RUDOLF CARNAP ON SEMANTICAL SYSTEMS AND
W.V.O. QUINE'S PRAGMATIST CRITIQUE
BOOK III - Page 7
Quine's Critique of Reductionism
Quine took Duhem’s philosophy of physical theory and made it a general philosophy of language, which implies the system-determined relativized nature of all semantics. Thus the third milestone in “Five Milestones” is the semantical shift from sentences to whole systems of sentences. This shift to a wholistic (or holistic) view of the semantics of language is a central characteristic of Quine’s philosophy, although it later went through some retrogression. The evolution of his thinking on this milestone is somewhat convoluted. He later came to think that his earlier and more radical pragmatism implies an unwanted cultural relativistic view of truth. Consequently in the 1970’s he attempted to restrict the extent of his semantical wholism, so that the semantics of theory is not viewed as contributing to the semantics of observation language, thus reverting to a positivist axiom.
His first statement of his wholistic thesis is what he later calls his metaphorical statement given in “Two Dogmas of Empiricism” (1951), one of his best known papers, reprinted in his Logical Point of View and often found in anthologies. The two dogmas he criticizes in this paper are the logical positivist theses of analyticity and reductionism. He defines the reductionist thesis as the belief that each meaningful sentence is equivalent to some logical construct based on terms referring to immediate experience. And he notes that Carnap was the first empiricist who was not content with merely asserting the reducibility of science to terms of immediate experience, but who actually took steps toward carrying out the reduction in the Aufbau.
Then Quine says that while Carnap later abandoned this radical reductionist effort, the dogma of reductionism continues in the idea that to each synthetic (i.e., empirical or nonanalytic) statement there is associated a unique range of possible sensory events, such that the occurrence of any of them would add to the likelihood of truth of the statement. Similarly for each synthetic statement there is associated another unique range of possible sensory events whose occurrence would detract from that likelihood. This dogma is implicit in the verificationist theory of meaning, and it survives in the thesis that each statement taken in isolation can admit of either confirmation or “infirmation”, which is to say, either verification or falsification.
The view of empiricism that Quine advocates as his alternative to reductionism is the thesis that statements about the external world face the tribunal of sense experience not individually, but only as a corporate body. Quine references Duhem in this context and his alternative view of empiricism has since come to be known as the “Duhem-Quine Thesis.” However, while Quine references Duhem in “Two Dogmas”, his wholistic view is more radical than Duhem’s, because Quine purges Duhem’s philosophy about physical theory of its positivism by ignoring Duhem’s two-tier semantics, which led to Duhem’s distinction between “practical facts” and “theoretical facts”. Quine’s treatment here of the difference between observation and theory is not a positivist semantical metatheory.
Furthermore, Quine’s radical wholism does not admit a distinctive semantical status even for pure mathematics and formal logic. Speaking metaphorically Quine says that the totality of our beliefs including mathematics and logic is a man-made fabric, which impinges on experience only along the edges. Then mixing metaphors he describes total science as a field of force whose boundary conditions are experience in which the laws of logic and mathematics are simply statements in the field that are more remote from experience. Any conflict with experience at the periphery occasions adjustments in the interior of the field, such that truth values must be redistributed over some statements, and a re-evaluation of some statements entails re-evaluation of others due to the logical connections among them.
The enabling feature of Quine’s wholistic doctrine of empiricism is his thesis that the total field is so empirically “underdetermined” by its boundary conditions, which are experience, that there is much latitude for choice as to what statements to re-evaluate in the light of any single contrary experience. And the criterion governing the choice of beliefs in the underdetermined system is entirely pragmatic, where the objective is a relatively simple conceptual scheme for predicting future experience in the light of past experience. The thesis of the empirical underdetermination of language can be traced to Duhem’s view of scientific theory. Duhem said that there could be many mathematically expressed physical theories, all equally empirically adequate, that explain the same phenomenon. But Quine furthermore extends Duhem’s thesis to include not just physical theory but all of language including contrary to positivists observation language. He maintains that no statement is immune from revision, and he notes that revision even of the law of the excluded middle has been proposed as a means of simplifying quantum physics. Quine notes that there is a natural tendency when making revisions to disturb one’s existing system of beliefs as little as possible, with the result that those statements that we are least likely to revise are those that have sharp empirical reference, while those that we are most likely to revise are those more theoretical statements that are relatively centrally located within the total network or web of belief. Later in his Philosophy of Logic (1970) this natural tendency becomes the “maxim of minimum mutilation”, an idea like James’ “minimum disturbance” in the latter’s Pragmatism (1907).
Quine’s most elaborate statement of his wholistic thesis is set forth in his first full-length book, Word and Object (1960). Instead of the metaphorical statement of his view in “Two Dogmas” a decade earlier, here he expresses his thesis in the literal vocabulary of behavioristic psychology. Much of the book is an exposition of his thesis of semantic indeterminacy as it is manifested in translation between languages, and thus appears as his indeterminacy of translation thesis. In the translation situation he portrays the field linguist in the same situation that Carnap postulates in “Meaning and Synonymy in Natural Language”, where Carnap attempted to describe how the field linguist can ascertain a term’s intension by identifying its extension from the observed behavior of native speakers of an unknown language. Carnap admitted that this determination of extension involves uncertainty and possible error due to vagueness, but he excused this uncertainty and risk of error because it occurs even in the concepts used in empirical science. While this admission of extensional vagueness in science made the fact unproblematic for Carnap, it had just the opposite significance for Quine. For Quine extensional vagueness is an inherent characteristic of language that he calls “referential inscrutability”, and which he later calls “ontological relativity.” And what Carnap called the intensional vagueness, Quine prefers to consider as a semantical indeterminacy in stimulus meaning but without admitting intensions.
Quine rejects Carnap’s thesis of intensions, explicates his own theory of meaning in terms of behavioristic psychology, and proposes his doctrine of “stimulus meaning.” Stimulus meaning is a disposition by the native speaker of a language to assent or dissent from a sentence in response to present stimuli, where the stimulus is not just a singular event but rather a “universal”, a repeatable event form. Stimulus meaning is the semantics of those sentences that Quine had earlier described metaphorically as positioned at the edge of the system of beliefs viewed as a force field, as opposed to the more theoretical sentences that are in the interior of the field. In Quine’s philosophy the idea of stimulus meaning is not a special semantics, but rather is an attempt to isolate the net empirical content of each of various single observation sentences without regard to the theory that contains them yet without loss of what the sentence owes to that containing theory. This attempt to isolate the semantics of observation language is a move away from his earlier critique of reductionism, where reductionism is understood as statements having a unique range of possible sensory events, such that the statements can be criticized in isolation. But at this stage Quine still retains his original empirical-underdetermination thesis, in which empirical underdetermination is integral to his wholistic thesis of semantical indeterminacy or vagueness.
The underdetermination thesis admitting multiple and alternative observation sentences for the same stimulus situation presents a question: how can the same stimuli yield alternative stimulus meanings? One of Quine’s answers is that the alternative theories or belief systems in which the stimulus situation is understood, supply different significant approximations. But there still remains the question of how stimulus meanings are to be construed as approximations. Quine has a theory of vagueness that he sets forth in the third and fourth chapters of Word and Object, which resembles the latter Wittgenstein’s thesis of paradigms, except that Quine explicitly invokes the behavioristic stimulus-response analysis of learning. On this analysis Quine rejects the view that stimulations eliciting a verbal response “red” are a well defined or neatly bounded class. He maintains that the stimulations are distributed about a central norm, which when a language is initially being learned, may be a very wide distribution. The penumbral objects of a vague term are the objects whose similarity to those for which verbal response has been socially rewarded in the learning experience is relatively slight. The learning process is an implicit induction on the part of the subject regarding society’s usage, and the penumbral cases are those words for which that induction is most inconclusive for want of evidence, because the evidence is not there to be gathered. And society’s members have had to accept similarly fuzzy edges when they were learning. There is an inevitability of vagueness on the part of terms learned by ostension, and it carries over to other terms defined by context on the basis of these ostensively learned terms.
Since Russell Hanson’s Patterns of Discovery (1958) the participation of theoretical concepts in the semantics of observation language is often expressed by saying that observation is “theory-laden”. And this semantical participation of theory in observation has made problematic the objectivity of observation, and therefore the decidability of scientific criticism. In 1968 in “Epistemology Naturalized” in Ontological Relativity Quine states that Kuhn and Hanson among others have tended to belittle the rôle of evidence in science and to accentuate cultural relativism, and that such philosophers represent a wave of epistemological nihilism. He notes Hanson maintains that observations vary from observer to observer according to the amount of knowledge that the observers bring with them. Thus one man’s observation is another man’s closed book or flight of fancy, with the result that observation as the impartial and objective source of evidence for science is bankrupt. At this stage of Quine’s thinking the semantical contribution of theory to observation is still problematic for him, but he continued to characterize observation language in terms of behavioristic theory of learning.
In the chapter titled “Observation” in his The Web of Belief (1970) Quine says that an observation sentence is a sentence that can be learned ostensively by the association of heard words with things simultaneously observed, an association which is conditioned and reinforced by social approval or successful communication, and which becomes habitual. And due to the social character of its learning, the observation sentence must be understandable by all competent speakers of the language who might be asked to assent to it. Thus according to Quine the sentence “That is a condenser” is not an observation sentence, even if experts agree to it. Quine maintains contrary to the positivists, that what qualifies a sentence as observational is not the lack of theoretical terms that may occur in theory formulations, but just that the sentence taken as an individual whole commands assent or dissent consistently, when the same global sensory stimulation is repeated. This behavioristic characterization initially enabled Quine to evade reference to semantics in his identification of observation language, and thereby to separate his view from that of the positivists, who defined observation language in naturalistic semantical terms. But in attempting to avoid a cultural relativist view of truth he thought he found in the likes of Hanson, Quine found himself getting back into the semantics of observation with the very positivist objective of keeping the semantics of observation uncontaminated by that of theory.
After Word and Object and Web of Belief Quine further developed the Duhem-Quine thesis in his “On Empirically Equivalent Systems of the World” in Erkenntnis (1975), the journal that had been made the official house organ of the Vienna Circle in 1930. This development of the Duhem-Quine thesis represents a further restriction on Quine’s earlier version on his wholistic semantical thesis of observation. Previously he had viewed empirical underdetermination as integral to semantical indeterminacy or vagueness in his semantical wholism. But in this paper he revises the concept of empirical underdetermination of language, and separates it from the wholistic view of the Duhem-Quine thesis. The scientific hypotheses that purport to describe things beyond the reach of observation are related to observation sentences by a kind of one-way implication, such that many alternative hypotheses may imply the same set of observation sentences, but not vice versa. Observation sentences do not uniquely imply just one theory purporting to explain the observable events.
It now is in this sense that natural science is “empirically underdetermined” by all possible events. Quine says that underdetermination lurks where there are two irreconcilable theory formulations each of which implies exactly the desired set of observation conditionals plus extraneous theoretical matter, and where no formulation affords a tighter fit. In Quine’s vocabulary the phrase “observation conditional” is an empirical generalization expressed in conditional form and implying an observation sentence describing an individual event. And his phrase “theory formulation” is a conjunction of the axioms of a deductive theory, which implies observation conditionals. This is a different sense of “empirical underdetermination” than what Quine meant in “Two Dogmas”, because it resurrects the idea of a semantically neutral observation language, which pragmatists such as Hanson, Kuhn and Feyerabend reject. When speaking of sentences implied by alternative theories these pragmatist philosophers find a phrase such as “same observation sentences” to be very problematic; they deny that different theories can have the same set of observations due to the contribution of the semantics of theory to the semantics of observation language.
Having revised “empirical underdetermination”, Quine then distinguishes his revised concept from the wholistic doctrine of the Duhem-Quine thesis. He reiterates that the wholistic doctrine says that scientific statements are not separately vulnerable to adverse observations, since it is only jointly as a theory that they imply their observable consequences, with the result that any one of the statements can be adhered to in the face of adverse observations by revising others. Then he states that wholism lends credence to the underdetermination thesis, because in the face of adverse observations we are free always to choose among various adequate modifications of our theory, and all possible observations are insufficient to determine theory uniquely.
Also in this work Quine considers several criticisms or “reservations” about the wholism of the Duhem-Quine thesis, and in his defenses he will pick and choose between underdetermination (revised) and wholism (unrevised). The first criticism is that some statements closely linked to observation are separately susceptible to tests of observation, while at the same time these statements do not stand free of theory because they share much of the vocabulary of the more remote theoretical statements. Quine answers that the Duhem thesis does not imply equal status for all statements. He says that the Duhem thesis applies even for observation statements, since scientists do occasionally revoke observation statements when these statements conflict with a well attested body of theory, and when the experiment yielding the observation cannot be replicated. This is such a weak concession to semantical wholism and the indeterminacy of observation, that it effectively limits wholistic theory participation in the semantics of observation language to the status of errors of observation.
A second reservation pertains to the breadth of the theory: If it is only jointly as a theory that scientific statements imply their observable consequences, then how inclusive must that theory be? Does the wholistic scope have to include the whole of science taken as a comprehensive theory of the whole world? Quine sees science as an integrated system of the world as science exists at any point in its historical development, but unlike the positivists he does not view it as integrated by reductionism into a single unified science. He says that Duhem wholism admits that science is neither discontinuous nor monolithic, but as “variously joined and loose in its joints in varying degrees”. Later in “Five Milestones” Quine elaborates on this idea by saying that all sciences interlock to some extent not only due to a common logic and mathematics, but also because small “chunks” may be ascribed their independent empirical meaning “nearly enough”, since some vagueness in meaning must be allowed for. This defense based on vagueness calls upon the semantical indeterminacy that enables wholism.
A third reservation is that the semantical and ontological wholism may imply a cultural relativistic view of truth. Quine denies that his wholism implies a cultural relativistic view of truth. His first argument is external to the wholistic thesis. He finds a paradox in the thesis of cultural relativism: if truth were culture bound, then the advocate of cultural relativism ought to see his own culture-bound truth as absolute. The cultural relativist cannot proclaim cultural relativism without rising above it, and he cannot rise above it without giving it up.
Quine then turns to the issue of irrationality of theory choice, the argument for cultural relativism that is internal to wholism. He argues that the choice between empirically equivalent alternative systems need not be irrational. He says he will settle for a “frank dualism”, and that oscillation between rival theories is standard scientific procedure, because it is thus that one explores and assesses alternative hypotheses. In this defense Quine switches between underdetermination and wholism. Rationality of theory choice is based on comparability of theories permitted by a neutral observation language, that is admitted by Quine’s revised underdetermination thesis, since it is only theories and not observations that are incompatible. The dualism is therefore merely one due to empirical equivalence. But the idea of empirical underdetermination as newly revised in this article is not the context in which the issue of irrationality of theory choice emerges. It emerges in the context of wholism where theory participates in the semantics of observation language. Quine then switches to the wholistic context, when he says that whatever we affirm, we affirm as a statement within our aggregate theory of nature as we now see it, and that there is no extratheoretic truth. Quine’s frank dualism has not been very frank in this defense. Quine’s revised concept of empirical underdetermination is not consistent with his semantical wholism. The revised concept of underdetermination permits a neutral observation language, while the Duhem-Quine wholism continues to permit theory to produce equivocation in observational description by resolving the vagueness in the semantics of observation language.
Quine eventually recognized this inconsistency. Just as he imposed logical one-way restrictions for his revised concept of empirical underdetermination, he found that he must impose semantical one-way restrictions in the semantical wholism of the Duhem-Quine thesis. In his “Empirical Content” (1981) in Theories and Things, which he notes contains “echoes” from “Empirically Equivalent Systems of the World”, Quine explicitly uses Hanson’s terminology saying that observation sentences are “theory-laden.” But Quine reconstrues the intended meaning of Hanson’s phrase to mean that the terms embedded in observation sentences may recur in theory formulations. Thus while Quine here says that observation sentences are theory-laden, he denies to the semantics of theory any participating rôle in the semantics of observation. In fact in Quine’s construing of “theory-laden” it is not observation language that is theory-laden, but rather theory that is observation-laden. At least he did not revert to the old Carnapian reduction sentences, to make theory observation-laden.
Still later in “Truth” in his Quiddities (1988) he is explicitly reconciled about refusing to admit theory any resolving function in the semantics of observation. There he says that we work out the neatest world system, and we tighten the squeeze by multiplying the observations. Tightening the squeeze in observation sentences is the progressive reduction of vagueness but only by the addition of information in additional observation sentences. Quine’s limitation on which contexts may resolve vagueness and which ones may not, is arbitrary and ad hoc. His wish to make observation sentences semantically uncontaminated by theory is a positivist atavism, even though his motivation is not characteristically positivist. His point of departure was not a preconceived semantics for observation; he attempted a behavioral (i.e., behavioristic) characterization of observation language instead. Still, he believed that an unrestricted wholistic, theory-dependent, context-determined semantics encompassing both theory and observation language implies a relativistic and subjectivist philosophy of truth. Fear of a relativistic view of truth led him to revise his original statement, the Duhem-Quine thesis.
Quine the logician always saw theory language as an axiomatic system with observation language serving as its derived theorems. Unlike for Hanson, Isaac Newton’s mechanics is for Quine still “theory” today. On the pragmatist concept of scientific theory, however, theory language is identified not by contrast to an observation semantics or by semantics at all, but by reference to its function or pragmatics in basic research: it is discourse that is proposed for testing in contrast to that which is presumed for testing. Thus, observation language need not be exclusively identified as either theory or nontheory language (unless the pragmatist simply chooses to define “observation” correlatively to his functional definition of “theory”). And all contexts consisting of explicitly or implicitly universally quantified sentences believed to be true operate to resolve the vagueness in the meanings of their common univocal terms. Quine’s view is not a pragmatist view of theory based on the function of theory in empirical basic science, but is better characterized as an archival concept of theory, or what Hanson called an “almanac” view. Correspondingly his concept of observation language is an archival concept of observation language. Quine believed that this archival view would enable him to make observation language a repository of permanent truth. And his motive is his wish to evade the relativistic view of truth, which he believed is implied by the unrestricted context determination of semantics.
More recently a member of Quine’s intellectual entourage, Donald Davidson, has attempted to evade semantical relativism with a turn to instrumentalism. Davidson’s principal statement of his thesis is set forth in his “The Very Idea of a Conceptual Scheme” (1974) and “Belief and the Basis of Meaning” (1974) reprinted in his Inquiries into Truth and Interpretation (1984), a book he dedicates to Quine with an inscription “without whom not.” He rejects the representationalist view of the semantics of language, which he considers a third dogma of empiricism after the first two referenced by Quine in the latter’s 1952 “Two Dogmas” article. Like Dewey’s rejection of the dualism of “experience” and “nature” Davidson rejects the dualism of “scheme” and “world”, of “conceptual scheme” associated with language and “empirical content”, of “organizing system and something waiting to be organized”, dualisms that he finds in the views of Whorf, Kuhn, and Feyerabend. In this manner he remains more faithful to Quine’s original behaviorism than Quine did. Thus the decision necessary for interpreting another’s discourse is to maximize one’s shared beliefs, such that there can be no basis for concluding that others have beliefs radically different from one’s own.
Davidson concludes that in giving up the dualism of scheme and world, we do not give up the world, but rather re-establish the “unmediated touch” with the familiar objects that make our sentences and opinions true or false. Thus Davidson argues that there is no conceptual relativism, because there are no representational conceptual schemes to be relativistic. But Davidson’s conclusion is a non sequitur. The knower can be a spectator of his ideas, but this inspection is a reflection ex post facto upon his firstly already having the inspected knowledge of the real world. Apart from this secondary reflective knowledge, the spectator thesis about knowledge of the real world is readily rejected, when we realize that what we know firstly is not our ideas, but the real world, and most notably that our knowledge is thus constituted by our ideas rather than the ideas being an object of knowledge.
Both Quine and Davidson are motivated to evade semantical relativism, because both mistakenly believe that a relativistic, context-determined, semantics implies a relativistic thesis of truth. Regardless of how culture-bound and context-determined may be the semantics of a language, it is not possible capriciously either to affirm or to deny truthfully just anything expressed by sentences made with those concepts. The empirical underdetermination of language implies that many alternative observation-reporting sentences can be said which are consistent with the same sense stimuli. Still, the empirical constraint imposed exogenously on sentences by the recalcitrant real world – even when not yet interpreted – forbids just any arbitrary distribution of truth-values over a set of logically related, semantically interpreted grammatical sentences. When any subset of these sentences is given definitional force to specify its semantics, only some of the remainder sentences containing the same descriptive terms can also be true.
Truth is always relative to what is said, but the real world in which all language users live forbids ingenuously asserting just any old thing in the semantically interpreted language. Therefore, semantical relativity does not imply relativism of truth, but just the opposite: with a metatheory of semantical description exhibiting the compositional nature of meanings, semantical relativity explains the partial equivocation that makes it impossible for the same sentences occurring in two different belief systems, to be completely true in one belief system and completely false in an alternative system. It explains how the same sentence is not simply and completely the same statement in each system, but is partially the same in each, and to that extent true in both systems. And for the same reason it also explains why the semantics of observation language need not be quarantined from the semantics of theory, in order to assert the objectivity of truth. Observation statements, which pragmatically defined are merely singular test-design statements, may be common to pragmatically defined contrary theories, such that belief in the test-design statements makes the test outcome contingent and not willfully validating, and makes a falsifying test outcome of one of the theories an objective truth.
Each person acquires the semantics of what Quine calls observation sentences from his own personal experiences, and he acquires it publicly and ostensively in the circumstances of his language-learning situation in his personal history. There is a wide variation among people between what is learned ostensively and contextually, but even for those simple statements learned ostensively by most people, intersubjectivity is increased with successive approximation, as the web of belief grows and imposes increasingly more shared truth conditions on the ostensively acquired semantics.
The entire web of beliefs may be viewed on analogy with an underdetermined system of conditional equations, in which the addition of new equations further restricts the range of numeric values that the set of variables may accept, until the system becomes uniquely determinate with a unique solution set. One difference between the mathematical system and the language system is that with just a sufficient number of restrictions the equation system may admit to only one solution set, whereas language is never restricted to a unique interpretation; it is always empirically underdetermined. Another noteworthy departure from the mathematical analogy is that the mathematical variables can take only one numeric value at a time without becoming ambiguous, while each of the univocal descriptive terms, including those used as measurement variables in applied mathematical theory in empirical science, simultaneously accumulate semantic values distinguishable in the explicitly related universal statements in the system of beliefs, subject only to the preservation of univocity. Thus all the terms explicitly related by the sentences in the web of belief may participate in one another’s univocal semantics, and thereby add to the resolving of one another’s vagueness, i.e., one another’s empirical underdetermination. Furthermore as implicit statements are made explicit by deduction, the vagueness in the meanings of the terms of the system is even further resolved.
But Quine viewed meanings as abstract or mental “entities”, and then developed his behavioristic theory of stimulus meanings, which he called “behavioral dispositions” to evade the representational function of language. He could not be expected to have developed a metatheory of semantical description enabling him to describe how meanings participate in one another. The closest Quine came to the idea of semantical participation was the idea of the resolution of vagueness. His rejection of the dichotomous analytic-synthetic distinction is a worthy start toward such a metatheory, but his rejection of the distinction was actually a rejection of analyticity as such, except in the cases that he called “analytical hypotheses” used for translations. As it happens, rejection of the analytic-synthetic dichotomy does not imply the rejection of analyticity as such. Universally quantified statements believed to be true for empirical reasons may also be used analytically to exhibit the complexity in the meanings of their constituent terms by displaying their component semantic values that constitute the discriminating capability in the descriptive function of the language. In other words all universal empirical statements in the web of belief are analytical hypotheses. And theories are those that are viewed as relatively more hypothetical than other empirical statements including notably those used in a test of the theory.