogenetic, ontogenetic, and cultural variables have led us to our present state of scientific knowledge, including all its various criteria for precision, confirmation, justification, verification, theorizing, and truth. Behavior-analytic approaches of logic and mathematics seize and improve upon Skinner's pioneering analyses, showing that they are part of the evolution of our ability to collectively deal with the world in increasingly complex and productive ways (Fantino & Stolarz-Fantino, 2013; Marr, 2015; Moore, 2015; Palmer, 2013; Terrell & Johnston, 1989). The evolution of such practices was greatly accelerated by the fact that we turned ourselves to the investigation of what we do when we think. It would be hard to imagine, for example, the systematization of Aristotelian logic and the subsequent developments in this field emerging in any other way. Thinking about thinking can improve it, with important practical consequences.

As Skinner (1957, Ch. 18) notes, the process of acquisition of a logic and scientific verbal repertoire is long and complex, for the individual as well as for scientific communities, and this emphasizes the social and behavioral nature of the process. Although Skinner (1974) regards the behavior of those who deal with logic, mathematics, and science as "the most difficult part of the field of human behavior and possibly the most subtle and complex phenomenon ever submitted to a logical, mathematical, or scientific analysis" (p. 234), he also states that there is simply no reason to prevent any attempts to understand these types of behavior from a behavior-analytic point of view. Skinner (1974) further observes that "traditional theories of knowledge run into trouble because they assume that one must think before behaving," and that the rational reconstruction of scientific behavior by scientific methodologists "seldom represents the behavior of the scientist at work" (p. 236). Such observations are especially apt considering Nagel's analysis of science as a purely rational enterprise.

Nagel argues that any "external" explanation of knowledge has to stop somewhere, because "that external view does not itself admit of a still more external view, and so on ad infinitum" (p. 20). The possibility that a behavioral epistemology may lead to an "infinite regress" (Zuriff, 1980), or an "infinite tail chasing" (Hayes, 1997, p. 43) is frequently recognized by behavior scientists (see also Barnes & Roche, 1997; Gifford & Hayes, 1999; Hayes, 1997; Hayes & Brownstein, 1986). The analysis of behavior is also behavior, subject to further analysis, and so on. Where Nagel finds his stopping point in reason, behavior scientists, following Skinner, tend to find it in the pragmatic goals of understanding, prediction, and control of natural and social processes. An evolutionary approach to knowledge would suggest that reason itself, whatever defined, has evolved as part of human efforts to solve everyday problems. As aptly summarized by Hackenberg (2013), Skinner's approach suggests that "scientific and logical behavior is continuous with other aspects of experience," and thus "does not take place in some logical-theoretical realm that transcends everyday life, but is instead part of everyday human contexts, messy and variable as they may be" (p. 279).

Nagel's claim that there is no neutral epistemic standpoint from which to analyze knowledge must be familiar to radical behaviorists, for it was repeatedly made by Skinner (e.g., Skinner, 1957, 1945/Skinner, 1972b, Skinner, 1974) himself. Behaviorists, Skinner (1974) remarked, cannot "step out of the causal stream and observe behavior from some special point of vantage, 'perched on the epicycle of Mercury'" (p. 234). In fact, the whole metaphor of an "external" or "outside" view is probably not adequate to deal with the issue. As Skinner's account suggests, humans simply evolved to the point at which they can analyze their own actions, and the implied circularity is no reason to prevent us from furthering our scientific self-knowledge:

The philosopher will call this circular. He will argue that we must adopt the rules of logic in order to make and interpret the experiments required in an empirical science of verbal behavior. But talking about talking is no more circular than thinking about thinking or knowing about knowing. Whether or not we are lifting ourselves by our own bootstraps, the simple fact is that we can make progress in a scientific analysis of verbal behavior. Eventually we shall be able to include, and perhaps to understand, our own verbal behavior as scientists. If it turns out that our final view of verbal behavior invalidates our scientific structure from the point of view of logic and truth-value, then so much the worse for logic, which will also have been embraced by our analysis. (Skinner, 1945/Skinner, 1972b, p. 379)

Thus we have that any epistemology is circular in this sense, as the very etymology of the word suggests: we aim to know what knowing is. The relevant point, as suggested by Skinner, is to "make progress": can we improve our understanding, prediction, and control of scientific behavior by means of our analyses?

However, in spite of the fact that we can analyze science and logic as behavioral processes, one can barely claim that we have reached the point of "invalidating" logic and truth-value, as Skinner rather boldly suggested could happen—and here Nagel has a point. Once we agree on a set of shared standards for the evaluation of scientific data (including logic and mathematics) the evaluation must proceed within these standards, and any attempt to "get outside" of them an evaluate them as "just" behavior would be pointless, albeit possible. We attribute generality to logic and mathematics in the sense that we expect to have the assent of other people whenever we present sound arguments or calculations supported by relevant data, and this is especially so in science—including, of course, behavior analysis. Although conceding that any language is a system of social conventions, with rules of correctness established by linguistic communities, Nagel draws a distinction between the content of thought and the meaning of words:

The fact that contingencies of use make "and" the English word for conjunction implies absolutely nothing about the status of the truth that p and q implies p. What is meant by a set of sentences is a matter of convention. What follows from a set of premises is not. (p. 40)

Here Nagel is only reasserting that the validity of a logical argument is a matter of logic, and there is no reason to quarrel with this. But again, this doesn't prevent a behavioral analysis of arguments and of logic in general, even if its goals would be different: in the first case, we want to know if a logical argument is valid (or a calculation correct, in the case of mathematics); in the second, we want to understand what behavioral processes lead a person or a verbal community to consider an argument valid or invalid (or a calculation correct or incorrect), and more broadly to emit logical and mathematical verbal responses at all. Nagel is justified in arguing that it would be irrational to abandon mathematics and logic "on historical or anthropological grounds alone" (p. 105), and behavior scientists have no reason to disagree. But the suggestion that we have to abandon an "external" view in order to preserve an "internal" one seems to have no grounding. The usefulness of logic and mathematics remains untouched by a behavioral analysis.

As another example of his point, Nagel states that there is a difference between a question like "How can human beings add?" and the question "How can electronic calculators add?," because "in ascribing that capacity to a person, I interpret what he does in terms of my own capacity," of which I cannot get outside, or about which I cannot have a neutral perspective. In face of that, how would one expect to "get outside" addition and explain it in anyone else? "To follow a rule is not to obey a natural law," concludes Nagel (p. 76). Attempts to explain addition in naturalistic terms would then be "impossible," or a "fallacy" (p. 42). But again, an analysis of mathematical behavior is no more circular than an analysis of, for example, philosophical behavior. Behavior analysis empirically demonstrates that the following of rules can be understood as a natural process. Why should one then refrain from applying this knowledge to understand how children can learn to add? Nothing seems to be taken from the validity and usefulness of addition by this procedure, and the fact that those who study addition must also know how to add doesn't seem to seriously preclude the analysis, for in general those who analyze verbal behavior must also behave verbally.

Part of the problem with Nagel's denial of an evolutionary account of reason seems to be that he regards language as a "vehicle" for reason, or as "displaying" principles of reason. In the author's words, "grammar obeys logic," and not the opposite (pp. 38–39). Moreover, to "have" a concept means more than to abide with the use of the concept by linguistic communities. For Nagel, the meaning of arithmetical concepts like "plus" and "two" is not based only in its use by communities, and although we may infer the possession of the concept from linguistic practice, it does not consist in this practice (p. 41). One is left to wonder, however, how to find the meaning of these concepts beyond their use in certain contexts, and how are we supposed to explain logic apart from verbal behavior. Nagel argues that "no 'language' in which modus ponens was not a valid inference or identity was not transitive could be used to express thoughts at all" (p. 39), but again, the problem is to suppose that language "expresses thoughts" that came before it. As suggested by Terrell and Johnston (1989), a rule of inference like modus ponens can only be understood as a product of behavioral contingencies. The rule is not a vehicle for a principle of reasoning, but a convenient way to control useful behavior with respect to the physical and social environment. Cruder forms of verbal behavior certainly came before logic as a set of rules, and this renders an evolutionary account of logic all the more necessary for its understanding. The fact that any newcomer to a scientific community will have to learn concepts and autoclitic rules with functions already established by the community doesn't imply that the evolution of the concepts and rules themselves must be left untouched by a scientific analysis. One wonders what would happen to etymology, a science to which Skinner frequently appealed in order to examine the historical function of words, if such an evolutionary approach was to be halted.

Palmer (2013) shows how the autoclitic frames provided by mathematical and logical rules "filter out" the variability of natural processes, generating "essentialistic models" that in turn control the scientific behavior of listeners in many useful ways. Truth and validity may thus be applied to these rules and to scientific model building in general, but we are dealing again with the selection of behavior by its consequences, not with essential principles of reason. In Nagel's terms, the difference between these approaches must be treated as one between alternative world views, but the implication that one of them better represents reality may itself be questioned by a selectionist approach. Even so-called "representations of reality" have useful functions under this form of explanation.

Nagel submits that mathematics and logic, given their universality, cannot be analyzed just as "language," but even if we allow for such universality this does not imply that a behavioral understanding isn't feasible. In fact, one would barely expect the development of mathematical and logical behavior in individuals not exposed to social contingencies arranged by verbal communities. Moreover, as Skinner (1988/Skinner, 1989) suggested, the universality of certain features of verbal behavior may be just a product of common environmental features. But even if we are uncertain about the relative participation of variables at different levels of selection in the explanation of mathematics and logic, the possibility of a behavioral treatment obviously remains open.

At some points in his book Nagel suggests a "conflict" (p. 87) or a "clash" (p. 139) between rationality and an evolutionary account of its development, but one wonders why this must be so. The author objects to the possibility of a behavioral explanation of rational procedures by stating that maybe "some things can't be explained because they have to enter into every explanation" (p. 76). Reason must then be "self-justifying" and "independently valid" (p. 136), in the sense that it must be considered a precondition of the acceptance of an evolutionary account, whereas the account itself cannot be used to justify or validate reason. But how is one to decide what is to be taken as a precondition of what? The fact that one could barely avoid scientific verbal practices when doing science should not prevent the functional investigation of its evolution. Nor should one refrain from proposing possible improvements in logical and mathematical verbal behavior just because one has to use them when doing so (after all, there would be no other way to do it). Logical and mathematical methods are useful rules that humans have developed to deal with the world in increasingly complex ways, and we have every reason to employ them to understand their own evolution.

Moreover, we need an evolutionary approach if we are to have any hope of understanding why humans—including scientists—do not always behave “rationally.” This is made clear, for example, by empirical research in the field of behavioral economics. The traditional “homo economicus” model, according to which humans are consistently expected to behave rationally to maximize their gains, in full awareness of the costs and benefits of their decisions, is now long outdated (Francisco, Madden, & Borrero, 2009; Furrebøe & Sandaker, 2017; Hursh & Roma, 2013; Reed, Niileksela, & Kaplan, 2013; Todorov, 2010). Most of our choices seem to be “irrational,” and we have to turn to selective contingencies to understand why. As Furrebøe and Sandaker (2017) suggest, we would be better off in dispensing altogether with the normative rational/irrational dichotomy, because it “carries the risk of substituting one mentalistic explanation for another” (p. 315) and may thus divert us from the analysis of behavioral contingencies.

Empirical Epistemology and the "Representation of Reality" in Science

As we have seen, one of Nagel's main arguments is that we cannot escape the notion that the aim of scientific language is to describe, in an impersonal and accurate way, an objective reality that is independent of our descriptions. The available data would then lead us to deem more likely that "to a considerable extent, the order that we find in our experience is the product of an order that is there independent of our minds" rather than "a framework we impose on experience" (p. 95). Science would then try to describe these orderly processes in the real world, although never in a final and absolute way. A competing subjectivist and relativist epistemological account, in trying to naturalize scientific knowledge, would amount to an alternative world view in direct competition with a rationalist and realist one. Such an account would, however, be self-defeating, because it could not avoid the idea of an objective reality.

Would all this apply to the "empirical epistemology" proposed by radical behaviorists? Selection by consequences constitutes a general framework from which behavior scientists talk about all behavior. It is an empirically based framework, and one that allow us to understand, predict, and control behavior with considerable success. As all scientists, behavior scientists will frequently talk and write in descriptive ways, which may be regarded as "representations" from a traditional point of view. Precise definitions (in the form of tacts and intraverbals) are an important part of empirical science exactly because they allow scientists to agree about the "truth" of their statements. As long as we have precise and shared definitions about our subject matters, we may agree about the number of planets in our galaxy, the types of organelles in a cell or the reinforcing effect of a consequence over a response class.

However, a behavioral analysis of the development of this kind of scientific verbal behavior (e.g., Normand, 2019; Skinner, 1957, Ch. 18) suggests that the only way to understand its emergence is through its practical, reinforcing consequences. A "precise" or "true" description of nature is not an end in itself; it is an important but intermediate step in order to coordinate our collective behavior in dealing with different aspects of the world. For most sciences the final test of empirical adequacy lies in their capacity to demonstrate experimental control over the processes they study—and, in some cases, to generate reliable technology derived from this control. As they get more sophisticated, however, sciences will frequently engage in the production of "true" statements before (or even without) achieving experimental and technological control. Scientists may then talk as "realists" because they have to coordinate their behavior towards the world in some way. This is a predictable consequence of the development of scientifically precise definitions. Logic and mathematics emerge as valuable tools in dealing with the world, but they are themselves only formalized descriptions of some of our relations with the world, not self-justifiable rational powers.

This is a rough synthesis of an epistemological "world view" supported by radical behaviorism and behavior analysis. In Nagel's terms, it can be treated as one of many possible descriptions of our epistemological interactions with the world, and is in fact in direct competition with other views, as it most certainly is with Nagel's. But if our epistemological approach doesn't have the benefit of a special "outside" view—if it's not "perched on the epicycle of Mercury"—why would it be preferable? Skinner (e.g., Skinner, 1957, 1945/Skinner, 1972b, Skinner, 1974) had repeatedly suggested that the answer lies in its practical consequences. This is, of course, the promise of an empirical epistemology: the functional analysis of the behavior of scientists would allow us to improve it, as it has allowed to improve behavior in many other settings. This improvement could well include our practices regarding ordinary uses of precision, objectivity, and truth. Against the charge of circularity posed by Nagel one can only say, with Skinner (1945/Skinner, 1972b), that progress is possible in such an analysis, and the very achievement of practical results would interrupt the infinite regress. The realist, on the other hand, has to face an old and ever-present caveat: he has no way to compare reality with what it states about it in order to judge the relative truth of his statements. He can only seek for agreement with its peers about the ordinary meanings of precision, confirmation, justification, verification, theorizing, truth, and so on, and in this he is side by side with any other scientist.

A radical behaviorist epistemological approach fully recognizes the historical, social, and cultural nature of human knowledge. To know is to behave, producing consequences in certain contexts—it is not to represent reality. However, in spite of holding a nonrepresentational epistemological position based on a behavioral account of knowledge, radical behaviorists feel free to go on talking "about the world" (or "reality," "natural phenomena," etc.) —as every person does, scientist or not. For example, we may assert many facts as truth: the world is not flat, the human species evolved from other species, some "news" are fake, and a host of other things, from trivial to complex, that we talk or write about with ordinary referential language. By the same token, we assert that some things are false, or we classify statements with different degrees of truth or certainty. It may seem, then, that even if we refuse to talk about "reality," we talk about it nonetheless (Barnes & Roche, 1997; Leigland, 2015).

The apparent contradiction is a product of the very process of dealing with the world through verbal behavior. Behavior scientists deal with behavioral processes with the general language of truth and facts simply because this is part of the verbal practices that were selected with respect to empirical evidence. But in doing so we are not talking about the world, or representing it; we are talking under the control of the world (including other people), and producing effects on it. This is a better picture of the scientific enterprise not because it's closer to truth, but because it permits us to identify and manipulate the variables that control the behavior of scientists, and to get rid of the impossible task of deciding when our descriptions better represent what the world really is. In the context of scientific verbal behavior a tact or intraverbal can be said to be more or less true, precise, accurate, or objective (Skinner, 1957, Ch. 5), but the criteria for these classifications are bound to their consequences in specific physical and social contexts.

But what about the regularities we observe and describe in natural phenomena? What about the fact that the results of our experimental procedures do not seem to depend on our verbal practices? Does this not suggest that, after all, our scientific verbal behavior must be in some way constrained by reality in order to be useful in dealing with it (as suggested by Marr, 2003; see also Laudan, 1990)? Science is a social enterprise, but the enterprise itself arose from our need to survive in a world that was here before humans came to be. The contingencies that shape scientific verbal behavior must include not only our social environment, but also the nonsocial, physical one. Presumably our descriptions of the world owe something to the world itself, not only to our social learning history. One may suppose that our descriptions of the world would be very different should the world itself be different than what it is.

Once again, however, our descriptions must be functionally explained. All verbal behavior, scientific or not, is part of contingencies that include physical and social relations—but the marks and signs that are the product of our verbal behavior can ever only affect the behavior of other people (or of ourselves as listeners), never the physical environment (Skinner, 1957; Guerin, 1997). This is the common function of all instances of verbal behavior, including our "descriptions of the world." To "describe," "specify," "refer," or "represent" are simply ways to produce consequences in certain contexts. As part of our ordinary language, these verbs are nonbehavior analytic ways to describe some functions of verbal behavior; a tact, as a verbal operant, never specifies, refers, or denotes—it only affects the behavior of other people in certain contexts. The nonsocial world is part of the contingencies that select different ways of talking or writing, but these contingencies are only put to effect by a verbal community. A solitary human on Earth (assuming he could survive) would have no need to behave verbally in order to change his environment. The emission of "representational" or "referential" words (tacts) is useful in many social contexts in which they are the standard way to produce certain consequences, but a radical behaviorist epistemology cannot employ "reference" as its basis for the explanation of knowledge, for it doesn't properly describe the relevant variables in action when someone "refers" someone to something.

In Nagel's terms this would be just another way of describing what happens when we "know" in competition with other possible descriptions. But it precludes an ultimate judgment of truth in terms of reference or representation. The explanation of knowledge in these terms is not false; it is simply not the best way to describe scientific verbal behavior if we are to identify its controlling variables. However, referential language is useful and may be more economical in much of ordinary scientific practice. The key to understand the apparent conflict between our epistemology and our ordinary use of referential language lies, therefore, in their usefulness in different contexts. Social agreement over our "descriptions of reality" is a necessary intermediate process so that we can coordinate our collective behavior in dealing with nature. Things and processes in the nonphysical world certainly take part in the control of scientific talk—but again, they remain unaffected by our words. The only function of verbal behavior is to alter the behavior of others. It allows us to coordinate our behavior in dealing with nature in many complex ways, thus achieving the impressive results displayed by science and technology.

If verbal behavior never describes, specifies, refers, or represents, what can radical behaviorists say about Nagel's argument that there is no escape from the idea of an objective reality, or "the idea that there is some way the world is" (p. 81)? Science is a verbal enterprise, and as soon as we begin to behave verbally, philosophical problems of this sort arise. But no matter how basic, universal, or "ontological" we deem our categories, their only general function is to affect the behavior of listeners in certain contexts (Skinner, 1957; Guerin, 1997). None of these categories can better represent what reality is. They can, however, affect listeners in many different ways, and that's their only relevance.

The fact that a word like "ontology" itself can be used with different meanings in philosophical discussions may be relevant. As Hayes, Barnes-Holmes, and Wilson (2012) point out, one may use the word either to refer "merely to explicit specification of conceptualizations" or to "the more traditional philosophical definition of what categories exist or can be said to exist in the world and the correspondence between analyses and these categories" (p. 4). Ontology in the first sense is, of course, necessary to any science. The second sense, however, is problematic in radical behaviorism: we can only create "categories" by interacting with the world and talking about it, but then our "ontology" (our conceptual divisions of the world) will be only a practical way of coordinating our collective behavior with respect to it. If one further assumes that ontological truth "by definition, is immutable, absolute, and final" (Barnes & Roche, 1997, p. 547), it's even harder to see how a behavior scientist could have any interest in discussing it.

As scientists we can barely avoid "talking about reality," however defined, and looking for some form of truth, precision, confirmation, justification, verification, and so on while investigating it. This should not contradict a behavior-analytic empirical epistemology, for in this approach any scientific verbal practice is analyzed in a functional way. As we suggested earlier, the mere use of representational language does not imply a subscription to realism. On the other hand, the recognition that verbal behavior does not represent reality does not imply that a behavior scientist cannot emit tacts or intraverbals that suggest the "representation" of ordinary language.3 Behavior scientists must feel free to talk "about reality" in any way that favors the achievement of their goals, without the burden of looking for the extent in which their words "correspond" to, for example, a Kantian noumenon. If the aim of science is to produce reinforcing consequences, Skinner (1974) is right in pointing out that the verbal behavior of the scientist is not closer to "what is really there" than any other behavior (p. 127). Thus, behavior scientists have no need to pursue scientific truth as verbal correspondence to an independent reality.

Finally, it must be noted that a naturalistic approach to scientific knowledge may also be criticized from the point of view of social constructionism: because science is a social process, there would be no way to distinguish scientific claims from any other kind of verbal practice, especially regarding its objectivity (e.g., Gergen, 1985, 1989; Howard, 1991).4 As Zuriff (1998) notes, however, it doesn’t follow from the rejection of a rationalist, representational epistemological approach that we must also reject any notion of objectivity in scientific practice. Scientific goals are of course human goals, and scientists may pursue them “while recognizing the role of human activity in the construction of descriptions of reality and criteria of objectivity” (p. 24). The fact that these criteria are socially constructed doesn’t make them arbitrary, as long as the goals they serve are clearly stated. It follows that these goals must be clearly stated, for only then we’ll be able to judge the adequacy of our objectivity criteria—and, more broadly, of all “precurrent activities (methods of observation, measurement, experimentation)” that are gradually selected in scientific practice (Zuriff, 1998, p. 19; see also Hayes, 1993). As we already suggested, these goals may be generally described as the understanding, prediction, and control of natural and social processes.

Conclusion

I have addressed Nagel's arguments regarding (1) the shortcomings of naturalistic explanations of scientific knowledge and (2) the impossibility of circumventing a realistic, representational epistemology.

Regarding (1), I argue that a naturalistic explanation of scientific knowledge is not only possible, but have important practical advantages, insofar as it allows the identification of the variables that control scientific behavior. The process of acquisition of a logic and scientific verbal repertoire is long and complex, and if we are to have any hope of improving it we must investigate it as a behavioral process. Nagel is right in arguing that there is no neutral or external viewpoint from which we can understand scientific knowledge, but this does not preclude progress in the understanding, prediction, and control of this process. There is no more circularity in this than in the behavior of analyzing behavior in any field. Logic and mathematics, as well as shared standards for the evaluation of scientific data (e.g., precision, confirmation, justification, verification, truth), are useful scientific tools. The fact that they may be analyzed as behavior does not preclude its ordinary use by behavior scientists. Verbal behavior is not a "vehicle" for reason, as Nagel suggests; it is reason in action. Instead of avoiding an evolutionary account of rationality, we have every reason to further pursue it.

Regarding (2), I argue that behavior scientists will frequently talk and write in descriptive ways, that may be regarded as "representations" from a traditional standpoint. However, a behavioral analysis of the development of this kind of scientific verbal behavior suggests that the only way to understand its emergence is through its practical consequences. A "precise" or "true" description of nature is not an end in itself; it is an important but intermediate step in order to coordinate our collective behavior in dealing with different aspects of the world. Scientists may then behave as "realists" because they have to coordinate their behavior towards the world in some way. However, verbal behavior never refers or represents in the traditional sense; it only affects the behavior of other people. The nonsocial world is part of the contingencies that select different ways of talking or writing, but these contingencies are only put to effect by a verbal community. The key to understand the apparent conflict between our epistemology and our ordinary use of referential language lies in their usefulness in different contexts. Behavior scientists must feel free to talk "about the world" in any way that favors the achievement of their goals, without the burden of looking for a fundamental "correspondence" between their words and other events.

At the end of his book Nagel states that mind or rationality may be "a fundamental feature of the universe" (p. 132) and advocates for "a more mind-friendly cosmology" (p. 133). In the terms favored by Nagel this suggests that we end up with two different pictures or views of the world, ourselves included: a mentalistic one, which supposes that humans have a mind or reason capable of reflecting reality, and a radical behaviorist one, which supposes that an evolutionary history of interaction with the environment lead humans to deal with it in a number of increasingly complex ways. Only the first picture actually aims to reflect reality. The second aims to increase our effectiveness in dealing with it. So, the metaphor of an epistemological "picture" or "view" may itself be misleading. Description of parts of the world and their relationships is an integral part of science, but it's not an end in itself. The ways we describe the world can only affect other listeners, allowing us to coordinate our interactions with it. Even if we insist in sorting out which among our "pictures" or "views" better reflect reality, in the end we'll still have to do this by comparing which ones better produce the goals we collectively pursue.

Compliance with ethical standards

Conflict of Interest

I declare that I have no conflict of interest.