The Luv Boat
or Unelected Affinities between Logical Positivists and Deconstructionists
II. Logical Positivism Revisited
La haine d’un homme médiocre est toujours une haine immense.
I don’t have a particularly good reason for focusing on logical positivism at this moment in the history of philosophy. My comments were occasioned by picking up and leafing through Rudolf Carnap’s article, The Unity of Science. I was bemused by both the conceptual confusion and the overwhelming vulgarity of the piece. Whereas the Tractatus reads like the poem of Parmenides, Carnap’s essay recalls nothing less than the Communist Manifesto. It goes so far as to employ the kind of groupspeak so beloved of totalitarian movements in the interwar period of the last century. The English translation occasionally inspires a kind of nostalgic irrelevance as when Max Black contends that problems in the Tractatus were solved by the concepts of verification and protocol sentences. I suppose, like your average slimebag politician, the philosophical agitator has to compensate for the weakness of what he says with the fiat like solemnity with which he says it.
But why even bother? The position in this and other rabble rousing articles from the Vienna days of logical positivism has been thoroughly repudiated even by the most hardcore adherents of “hardnosed” logicist philosophy. In fact it was partly repudiated by Carnap himself. He changed his views frequently on a number of issues and not even all members of the Vienna Circle, the early Moritz Schlick in particular, conformed to Carnap’s ultimate program on all points. Well, even within academic philosophy much of the repudiation has been in the details, such as, for example, the theory of protocol sentences. The spirit of the original program remains intact. Moreover the logical positivist viewpoint has seeped into the subconscious of working scientists. The logical positivist dictum of staying close to the home of immediate observations and what can by simple rules be deduced from simple observations is what empirical scientists, particularly in their most defensive postures, believe they adhere to. What goes unsaid is that logical positivism leaves no room for science at its most creative, least Baconian and most glorious, namely the recognition of laws and the construction of theories or models (I understand the point of replacing the word “theory” with the word “model” for the empirical sciences, but I am cautious about the possible confusion with models for logical systems and mathematical models). Failure to distinguish between law and theory (or model) accounts for the failure to capture the intuitive, insightful quality of science - or the conceptual reordering involved in scientific revolutions.
Is Carnap M. Homais? He sure sounds like him. As a philosopher Carnap cornered the job of the provincial apothecary. And, while I don’t think M. Homais is entirely bad or worth dismissing, the comparison might be kept in mind. There are what some may consider bigger fish to fry from the foundationalist pond including the detailed reductions in Russell and Whitehead’s Principia and the program of epistemological foundations in Husserlian phenomenology. Since Cartesian foundationalism relies crucially on the goodness of god, it is to be regarded more as a museum exhibit than a serious contender for philosophical evaluation. There are also specular metaphysical theories like those espoused in Wittgenstein’s Tractatus and Davidson’s writings. Perhaps I shall turn to those some day. For the nonce it is the very vulgarity of logical positivism and its easy absorption as sound bites into the popular imagination that impels settling the question of its validity once and for all.
Part of logical positivism is dead. Part of it has turned into a collection of idées reçues. As far as philosophy is concerned, the logical positivist prohibition on productive speculation about the nature of scientific theories remains very much in force. The unified science program of the Vienna Circle determined the parameters within which much academic philosophy continues to operate despite the ultimate rejection of most of its basic tenets and procedures. (In our day physicists and cosmologists have taken up the speculative challenge with equal portions of admirable verve and philosophical naiveté. The elegant speculations of Stephen Hawking have been more than counterbalanced by innumerable Zen interpretations of quantum mechanics at the hands of respected physicists. (Not that there isn’t an important and unmined issue in understanding Buddhist and Taoist ontologies. It’s just that simply pointing out apparent parallels is grossly superficial.)) Indeed the limitations of the unified science outlook may also be largely responsible for some of the perceived current problems in devising a satisfactory general philosophical theory on the order of the grand old systems – that is a philosophy not limited to the “foundations” of science - and the relations of academic philosophy to the wider culture. So at the risk of being accused of spitting on a corpse I thought I might revisit the founding documents of so-called logical positivism (a name chosen by later editors) and review some of the problems with the program espoused.
I suspect there’s nothing I have to say that hasn’t been said before. But at the present juncture, where academic philosophers tend to get lost in the weeds, there may be some value in recalling the historical roots and original purpose of modern philosophy. Some of the following comments about metaphysics in particular appear most notably in Heidegger but I don’t think he was particularly clear about it nor did he provide reasons to back up his assertions (Cf. e.g. “Überwindung der Metaphysik” in Vorträge und Aufsätze pp. 67-95 where Heidegger seemed to be more bothered by cybernetics than logical positivism). In addition, some of what Derrida has to say about Husserl parallels my comments on Carnap’s 'pataphysics. Does this make Heidegger le curé Bournisien? Quite possibly. I’m sure the very idea of having Heidegger and Carnap sit together at a communal philosophical table has set them both spinning in their graves – in opposite directions of course.
So let’s jump into our cigarette boat and skim around Salamis.
1. When Saying That You’re Not Doing Something Means That You’re Doing It. Carnap’s gripes have nothing to do with metaphysics at all – i.e. with speculation about the nature of entities outside our world and their causal effect on events in our world. Unfortunately it is specifically metaphysical dogmas that caused much of the trouble in the world when they became part of political and social programs; and it is metaphysical dogmas that were the target of Enlightenment philosophers who in many ways serve as the model for the logical positivists. Rather, the issues he specifically identifies as metaphysical are really 'pataphysical in nature. One such issue was whether the science of biology requires the concept of a vital force for a satisfactory theory of living entities. Carnap probably wins on that one although it remains an open question whether the concept of evolution and its relation to animal behavior has as yet been satisfactorily reduced to chemical laws. Another is essentialism. A third is whether lexical definitions of supposed essences are really informative. These two ideas are related but not entirely the same since a version of essentialism can hold that at least some essences are ineluctable. Carnap’s alternative, which addresses only the definition and not the essentialist issue, is to replace lexical definitions with what today are called operational definitions – a useful but limited solution which I will discuss in more detail below. Moreover, essentialism both with respect to particulars and kinds is still, rightly or wrongly, very much a part of the philosophical landscape. One has only to think of Geach’s defense of Aristotelian essentialism and the important role essentialism plays in Wiggins and the interpretations of modal logic proposed by Kripke and others. Another aspect of ‘pataphysics that Carnap does not mention most likely because he doesn’t notice it concerns generalities about things in this world. The result is that he makes broad ontological assumptions that are not obvious, not justified and probably not necessary for the successful pursuit of scientific research.
Carnap (p. 22) makes a blanket statement that he and his colleagues intend to restrict their activities to logical analysis – presumably of the observation reports and rules of inference that compose the thesaurus of the sciences. If that means they are not practicing what we others consider to be philosophy, so be it. Specifically excluded from their pursuit are opinions about doctrines such as monism or dualism, spiritualism or materialism (although he walks this back later in his essay: p.93. For such a hardnosed philosopher Carnap is not a marvel of clarity. He flirts with rejecting materialism and later proposes physicalism, or at least physicalist language as the intersubjective solution for all the world’s ills. What is the difference between materialism and physicalism? None as far as I can see. So the even minded umpire who rejected both phenomenalism or spiritualism and materialism was just practicing a feint before awarding the ball to the physicalist/materialist side. (Carnap was to use a phenomenalistic framework in various of his later incarnations. Goodman (pp. 101-107) rehearses some of the objections to the claim of the primacy of physicalism, particularly in Carnap’s version that includes observation statements. Goodman’s claim that his own system can be interpreted either phenomenalistically or in physicalist terms, i.e. that the logical structure of physical objects is identical to that of mental phenomena, is never demonstrated. But either it is the case that one of the two interpretations is defined in terms of the other, in which case the principle that their structures are identical is a vacuously analytic truth, or that the two interpretations entail different consequences, in which case their identity definitely needs defense. Goodman also deals with the claim which motivates much phenomenalism to the effect that knowledge of phenomena is certain knowledge and that it can serve as a basis for knowledge of physical objects. But he does not address the other motive behind the notion of mental ideas (or sense data or Vorstellungen), at least among those of a Kantian persuasion, to the effect that mental ideas are described differently from and do not enter into an ontological one-to-one correspondence with the physical objects they “represent,” so to speak, and that this difference, furthermore, allows philosophers to “logically” construct physical objects (or their descriptions) from the mental images (or the descriptions of those images) that represent them. Sellars addresses this issue (Ch. 9), but his solution comes at the expense of accepting a somewhat ineffable (What qualia/essences are there besides greem and kleem?) version of Aristotelian essentialism for his basic particulars.)
There is another locus of confusion. Carnap distinguishes between a material and a formal mode of speech, on the one hand. Material speech is supposed to be about things and formal speech is supposed to be about statements. He also talks about physical language as the ultimate reductive language of science, on the other. He never lays out what the domain of reference of physical language is supposed to be though he implies strongly that it is the world of objects and events that physicists talk about. And of course that world is what we are left with after a materialist reduction. Presumably physical language can come in both a material and a formal mode. By way of aside, as if anyone cares about my opinions (as opposed to my arguments), I thoroughly agree with a reduction of our ontology to the physical or material world or, more cautiously, at least away from the ideal worlds we have come to know from the history of philosophy. I don’t believe in gods or angels or demons and I don’t believe that Lockean or Husserlian ideas are not reducible to physical counterparts, viz. an amalgam of neurophysiological events and the objects those events refer to. That said, I do believe that a strong, indeed overwhelming case can be made for the impossibility of a methodological reduction. Many sciences require concepts that cannot be fully translated into the language of physics. My argument is against Carnap’s claim or assumption that he somehow proved (via intersubjectivity) the ontological and methodological primacy of physicalism. He did nothing of the sort. (Black translates “methodical materialism” or “methodical positivism” when he presumably means “methodological materialism” and “methodological positivism”; I will substitute the latter terms for the former throughout)), idealism, positivism (a name which he also sees fit to adopt later on (p. 27)), epistemology and ethics including utilitarianism and intuitionism. Pretty much the academic equivalent of the St. Bartholomew’s Day Massacre. It is not specific views that Carnap rejects but the questions those views are meant to answer since those questions, by his criteria of meaning, are meaningless. The problem is that in the course of the “logical analysis” pursued by Carnap and the observation reports and theories constructed within the empirical sciences, ideas are assumed and truths are presupposed whose evaluation and consideration has not and perhaps cannot lie within the powers of either logical analysis or the empirical or even mathematical sciences. The –isms Carnap rejects (and all those he mentions specifically) are meaningless by positivist criteria of meaning largely not because they deal with transcendent entities, which were the valid target of 18th century skepticism, but because of their generality. You cannot verify or falsify or logically parse a doctrine that there is no external world and everything is just a figment of my imagination. But Carnap misses the methodological error of solipsism. He thinks it boils down to non-verifiability or, as he argues later in the text, the possibility of constructing a non-solipsist “intersubjective” language, whereas what makes doctrines like solipsism “metaphysical” is an inappropriate use of the word “all” and the concept that “all” is purportedly meant to signify. You can’t verify or bring evidence against solipsism because all things are in my mind. Any evidence to the contrary is simply in my mind as well; likewise all evidence in support of solipsism is not in my mind but in the physical world. Because he is in error about what constitutes the fault of many metaphysical –isms, Carnap commits exactly the same fault. He uses at least implicitly the term “all” to characterize the meaningfully real world: All things are given, all things can be represented in a logically perspicuous language, all things can be observed or be reduced to what is observed. Many of these assumptions have been dealt with by Carnap’s critics and successors. But logical positivism and many of Carnap’s critics make equally fatal ‘pataphysical assumptions that have not received the attention they deserve. Primary among these is the status of the objects the sciences purport to deal with. Specifically in the case of the social sciences a case can be made that their subject matter is a construct that simply appears along with the primary observations and basic laws and theories of the science. But that fact does not eo ipso ruin their scientific status. In a broader sense suppositions are made about the real existence and distinct specifiability of the broad genus particulars such as constitute the members of sets in the foundations of mathematics or such that groups of which are the primary referents of the individual sciences. There is nothing in the program Carnap lays out for dealing with issues regarding either specific ontologies or our ontological assumptions in general. It is just possible that one could reject these issues as meaningless, but there is nothing in Carnap’s criteria of meaning that specifically excludes ontology (Strawson would later attempt a proof that the notion of a particular is a necessary category of thought). Particulars (as well as universals if some readings of Frege and Gödel are to be believed) are experienced. One ontological question is to what degree the experience of particulars is subject to variation or conditioning. Carnap does not mention this issue in his list of metaphysical –isms. Indeed it is significant that it is not amenable to treatment in the formal mode only, which is the technique he proposes for dealing with other broad issues regarding the foundations of science, for the supposition of specifiable particulars is built into the grammar of the formal languages Carnap takes to be basic and Carnap never questions whether those formal languages may or may not synchronize with “the given” or whether they may be considered as artifacts “out there” in the world along with the objects their signs refer to. He never questions whether they provide the only possible access to the world or, more modestly, a privileged access to the world or rather an obfuscating veil that distorts our access to the world.
Carnap could take the position that he need not deal with his own presuppositions. They are working presuppositions and as long as they function successfully to clarify (not justify) the theorems of the sciences, then his goals are met. This is almost always the position taken by practitioners of the individual sciences and you can find echoes of Carnap (and Comte and Mach and Russell) in the introductory chapters of many college textbooks. But this effectively obscurantist drawing of the line can also be taken to justify theosophy, transcendental meditation, Mormonism, fundamentalism, Gaia theology and any other nutty pseudo-intellectual movement out there in the land of the semi educated. Carnap’s self-imposed inability to go beyond sloganeering and party politics deprives him of the ability to engage with forms of life and sets of beliefs whose presuppositions are different.
The particular ineffectiveness of the Carnap refusal is on display with regard to ethics. Carnap rejects what he calls regulative or normative ethics in favor of a scientific study of what our ethical beliefs in fact are (p. 24). The philosophical study of the “essence” of ethical maxims (which at a certain level of abstraction is what the leading ethical theories, Kantianism and utilitarianism, are about) seems to fall into prohibited territory as far as Carnap is concerned. And, of course, individual ethical canons are not evaluated; they are simply studied. This deprives us of the ability to engage, employing the rigorous kind of argument which is practically synonymous with philosophy, with those ethical canons we may find objectionable and whose defeat is particularly pressing when they begin to gain ground in the popular mind. Logical positivism proposes to simply study Marxism, Fascism, Xtianity, Mohammedanism and all the other dubious ethical canons. It destroys the very philosophical tools to question their validity. One might ask is the abdication of ethical engagement by philosophy departments at least partly to blame for the current ascendancy of the Religious Right in supposed civilized nations like our own?
Carnap’s unified science program is in direct conflict with the idea of distinct scientific domains which is at the heart of Dilthey’s philosophy and for which Heidegger tried to provide an ontological foundation in Sein und Zeit. There is, however, a difference between methodological reduction and ontological reduction. There is reason to believe that Dilthey’s view focused on the former let the ontological chips fall where they may. Even if he didn’t do so specifically there is room for that position. Unbridgeable gaps between the methodology of history, for example, and the methodology of physics does not mean that there are objects numerically in excess of those defined by physical theory. It would imply the need for alternative non-reducible observations and theories.
2. When You Get Caught in Your Own Net. This sentence encapsulates one of the most obvious faults of the presentation of the logical positivist program: “…our position...requires that every scientific statement should be based on and reducible to statements of empirical observations. “ (p. 27. The word “statement” is vague in context. Carnap or Black is not quite clear as to where on the spectrum from propositions through sentences, sentential strings and sentence tokens to utterances, statements may lie. I assume the original German is Satz in which case it is most likely Carnap means “proposition” in the sense of what is meant by sentential strings in different languages that mean the same thing.) As phrased, Carnap’s sentence satisfies its own criteria (Cf. also p. 75 for another instance of this claim). It is verifiable. We just need to check the corpus of writings of the logical positivists to verify that their position indeed requires that every scientific statement should be based on and reducible to statements of empirical observations. The “should be” in the subordinate clause is probably a literal translation of the German subjunctive. But unless Carnap believes that he and his buddies are just a bunch of cranks, his sentence effectively presupposes this proposition:
(a) “Every scientific statement is based on and reducible to statements of empirical observation.”
But (a) is not an empirical observation or based on an empirical observation. It is only “true” if you make it true, that is in case you use it as a criterion to select statements you regard as scientific. In that case it is not an empirical observation but a stipulative indeed lexical definition. And Carnap doesn’t see any place for lexical definitions in science. The conclusion is that (a) is not verifiable. There is no empirical observation in light of which it is true; the only thing that makes it true is definition by stipulation. If (a) isn’t scientific what is it? Among those statements that make positive assertions about the world Carnap allows only two possibilities, scientific statements and nonsense or metaphysics. In fact (a) shares other properties with the putative assertions of doctrines such as subjective idealism, for example, which Carnap denounces as metaphysical. That is, it asserts something about all statements and, by extension, about the whole world. The subjective idealist in effect defines the whole world as being in the mind, so nothing you can discover in the world will falsify his statement. He will continue to assert that the falsifying fact you discovered is in the mind also. Likewise the logical positivist defines the whole world or everything in it as expressible by and only by empirically verifiable statements. Nothing you can find in the world will falsify his statement since he will continue to assert that whatever you find is either reflected in a verifiable statement or that your report of your discovery is nonsense. The embarrassing conclusion is that (a) does not live up to its own criteria of meaning and therefore qualifies as nonsense. It is easy to see that this applies to any assertions defining the logical positivist criterion of meaning.
By way of aside, Carnap cites (p. 26) Wittgenstein as a source of his claim that only verifiable statements are scientific. Verification, of course, was a notion totally foreign to Wittgenstein. There is greater kinship in Carnap to Russell and his notion of acquaintance.
Another sentence (p. 28) exhibits a stronger version of the same problem: “Logical analysis shows that every statement is either empirically verifiable (i.e. on the basis of protocol statements), analytic or self contradictory.” Assuming logical analyses are always valid this sentence entails:
(b) “Every statement is either empirically verifiable (i.e. on the basis of protocol statements), analytic, or self-contradictory.”
The problem is (b) is neither empirically verifiable nor logically analytic or self-contradictory. It cannot be verified because it is the token of a universal proposition. We cannot verify its truth status by examining each of the infinite (or very large finite) number of statements and possible statements. Moreover, there is no obvious recursive method for such a verification based on some manageable finite number of core verifiable statements. The only way it could be analytic is if it were like (a) a stipulative lexical definition. But accepting it as a definition would be a choice for which we could reasonably require reasons outside of what Carnap calls “logical analysis.” We could as easily say: “Popish analysis shows that all statements are either ordained by god or obviously false and the work of the devil.” Incidentally it is not clear which logical analysis Carnap is referring to. (b) is not a theorem of the Principia nor of any other first order logic I am aware of. Carnap’s original phrasing and the fact that it is a semantic claim tell us that (b) is a second order or metalinguistic proposition. But there is no concept of verification in any of the standard metatheories of first order logic. (There may be some gold in intuitionistic logic, but I’m not going there.) These are concerned solely with the formal structure of the truth or falsity of first order propositions and molecular propositions in particular, not the circumstances under which a first order atomic proposition is true or not or may or may not be truly asserted. The closest thing I could find to (b) is a comment let drop by Russell to the effect that anyone dealing in scientific matters needs “a robust sense of reality” (Introduction to Mathematical Philosophy, p.170. Cf. also p. 169). I am sorry to report that (b) is exactly the sort of metaphysical statement that Carnap deplores. And the fact that it is neither empirically verifiable nor analytic or self-contradictory casts doubt on his theory of meaning. It expresses a threshold semantic paradox. If it’s true it’s false, but it could simply be false (or meaningless) without entailing its own truth.
By way of aside I believe that the felt need on the part of the Logical positivists for a concept like verification is symptomatic of the fact that more than a purely formal semantics of first order formal logical systems is required to tie logic to “the world.” However, a metalogical semantics of the Tarski sort is itself a purely formal system that mostly provides an alternative structure for propositions containing logical connectives (what Russell called molecular propositions) and other syncategorematic terms like the quantifiers on the basis of their truth value instead of their deducibility from axioms by way of rules of inference. Tarski’s argument for Convention T stands outside of a formal semantics. It is a minimal condition for the validity of the theorems of a formal semantics that assert the truth of an atomic (or, by extension molecular) first order proposition. Do we need anything more than the minimal? For a definition of the conditions under which the truth of propositions containing contextually defined logical terms is derived from the truth of their embedded atomic propositions, probably not. For a theory of the truth or meaning of referring terms or atomic propositions with actual content, I’m strongly inclined to believe that we do. The crude idea of verification (or falsification) doesn’t fit the bill. Nor, in my opinion, does Davidson’s attempt to bypass verification by applying Convention T directly to theories of meaning of natural languages.
For similar reasons I’m inclined to agree with those (Cf. e.g. Barrow, New Theories of Everything pp.51 ff.) who think that Gödel incompleteness (by which I obviously mean the Second Incompleteness Theorem) doesn’t have the implications some of the more wild eyed on both sides of the political fence claim for it. Completeness is a notion from the formal metatheory of logic. Those propositions that the Second Incompleteness Theorem shows are not theorems of a sufficiently rich first order system even though they are true according to a model of such a system are generated on the basis of a version of Cantor’s diagonal proof. That is, they do not, at least when they are initially identified, belong to the set of first order propositions the domain of whose models are countable (the only kind that qualify as theorems). Whether there are uncountably many states of affairs in the physical world is an open question. I hazard to guess that the same doubt can be raised for many ontologies that are not restricted to the natural world of the physical sciences. Of course at least one such expanded ontology would include the symbols of first order systems, which Gödel insightfully selected as the model for his meta-proof. (Actually the model can be any set the number of whose members is uncountably infinite; the symbols of the first order system are just an illustration that such a set exists, in the set theoretical sense of “exists.”) Again I am led to wonder whether the proof does anything more than show the impossibility of reducing all of mathematics to first order logic. Of course it doesn’t “prove” that there is a buzzing hive of Platonic entities out there somewhere over and above the purely abstract members of countable sets. And it is worth remembering that, in the terms of the proof itself, each string of symbols, once generated by the diagonal method, immediately joins the set of countably many strings of symbols and may therefore be associated with a theorem of the first order system. If as a result Gödel incompleteness doesn’t have much to do with the meaning of first order atomic propositions, it certainly doesn’t satisfy or prove unfounded our felt needed for something like verification in a theory of linguistic meaning.
3. Just What Are You Up To? Carnap is unclear whether his anti-metaphysical rhetoric also means that philosophy is to be rejected. Lack of specifics on his part compels us to draw up a list of what seem to be the options, perhaps the only options. It may be that philosophy and metaphysics are identical. In that case, it is an open question what we should call, or more relevantly how we should regard essays like The Unity of Science and Carnap’s other writings. If they are scientific writings we may ask: Which science do they belong to? If not, then either Carnap is contradicting himself or else he has given us no reason – no reason based on observation or logic or metaphysical reasoning - why we should believe what he says. If there is a valid type of philosophy Carnap provides no description of what a non-metaphysical philosophy looks like. Non-metaphysical philosophy may be no more than the creation of logical systems in which case there is good reason to believe that non-metaphysical philosophy is a branch of mathematics. It is not irrelevant that first order two-valued quantified logic or the predicate calculus is now a mature discipline from which we probably can’t expect very much that is new. Developments would have to come from the outside in the form of new mathematical intuitions or a radical shift in perspective. Branches of mathematics do in fact exhibit a life and reach a stage of maturity beyond which they don’t really develop. Euclidean geometry is a mature science. Cartesian analytic geometry and non-Euclidean geometries did not add to the theorems of Euclidean geometry. Newtonian/Leibnizian calculus is another example of a mature mathematical science. The conclusion may be that the maturity of the predicate calculus also means the end of philosophy. Another alternative is that Carnap is defining or contributing to a discipline that may be called the foundations of science but that this discipline is not philosophy. It expresses in a prescriptive manner what the basic concepts and rules of scientific research are much like the rules of a game define its behavior. What makes the foundations of science practiced in this way not philosophical is that following the rules is not justified by argument. It is something you have to do if you want to play the game. Carnap’s prescriptions are something like the opening pages of Julia Child’s Mastering the Art of French Cooking.
But a further problem is that Carnap does address philosophical issues and claims to decide in favor of one side of those issues despite the fact that the terms of the game he sets up imply that those issues are undecidable, that in fact any discussion of them – presumably including his own – is meaningless. One such issue has to do with the reduction of scientific language to the language and conceptual structure of what he calls physical language or methodological materialism (pp. 28-29). This is a philosophical issue because no observation or logical system will support reduction to materialist language, because at least one alternative to materialist language, viz. phenomenalist language, is practically defined in terms of its relation to materialist language and because Carnap condemns phenomenalist language as metaphysical. Since metaphysical arguments are meaningless Carnap cannot even express materialism as a metaphysical preference to phenomenalist language. It boils down to a methodological preference like a taste for grilling over frying. Nevertheless the adoption of a materialist language in place of any other would seem to have no effect on either the conduct of scientific research or the statement of research conclusions. It is unlike the preference for operational definitions in psychology which do have an immediate impact on how experiments are designed. Moreover, German physicists at the end of the 19th century were perfectly comfortable in reporting their research and discussing their theories in a version of phenomenalist language. In fact Carnap’s preference may not even be valid methodologically. The so-called Copenhagen interpretation of quantum mechanics assumes a type of phenomenalism. Indeed the fundamentals of quantum mechanics may require a phenomenalist starting point.
Let’s do a quick thumbnail analysis of different sorts of languages to perhaps shed some light (That’s a quantum joke, get it?) on the matter. Carnap proposes something he calls materialist or physicalist (p.95) language as a sort of language into which all the truths of science can be translated. Let us call it an example of a reductive language. A reductive language is a language into which you can create a function from all the true statements of a given language. But there are reductions that go beyond a simple translation function. There is a big difference between reducing meanings to extensionally defined sets, reducing the basic concepts of chemistry to those of physics on the one hand and eliminating some non-physical entities on the other. Some reductions only work by changing the truth value of purported observation statements from the reduced language. Literary practitioners protest regularly against the reduction of Hamlet to a Shakespearian brain wave. There is also a difference between a reductive language and what we might call a minimal language. A minimal language is a language that has the least number of elements capable of expressing all the truths expressed by the language (e.g. natural language) which constitutes the domain of the reductive translation. Quine’s concept of ontological parsimony is the basis for a program to construct a minimal reductive language. The language Carnap devises in Meaning and Necessity may not be minimal in this sense. A reductive language can be rich and not minimal. A minimal language such as a language representing the predicate calculus may not be reductive. After all from a formal point of view you can quantify over anything whose description is does not entail a contradiction. Scientists historically prefer reductive and minimal languages because of a preference for simplicity qualified by the ability to capture and explain all observations.
Let me propose the notion of homology between reductive languages. Two reductive languages are isomorphic, to use Carnap’s term, if a one to one correspondence can be established between the sentences of those languages. Material language and phenomenalist language are isomorphic. Physical language and Carnap’s straw man of non-physical quality languages (pp. 65-66) are not (but this non-isomorphism does not address the question of whether physical language captures everything we mean by qualities). In fact, it is easy to define any number of languages that are isomorphic to physical language. You can create such an isomorph at home. Just add an operator that does not appear in the physical language to every sentence expressing a proposition in the physical language. Let’s do that right now! Let me introduce an operator symbolized by ю. I will call this the phuk operator. An axiom of my meta-theory is: For any proposition p in the object language, p is true if and only if юp is true in the ю language. Thus p and юp are truth value equivalent. You can read юp as “Phuk it that p” if you so choose. But let me stress that the assertion of юp doesn’t have to be a value judgment on the part of whoever utters it. It says what it says and no more. Let me call the set of all the well-formed propositions in the form юp the thesaurus of the ю language. Another axiom of my meta-theory is that for every proposition p in the physical language there is a proposition юp in the ю language and vice versa. The thesauruses of the languages are extensionally equivalent. The physical language and the ю language have the same symbols with the exception of the addition of ю in the ю language. The grammars are also equivalent within the scope of the ю operator. Therefore, since there is no philosophical basis argued by Carnap for the use of material/physical language over any other (Material language is just a preference like a taste for Sacher Torte) and since there isn’t even any content given to the notion of material language outside of the unexplained claim that it is supposed to somehow represent “reality” (but cf. (7) below), the choice of a material/physical language is a lexical and not a philosophical or scientific issue. Carnap’s claim that apart from physical language no intersubjective language is known (p. 66) is hereby simply false.
Carnap does assert that he does not advocate the “metaphysical components” of materialism, merely its “logical components” (Ibid). Actually I wasn’t aware that materialism had any logical components, or at least any that aren’t shared by nearly every competing broad generalization about the world. On pp. 93-94 Carnap makes a peculiar qualification that immediately inspires the reader to ask, “Well then, what’s the point” The qualification is that the use of his physicalist language does not involve existential commitments. His actual words are, “…we are referring to a thesis which speaks simply of the logical possibility of certain linguistic transformations and derivations and not at all of the ‘reality’ or ‘appearance’ (the ‘existence’ or the ‘non-existence’) of the ‘given’, the ‘mental’ or the ‘physical’.” So Carnap’s protocol language is nothing more than a translation machine of the kind your smart phone has programmed into it so you can order lunch in Shanghai. If the translation is purely lexical, then the logical positivist program seems like a pointless exercise since physicalist protocol sentences do not tell us anything more than is contained in the original sentence. They do not tell us, for example, that the objects referred to in the sentence translated from can be understood as physical objects. If they did, that would imply an existential commitment. At least a successful translation should tell us that it is possible to regard apparently non physical objects as physical (but, as I noted above, if the translation were one-to-one, the reverse would also be possible). What is lacking is what you might call a hard core semantics of the physicalist language, i.e. a proof that physicalist language does indeed entail existential commitments. In fact Carnap confuses two types of translation which can be distinguished on the basis of the language translated from. We might call these metaphysical (Of course, the correct term is “‘pataphysical,” but for simplicity’s sake and to maintain consistency with Carnap’s text I will use “metaphysical.” The informed reader is invited to substitute “’pataphysical” for “metaphysical” throughout this paragraph) translation and reductionist translation. A reductionist translation occurs when the terminology of a given science (chemistry, for example, or in Neurath’s world sociology) is translated into the language of physics (and, if Russell were right, ultimately into the language of ramified first order logic). This is presumably what Carnap means by “unified science.” The terms and concepts of sciences like chemistry and biology are, where they are not found in the vocabulary of physics, mere shorthand, a convenient way of expressing observations and laws whose expression in the language of physics would be unwieldy. This may be possible (although the ideas of complexity and organizational principles that are beginning to make their appearance in current physics and cosmology may include irreducible terms). However, it is distinct from metaphysical translation, which is the other kind of lexical transformation Carnap gets pretty hot and bothered about. Roughly speaking, a metaphysical translation occurs when a vocabulary that is meant to refer to the whole world or a significant portion of it is functionally mapped on to another vocabulary that is meant to refer to the whole world or a significant portion of it. What is different is that the former vocabulary is not the vocabulary of an individual science. It shares a status with physicalist language if the reductions of the first kind of translation are successful. That is both the language translated from and the language translated to apply to the whole of reality. It is this feature that leads me to call this kind of translation metaphysical. If that is an appropriate term, then Carnap’s proposed translation is a metaphysical maneuver and his arguments for it are metaphysical arguments. Carnap is not rejecting metaphysics. He’s practicing it.
Carnap asserts that there are no alternatives to physicalist language (“…any universal language would affect a unification of science but no language other than the physicalist is known.” (p. 96)), but he doesn’t prove this view. He doesn’t even try. His mistake probably stems from the fact that he doesn’t recognize the difference between reductionist and metaphysical translation. Clearly any metaphysical language like the language of intersubjective phenomenalism à la Husserl or the neutral operator languages for which I provided a translation algorithm above would qualify as competing universal languages. Both competitors would satisfy the transformation requirements Carnap suggests later in the same paragraph (pp. 96-97). It should be obvious that as long as a reductive language is ontologically neutral and presents itself as simply a lexical translation of an alternative language, its uniqueness cannot be proved. (On p. 84 Carnap seems to concede the possibility of alternative reductive languages, viz. a purely experiential language and a purely physical language. But he doesn’t raise the question of an isomorphism between the two. Also these comments contradict everything he says on the subject in the rest of his essay.) Indeed the very possibility of a one-to-one translation proves the opposite, namely that it is not unique. It is almost as obvious that the only way to demonstrate uniqueness is via an ontological argument, i.e. an argument to the effect that a reductive language has a unique set of referents and these referents are all that in fact exist. At the end of his essay (pp. 98 ff.) Carnap states the relevant ontological conclusion without a supporting argument. Ostensibly he believes that no argument is necessary since he already claimed that there are no competitors to physicalist language as a reductionist language, a claim we found to be false.
Let me suggest a different term to replace those of reality or Wirklichkeit or the external world or objectivity. That is the concept of autonomy. Autonomy does not have the doubtful overtones of objectivity as opposed to subjectivity or relativity. Nor does it have the spatial overtones of externality as opposed to in the mind. It simply serves to characterize those objects whose features and behavior are independent of or not in a way I will not define here under the control of the observer.
Recent developments in physics such as string theory some of whose hypotheses are unverifiable do not directly contravene Carnapian principles since those hypotheses have a relation to observations even if that relation is not one of straightforward deduction or induction. But they do highlight a glaring defect in the logical positivist view of science. That is the lack of a place for theory (or model) construction. Carnap writes as if science were just observations and laws (i.e. inductive generalizations from observations). He reserves no place for the creation of theories that provide the conceptual framework (mass, force, energy or evolution) for the statement of laws and what to look for in data gathering. States of the universe that are not just unobserved; rather they are unobservable and cannot be linked to observations because they involve broken symmetries (Cf. Barrow p. 59). The same can be said about the concept of complexity that is current in theories – or models – of science. The idea is that some systems such as the brain, the cosmos, processes occurring in biological populations such as natural selection and indeed certain human social systems may reach a level of complexity such that they cannot be explained by laws translatable into the laws of physical theory. That is the outcomes may be irreducibly random from a physics point of view (Some would say from any point of view). The implications of this development for the unified science program remain unclear. A lot depends on whether the complexity is essentially irreducible or is simply the result of the fact that an infinite amount of data would have to be accounted for. We would also need to distinguish whether an infinite amount of data is may be generated by recursive operations or some other effective method or not. Other new ideas that threaten the unified science program are randomness, the distinction between a law and the outcome of a law, asymmetry, chaotic systems, doubts about algorithmic compressibility (which is just another name for reduction), the limits of prediction even probabilistic predictions and so on. Some theoreticians of science have gone so far as to use the term “naïve reductionism” (Barrow, p. 173), which is the project of reducing “…everything to its smallest constituent pieces (sic)…” and to characterize this project as “misplaced.” However, not only Carnap’s unified science but the mighty structure of Russell and Whitehead’s Principia boils down to an instance of naïve reductionism. We might regard the Principa as proposing extreme naïve reductionism since its end point is not the laws of physics but the axioms of first order logic.
There are two types of theory. The first is the theory that defines an individual science. In this sense Newtonian mechanics was the theory of physics or natural philosophy and evolution is the theory of biology. Then there are broader philosophical theories. One such is general ontology. Carnap never distinguishes between the two sorts of theory. He appears to reject general ontology when he rejects metaphysics, but he does not recognize or does not address the status of general ontologies as theories. He does not defend his polemical position against the possibility that different sciences may subscribe to different ontologies.
In a similar vein Carnap confuses scientific laws with scientific theories. The former can legitimately if roughly be described as generalizations inductively arrived at from multiple single observations. A scientific theory, on the other hand, is a matrix of concepts and principles that define the science. Newton’s laws of motion are scientific laws. The concepts of force and matter are part of his scientific theory. Most scientists regard theory as resulting from insight. Theories are what constitute the creative side of science. Scientific revolutions can mostly be identified as changes in theory.
4. When Clear Distinctions Get Muddy. Carnap opens his programmatic statement of the unity of science by distinguishing between the sciences of form – logic and mathematics – and the sciences of content – the so-called empirical sciences. The difference is that, while the true propositions of the empirical sciences tell us something about the world (more specifically the material world), the theorems of logic and mathematics convey no information at all. They consist only in the rule-determined application of certain compounding operations. So an instantiation of the Law of Non-Contradiction tells us nothing about the world even if an atomic proposition instantiating the variable position would do so if stated alone. For example, the proposition “¬((Jenna Jameson was a porn star at timet) & ¬( Jenna Jameson was a porn star at timet))” conveys no information about Jenna. This has become something of a trite distinction for many people, but at the same time it is a matter of serious doubt among some logicians and mathematicians. I won’t go into how new mathematical disciplines such as fractal geometry are created on the basis of intuitive pictures of the world (what we used to call models without ambiguity) and the degree to which the axioms derived using such pictures can be deduced from number theory. We need only glance at elementary logic. Carnap’s only example of the formal nature of logical theorems is the Law of Non-Contradiction itself. That is to say he chooses an example based on a purely contextually defined propositional operator and an arbitrary atomic proposition.
Things are not so obvious, however, if we move to another level of theoretical complexity. The operators of the predicate calculus are defined by way of the notion of a set or class. This is the case whether sets are part of the semantics of the predicate calculus or are treated as part of the first order theory à la Russell and Whitehead. Now it is not at all clear that the notion of a set can be contextually defined. Russell (The Principles of Mathematics pp. 18 ff.) takes the notion of a set as a primitive. Yet we may be equally justified in deriving our notion of a set from empirical observation; it may come from a primitive observation of not just rabbits, so to speak, but a group of rabbits. What is the difference between a gaggle of geese being a logical primitive and a gaggle of geese being something we observe in the world? Consider the equivalent question: What is the difference between a goose being a logical primitive and a goose being something we observe in the world? In either case, the third alternative, namely that either “gaggle” or “goose” is contextually defined by way of the axioms of the predicate calculus is excluded. “Goose” is not a logical concept and, by way of the same argument, neither is “set.” Of course, the definition of “set” (or alternatively “class”) is a complex issue beyond the limits of this discussion (Cf. Church pp. 28 ff., Russell, Principles of Mathematics pp. 66 ff. and Suppes pp. 17 ff.). In particular defining sets by way of set identity may be an adequate substitute for empirical ostension. But there remains room for serious doubt. How can you know if two sets are coextensive if, lacking a meaningful observation, you don’t know what a set or a member is? Plato of all people or whoever wrote Epinomis provides an elegant empirical description of how we learn to count (977D) and a plausible model of how sets are discovered through experience. Any number of working definitions in the natural sciences and psychology are framed in terms of identity. But no one will deny that a phenomenon so defined is “in the world” or that its concept is not purely formal. Once we engage in this path moreover, it is possible to look at the concept of a primitive proposition and wonder whether the idea of a state of affairs such as satisfies it or in virtue of which it is true is not itself something we understand due to observation. (Obviously I don’t mean particular states of affairs, but the state of affairs as such, so to speak, just as, when I say “porn star” in some contexts I don’t mean a particular porn star but porn star as such.) So even the propositional calculus, or at least the interpreted propositional calculus, may not be purely formal. To put the matter somewhat prejudicially: Do we come into the world with the innate idea of states of affairs or do we understand what the term “state of affairs” means on the basis of observation and learning? I won’t go further since this should be sufficient to unravel Carnap’s neat distinction between formal and empirical science. I might just conclude by noting that the distinction between formal and empirical science causes serious problems for the extreme naïve reductionism I discussed in (3) above.
5. Ontological Weight Loss. Carnap’s essay contains at least two different projects. The first is reductionism, both ontological and methodological. The second is the distinction between the material and the formal mode of philosophical and scientific speech. These are not the same.
Carnap’s early distinction between a formal and a material mode of scientific speech is curious because it signals a radical departure from empiricist principles that the meanings of our words are ultimately determined by what we experience in the world. Of course, the specifics of Carnap’s view would come to be rejected not least of all by Carnap himself. And Carnap seems to be mostly motivated, as are other philosophers who choose to confine their remarks to observations about language, not by a rejection of speech about material or physical objects (That is the confine of the empirical sciences) so much as a rejection of talk about essences and unobservable forces and entities - what they consider to be the stomping ground of traditional metaphysics. But there is a legitimate question of whether the baby of the physical world doesn’t get thrown out with the bath water of essences and their kin. So there might be some value in recalling a few of the stranger aspects of Carnap’s reasoning. In the first place, Carnap’s view is that specifying the meaning of all referring terms is a matter of translation whether across languages or within a given language. Even ostensive definitions, the crucial hook whereby the later Russell (An Inquiry into Meaning and Truth) tied the language of science to reality, turn out in Carnap to be simple translations between phrases using indexical terms and phrases using definite descriptions. The result is a linguistic idealism that acknowledges no reality outside the syntactical rules and semantic transformations within a given language. Does this demonstrate affinities with pragmatist philosophies and so-called coherence theories of truth (or in fact the differential semantics of French structuralism)? Very possibly. It also seems to fit with Hilbert’s formalist views of mathematics (in which case it may also fall to incompleteness results). None of those approaches to philosophy or mathematics sit comfortably with the positivism to which Carnap claims he adheres. Furthermore, even if we can swallow the overall goal of the construction of a formal mode of speech parallel to the material mode, Carnap’s version is nothing if not crude. Consider Carnap’s translation of a statement about the object domain of economics from the material to the formal mode. The material version runs “The propositions of economics describe economic phenomena such as supply and demand.” The formal version runs, “The sentences of economics can be constructed from expressions such as ‘supply and demand.’” If indeed these two versions are equivalent as far as truth value is concerned, the only distinction is that the latter is far clumsier than the former.
But is material speech more “metaphysical” than formal speech? Carnap states that an advantage of the formal mode of scientific speech is that, when we use it, we are not deceived into thinking that we are looking for the essence of things like supply and demand. However, if the material and the formal are indeed truth functionally equivalent, then if the former involves a search for essences, so does the latter. The only difference is that the latter uses a lot more words. (There is an open issue as to whether truth value equivalence is the correct analysis of certain pairs of statements. I don’t believe that the reasons for this controversy have direct bearing on my point about the truth value equivalence of paired statements in Carnap’s formal and material modes. An Austinian analysis could very well be applied to paired material and formal mode statements while maintaining their truth value equivalence. It would simply require restrictions on the utterances of such statements. For more on this, cf. Strawson, Logico-Linguistic Papers pp. 165 ff.) The one type of research I can think of offhand where direct observational language about things should be replaced by an equivalent is where the phenomenon to be studied is for one reason or another inaccessible, such as in quantum mechanics or cognitive psychology. The translation in these cases, however, is not into a formal mode of speech. It is, rather, into another material mode, namely the language of indirect observation. And truth value equivalence is not a problem; it is in fact the point of the translation.
Now let’s turn to reductionism. Carnap states that his brand of reductionism is purely methodological and does not entail ontological commitments. But in the course of working out and defending methodological reductionism he throws, perhaps unavoidably, ontological conclusions into the mix. My feeling is that the background of Carnap’s reductionist project is the more venerable philosophical project we call foundationalism. Ockham and Hobbes may be considered reductionists but not foundationalists. Descartes and Husserl are foundationalists but not reductionists. Clearly foundationalism and reductionism are not the same thing. Descartes’ foundationalism was based on proofs of the existence and benevolence of a deity and was completely compatible with his rich and ramified dualist ontology. Husserl’s foundationalism was almost entirely methodological and was compatible with each of the many ontologies he espoused during the course of his philosophical career. Likewise ontological reduction by itself produces unassailable foundations for knowledge - the kind Russell required – only if a certain class of propositions are, for all intents and purposes, necessarily true and our knowledge that those propositions are true is certain knowledge. Moore and Russell certainly believed that sense data reports fit the bill and Husserl held the same beliefs about the results of phenomenological introspection. Assuming Carnap avoided making the same claim for his protocol sentences and the experiences they report (This is somewhat ambiguous), it follows that, while the ultimate overall theory might be unified and, to use Quine’s term, ontologically parsimonious, it is not founded in the sense demanded by Russell. Hence my intimation above that Carnap’s views on the language of science look more like pragmatism than hard core positivism. That is not in itself a bad thing, but it does raise the question of why admissible protocol sentences should be any more valid or acceptable than other possibly justified propositions whose meaning is determined by convention or indeed propositions, such as propositions about essences, that do not derive from direct observations.
If reductionism is to be foundationalism, there must be something unassailable about physical language (as well as at least some rules of deduction and induction). The reduction cannot just be a working tool. It is possible that Carnap is not interested in providing firm foundations for science. (He never clearly disentangles epistemological foundationalism from his ontological project.) But then he is exposed to the response that other philosophers and other scientists could express preferences for other working tools. And he has provided for no defense against powerful counter arguments such as the argument for levels of complexity. (On p. 87 he remarks on the current non-existence of descriptions of a brain such as could constitute the isomorphic counterparts of sensation descriptions. He also seems to concede that the physical counterparts are not located in a single part of the body. However, he does not raise the possibility of irreducible complexity.)
Let me introduce what appears to me to be a subtle argument that Carnap would probably reject as metaphysical, but if valid, would undermine a foundationalist project involving language or linguistic tokens such as protocol sentences. You might call it the veil of language argument. There is a difference between an immediate experience and the protocol sentence that reports that experience. Consider the possibility of a level of experience that is not predetermined by the grammar in which the protocol sentences are couched (for example, an experience of a world that is not populated by the “logical fiction” of distinct particulars). A skeptical anthropic argument might be used here. We perceive and conceive a world populated by particulars, the referents of our referring terms, and as a consequence we have constructed our language (Language is an artificial human construct, after all) to mirror the possible necessary category of particulars. But this may be no more than a result of the arbitrary way we have evolved. Perhaps the fiction of the logical particular was important to our survival as a species. In that case the very grammar of the protocol sentence would be based on a fiction. Is the anthropic version of the veil of language argument metaphysical (i.e. ’pataphysical)? Well, our current existence with a certain mental constitution is an empirical fact.
Carnap does not sufficiently distinguish between his anti-metaphysical attitude, reductionism and materialism (or physicalism). While you may reject metaphysics and espouse reductionism and materialism, you are not obliged to take all three positions. You can be anti-dualist or anti-spiritualist without endorsing reduction to the language of physics. You can be anti-transcendentist and so reject metaphysics in that sense while accepting the possibility of a phenomenalist description of the immanent universe. One of the faults of logical positivism was to pour all these distinct views into a single bowl, stir vigorously and assume you had to accept all of them if you accept any one.
Carnap does not mention arithmetic and other fields in pure mathematics as examples of sciences. Perhaps he assumed that the reduction of arithmetic to logic was a done deal. But that is not obvious. The set theoretical paradoxes as well as the metatheorem that, roughly speaking, some truths of mathematics are not theorems of any sufficiently rich first order logical axiomatic system, continue to cause problems for the logicist. It is significant that these issues arise from the idea of a totality of objects, the core ’pataphysical issue that Carnap ignores. Another reason for Carnap’s silence on the issue may be that Russellian Principia style reductionism is not the direct influence on Carnap and logical positivism. The direct influence was the Tractatus which upheld not so much a deductive type of reduction as what might be called a specular view of the relation of language to the world. According to the specular model a logically adequate language is not one from which the theorems of the empirical sciences may be deduced. Rather, when the theorems of the pure and empirical sciences are reconfigured into a logically perspicuous form, they constitute a mirror of the world. Moreover, even though Carnap ignores in this text a Principia style reduction, he does advocate, as I shall discuss below, a different sort of reduction, viz. a reduction of all statements that qualify as scientific statements to propositions solely about the physical world. This type of reduction is not the same as logicist reductionism, although one can safely say its conclusions functioned as assumptions on Russell’s part (Principia ends with the reduction of physics to geometry). We need only think of his injunction that we maintain a robust sense of reality.
Whatever Carnap’s attitude about logicist reduction may be, the Russell/Whitehead attempt to translate the basic concepts mechanics into the language of pure geometry (“Matter is whatever occupies points in both space and time” etc.) has serious defects. The translations of the Principia apply only to the basic concepts of a physical theory (or at least a physical theory that incorporates geometry). The validation of specific theories such as Newtonian or non-Newtonian mechanics and the laws involved in those theories remains a matter for observation, even though, strictly speaking by the premises of logicism, it shouldn’t. The laws of mechanics should be derivable by pure deduction, i.e. the application of rules of inference, from the axioms of first order logic. Moreover, the reduction is never pure. Matter, for example, is not defined in terms of space and time. It is defined in terms of space, time and the concept of occupying. The concept of occupying is not geometrical or logical. It is a new concept.
Within the reductionist project there are two separate issues. The first - and I suspect this is the original motive for reductionism – stems from the desire to eliminate “unreal” entities from our ontology. The worst culprits among unreal entities are god, gods, angels, devils, demons, souls, spirits and the perfect woman - the entire menagerie of anthropomorphic feux follets and the volitionist or intentionalist cosmological theories that went along with them (Think of Anaxagoras or Plato’s Timaeus or the Xtian god if you want an idea of what an intentionalist cosmology looks like). Later positivism added so-called ideal entities to the list: spirit of the age, culture, history, mind, élan vital, will, senses and meanings (notably a part of the later Carnap’s own ontology), intentions (including Brentano’s intentions) etc. It is important to remember that these latter are not part of the volitionist cosmologies of the religious stage of human thought (Remember Frazier’s three ages of magic, religion and science?). They don’t play much of a role (Scholastics such as Aquinas did have a concept of intention) in the suffocating theological ideologies that scientists have worked so hard to combat. And for that reason the only benefit in jettisoning those entities from an acceptably scientific ontology is that they are superfluous, that they overcomplicate theories that could get along well enough without them, that they violate the mandate of simplicity, and ultimately they are non-explanatory. These benefits are substantial but mostly operational. They are in no way comparable to the benefit of unmasking and undoing the very real political terror that religious tyrannies are capable of visiting on credulous populations.
The problem of ideal non-volitional entities introduces the second and very distinct issue that also subsists under the name “reductionism.” That issue is whether – assuming the objects of scientific theories are limited to so-called “real” entities, the entities referred to by conceptual framework of physics, and assuming that the other Principia reductions are valid - the remaining conceptual framework of pure formal logic is sufficiently rich to capture a satisfactory general theory of the world, that is a theory from which can be deduced all the agreed truths of science. It is important to be clear about the distinction. Most psychologists, for example, would agree that our cognitions and emotions are reflections of our brain states. They do not have a separate ideal existence in some sort of mind. They are in principle reducible to sufficiently rich concepts of neurophysiology. But that does not mean that a theory capable of capturing all the truths of basic organic chemistry would be capable of deductively capturing the truths of psychology, even in minimalist versions like cognitive psychology. There was a time when scientists believed that behaviorism was an example of a successful scientific reduction. Even if that project had succeeded, however, it is important to remember that there was never a concept of behavior in organic chemistry or even a proposal to translate the concept of behavior into the theorems of chemistry. The path from the chemistry of plant metabolism to the evolution of plants seems clearer, but there are gaps introduced by the concept of animal behavior in the theory of animal evolution. The culprit may be a qualitative jump in complexity such as many physicists have begun to argue is the case in physical theory. Or we may be at too crude a level of neurophysiological observation. We may still be waiting for our neurophysiological Hubble telescope. Nevertheless, even assuming refinements in observation, just as with the concepts of set or space/time occupation, new concepts seem to be inevitably introduced at each level of “reduction.”
Finally it is important to understand the concept of the mathematicization of nature and to distinguish that from the reduction of all the sciences to (effectively) formal logic. Mathematicization as it is practiced by almost all sciences involves the application of a mathematical model to explain observed regularities and make predictions of future events. The model chosen may be ultimately reducible to basic arithmetic à la Russell. However, the observations are not part of the model; they are measured by the model. The observations contain concepts that are not deduced or completely defined within the axiomatic system of the mathematical discipline by which they are measured. The creation of a successful model for predicting future observations does not constitute a reduction of the entire science (model plus observations) to pure arithmetic.
6. Or How to learn to Love the Bomb. There is a picture in the popular imagination of scientists as heartless technocrats accurately captured by the figures of Mr. Spock and Dr. Strangelove - individuals whose only concern is the most efficient way to achieve a predetermined goal and who have no interest in whether the goal itself has merit outside of its logical compatibility with other predetermined goals (that is to say, the assertion of that goal does not contradict the assertion of other goals). Heidegger gave philosophical expression to this popular image in the two opening essays of Vorträge und Aufsätze but did not suggest a particularly satisfying alternative (Living in The Black Forest and making coo coo clocks is not my idea of a satisfying alternative). The attitudes of scientists thus understood are in fact those professed by the various manifestoes of logical positivism. The popular prejudice found itself justified in the failures of newly empowered technocrats in the management of the Viet Nam war and the failures of social engineering and economic centralized planning in Europe and the Communist countries. Opposed to the technocrats were the more romantic and disorderly forces of people’s revolution in Viet Nam, on the one hand, and the economic hidden hand so eloquently defended by Hayek, on the other. Right at the beginning of The Unity of Science is a paragraph (pp. 23-24) that metaphorically lights the fuse of the scientistic (not scientific) fiascoes that were to come at the end of the 20th century. This is the paragraph where Carnap rejects normative ethics as “metaphysical.” Much can be said for a distinction between normative ethics as a collection of ethical maxims and philosophical ethics as an investigation of the form of all possible maxims, and so a case can be made for pursuing moral philosophy at a level of abstraction that does not necessarily commit an individual philosopher to a specific ethical canon. But that doesn’t really make much difference. Carnap makes clear that he rejects both approaches; the only pursuit he will allow is the empirical, i.e. sociological or psychological study of the nature or human behavior and moral judgments. This is Strangelovism at its most absurd. “Don’t bother me with fuzzy questions about what we should do. All that interests me are the best (presumably most efficient) ways of reaching your goals, whatever those goals may be.” The dangers of just putting the technocrats in charge was exploded as early as the 3rd century BCE in the apocryphal Platonic dialogue Alcibiades II (144C ff.) where Socrates makes two important distinctions. The first is between knowledge (ἐπιστήμη) and knowledge of what is best (ἐπιστήμη τοῦ βελτίστου), i.e. knowledge of natural facts and knowledge of what to do with those facts. The other distinction is between knowledge of the art of moral suasion (πότερον βέλτιον) and an understanding which moral conclusions should be drawn (ὅτε βέλτιον). When these distinctions are not made, Socrates concludes, people in democracies allow themselves to be persuaded by those who pretend they have the moral answers, namely slimebag politicians (ῥητορικῶν πολιτικὸν φύσημα φυσώντων). Carnap commits the philosophical error of believing that since the logical positivist supposedly knows the answers to the questions posed by the first disjunct of Socrates’ distinctions, he also knows the answers to those posed by the second disjunct. The successful campaigns waged by peoples’ revolutions and libertarian movements on the opposite ends of the political spectrum have exposed the vacuity of the efforts of often well intentioned Carnapian technocrats. (Obviously the conclusions drawn by right wing religious bigots to the effect that scientific facts are themselves mistaken is a patent absurdity of the other sort; what is at issue is the application of those facts in a moral context not the facts themselves. Likewise the left - much romanticized when it took the form of peoples’ revolutions - almost inevitably constructed a far more technocratic economic system than the regimes they replaced. The issue in either case is not right or left but logical positivism itself and the foolishness of forcing people who had rejected moral reasoning into the role of moral arbiters.)
It is worthwhile remembering Dworkin’s criticisms of what he calls “legal positivism” in this context. For legal positivism as Dworkin defines it is a descriptive theory of actual laws and cases that denies the existence of any jurisprudential norms not enshrined in the facts of written legislation and judicial case books.
7. There is no such thing as “the given”. Accordingly the idea that science can be based on protocol sentences or primitive protocols that are direct reports of “the given” is a myth. Carnap agrees to a point. But there are two distinct arguments in play.
The first has to do with protocols that, for argument’s sake, we will assume no one questions. A version using Carnap’s wording would be (a) “I now see a thin, long brown body.” This is supposedly a more “authentic” version of (b) “I see a copper wire.” Carnap calls (a) a primitive protocol because it involves (supposedly) fewer inductions than (b) in order to be asserted as true. (a) is somehow a more authentic version of (b). But that is just ridiculous. In the first place, it is not obvious that (a) involves fewer inductions than (b). It seems to be the case because under most circumstances (b) is true only if (a) is true but not vice versa. (a) may be compatible with (c) “I see a stalk of wheat.” But there is no reason that both (a) and (b) could not both “involve” a countably infinite number of inductions (that is, we have no rule for when to stop looking for presuppositions). Moreover, time throws a wrench in the works of all protocols. I may be sure of both (a) and (b) at timet1 but have no justification for asserting at time t2 the truth of “I saw a thin, long brown body at timet1” Finally there is nothing primitive, obvious, needing no justification or given about the concepts thin, long, brown and body. These are not part of the world, part of the given. Like the concepts copper and wire they are part of our conceptual store for classifying “the given.” Bedfellows as strange as Quine and D.T. Suzuki have called the concept of an object an artificial construct. Hegelians would say that, like all concepts, the concepts in primitive protocol sentences are not directly given, they are mediated.
Carnap would agree or at least not seriously disagree with the first argument. In fact he states a version on pp. 46 ff. where he credits the first argument to the findings of Gestalt psychology. The importance for Carnap of protocols is not for deductive foundational purposes but for purposes of verification. Protocols are deduced from laws of nature considered as hypotheses, not the other way around. To be more precise, Carnap does not even call this verification “in the strict sense.” There is always, he says, an element of convention in the verification of a law. The troublesome notion of meaning as internal to a set of statements has a way of popping up in the darnedest places. Furthermore, Carnap does not decide – which in my opinion he has to in order to clarify the notion of a protocol – whether an element of unadulterated observation can be extracted from the conventional mix or whether everything stateable (even protocol-wise) is thoroughly conventional.
Carnap goes on to entertain the notion of a single individual performing the verification of a law (or any statement) by determining whether protocols can be deduced from it. After telling this story he rather abruptly concludes that if “several persons” perform the same acts of verification and reach the same conclusion, then the laws or statements in question “have sense” and are presumably verified for those persons “intersubjectively.” The next stage in his argument will be to claim that physical language is the only intersubjective language and is therefore the only possible candidate for an intersubjectively objective language of science. He never actually proves the oft repeated shibboleth of the intersubjective uniqueness of physical language. And that point is false. Brain in the vat theorists contrive ingenious thought experiments of electrode links between brains such that create a phenomenalist intersubjectivity and intersubjective language between linked brains. And, of course as we saw in (4) above, Carnap can find no succor in reductionism.
There’s also something wrong with the notion of “several persons” independently performing acts of verification (deducing protocols from laws and other singular statements), for that leads directly to the second argument.
The second argument occurs at the object level. We can accept the conclusion that there are no immediate perceptions of anything other than objects (copper wires, tables, red cubes etc.). But a problem still comes up when we imagine the possibility of someone disagreeing. Suppose your lab partner were suddenly to say that he doesn’t see a thin, long brown body or a copper wire? What he sees, he says, is the son of god. The logical positivist would probably suggest getting a new lab partner and I heartily agree with the suggestion. (For more on the subject of alternative observations see my “Find the Pope in the Pizza.”) However, the problem doesn’t go away just by calling the person who disagrees with your primitive protocol a kook. (Would a theory of observation depend on a theory of rational behavior? There might be a nasty circularity in that but it might nevertheless provide some insight.)The reason lies in the logical positivist’s purpose for proposing the idea of protocol sentences. Let’s give him the floor: “…protocol statements (are – WD) statements needing no justification and serving as the foundation for all the remaining statements of science.” (p. 45) Although presented in the context of regarding protocols as the hypotheses for deduction, which Carnap rejects, that assertion still holds true, actually it must hold true, if we regard protocols along with Carnap as the conclusion of a verifying deduction. Without justification there is no defense against someone who says he does not see what you see. Otherwise you haven’t intersubjectively verified your law. You could say that reasonable people or the scientific community or all my buddies see what I see, but that’s not good enough for statements that purport to need no justification and can serve as a secure foundation (verification) for all the remaining statements of science. (You might think that Quine fixed problem in Carnap’s theory of observation by throwing out notions like sense data the certainty of protocols (e.g. Word and Object pp. 2-3). But he never questions the assumption that “well-placed observers” will always agree or agree in sufficient numbers about the truth of observation statements, or indeed what qualifies someone as a an acceptable observer outside of his disposition to agree with his colleagues. He considers “occasion sentences” as expressing observations about which “…there is pretty sure to be firm agreement…” and on which “…a scientist will tend to fall back when pressed by doubting colleagues….” (Ibid. p.44. Cf. also p. 55 on incoherent behavior.)) If the logical positivist were to resort to majority agreement about protocols, he ends up with the disaster of a sort of coherence theory of truth, the very opposite of what he wanted to achieve. Moreover, my example above gave the logical positivist every possible advantage. Alternative observations may not originate from isolated nut cases. Perpetrators of the psychotherapy scam “see” traumas like sex abuse where others see someone unduly influenced. The issue is one of verifying observation, not one limited to theory. How many in a party of those who disagree need there be before the band of rational brothers begin to doubt their own primary observations?
8. The Colors! The Colors! Carnap’s argument for the primacy of physical language over, say, phenomenal language is that every perception statement can be translated into statements about physical objects. The point is not so much that the reverse translation cannot be made into phenomenalist or ю statements. The idea is that physical statements are about objects “out there” that constitute an objective touchstone for every scientist’s personal perhaps subjective perceptions. Physical language is about autonomous objects accessible by others; the language of subjective impressions is not. This is a feature of physicalism that doesn’t belong to phenomenalism (although to my mind it does belong to ю-ism). The use of physical language doesn’t refute the challenge of solipsism or phenomenalism. It simply takes advantage of a feature of physical that behaves as if that challenge were invalid. Even though everything that appears to belong to an external world and to the perceptions of physicist B may just be in physicist A’s mind, nevertheless, the use of physical language takes advantage of a feature that treats objects as if they were extra-mental. At a second level of skepticism the entire complex of external objects, perceptions and even physicist B himself and his perceptions may be “internal” to physicist A. Nothing can be done about that except to act as if it were not the case. For this maneuver is not available to the subjective idealist. Likewise, even though the immediate domains of reference of physicists A and B when they utter protocol sentences may be the phenomena they sense or perceive, nevertheless by dint of using physical language, they agree to a secondary domain of reference which is the physical world. The spatio-temporal reference points of physical language may themselves be imaginary or in the mind of some individual or even the shared transcendental ego, nevertheless it is part of their meaning that at one level of analysis at least they are independent of the observer. (Note this is Husserl stood on his head.) And this is all that is needed. It is the methodological materialism promised in the beginning.
Carnap’s argument for standards of measurement as somehow providing additional support for the superiority of physical language is not so successful. Carnap believes that quantitative physical statements are somehow more objective than phenomenalist statements. Compared to so-called secondary quality statements, this is true. The argument is familiar from Plato. But at the meta-skeptical level quantitative statements possess no special status. A particular numerical value for a color frequency could very well be in physicist A’s mind; the difference is that the color it stands for is in a version of A’s mind that itself is in A’s mind.
But, measure vs. comparison or quality aside, even this ingenious formulation is subject to the second argument above because the nutty lab partner may insist the objective referent of his perceptions is the son of god or Satan or whatever whereas all you can see “out there” is a copper wire or a long thin brown thing (I nobly resist the temptation to obscene reference). And there is some slight urgency to this situation that goes beyond intramural philosophical theorizing. For there are disturbing numbers of nutty lab partners out there who vote (or assassinate in the case of nutty Mohammedans) and their very real activity threatens our real freedoms. There is also room for doubt that comes from Carnap’s choice of the febrile example of secondary quality perceptions and color perceptions in particular. Carnap’s hearty view is that every subjective color perception can be “physicalized,” i.e. set in a one-to-one correspondence with some quantity of light wavelength. But that is woefully insufficient. We also require a neuronal configuration that receives the wavelength stimulus and under the influence of the stimulus causes under certain circumstances a person to utter a color name. Carnap’s other example of a secondary quality expressed in the statement, “It is cool here,” is equally unfortunate. That statement is supposed to be equivalent to the statement asserting a certain temperature (more accurately range of temperatures) expressed in the supposedly universal physical language. Obviously the secondary quality statement and the physical statement are not truth value equivalent. Carnap’s generalization to the effect that in his successful construction of a physical language “…qualitative determinations in protocol language are single-valued functions of the numerical distribution of coefficients of physical states.” (p. 61) is simply false.
The doubt is how much more the internal color perception has to it than the objective light reaching the retina. This doubt becomes stronger when we think of more complex “secondary qualities” such as are the subject of the psychology of the emotions. The doubt is whether – because of complexity – we need to devise or discover other languages that are not physical even in Carnap’s methodological sense but retain the virtue of intersubjectivity.
Let’s grant that Carnap is not making a “metaphysical” point but simply a methodological observation. Let’s also assume that everything he says is not contrary to the nature of observations at the quantum level (His failure to address this is a serious defect. The Maxwell equations are one thing; the Schrödinger equations something else entirely. Probabilistic calculations do not address the issue of initial observations). We are tempted to ask, “What’s the point” Yes, physicists write up their experiments, make predictions and induce laws from their observations in language that does not imply they dreamt the whole thing. A physicist can be a thoroughgoing solipsist (and I suspect some are) or, like Newton, believe that god is the efficient cause of every single event in the universe, and yet not commit their results to paper with any such implication. Doesn’t Carnap’s methodological point come down to saying that people do what they do? A M. Homais banality if there ever was one. If Carnap’s remarks are to have any content one would think he needs to take the methodological leap and make a strong metaphysical claim. But that leap, as we saw above, would immediately expose him to a barrage of enemy fire.
9. Man-Machine Redivivus. In his Section 5 Carnap applies his reductionist views to the specific case of the reduction of the observations and laws of the science of biology to counterparts in the science of physics. The idea is that biological concepts can be subject to progressive translational reductions until we arrive at purely chemical concepts. Interestingly enough he gets down to defining spermata and eggs as cells with certain features (Which features and how can these be further reduced?), but he stops there. The cell is not a chemical concept and Carnap never tells us what cells can be reduced to. The other very limited set of concepts he assures us are reducible are those of species, organism and event by which he presumably means biological event. His theory for reducing the last concept is remarkable for its simplicity if not accuracy. Carnap states (not argues) that the very limited set of biological events he singles out are no more than instances of "spatial redistribution" and that they can be adequately explained by invoking (presumably physical) laws of spatial redistribution.
The faults are legion. Spatial redistribution of parts inadequately leaves out causality and the lawlike quality of event sequences. Carnap does not discuss whether a biological event can be predicted using current physical theory alone. He does embrace the Humean idea that laws are the inductive generalization of individual observations and that we don’t need anything else. We predict future occurrences of meiosis, for example, because we have seen it happen several times in the past. The presumption is that induction alone is adequate. Here again Carnap does not recognize the distinction between law and theory. A law is a generalization that is non-explanatory. A theory provides an explanation of the individual events. Newton's laws of motion are, as the name implies, laws. But universal gravitation and the hypostasis of the concept of force is a theory. As far as biology is concerned evolution goes part of the way to filling the theoretical gap. However, evolutionary theories like natural selection use concepts that cannot be found in chemistry.
Furthermore the point I made above about alternative reductive languages has not been adequately resolved. Simple translation into language of spatial redistribution does not of itself bring about a reduction to the language of physics. Consider the language of animism as an alternative. I presume there is a version of animism that can be adequately stated using the language of spatial redistribution and the concept of volition. If so, we can effect an equally complete translation of biological concepts into those of animist spatial redistribution wherein the idea of volition replaces the idea of mechanical cause. Carnap provides no reason why this kind of reduction should be any less desirable than physicalist reduction.
On p.71 Carnap expands his methodological reduction regarding biology to the so-called science of psychology (Psychology is not a science as opposed to a grab bag of proto scientific observations because there is no overall psychological theory equivalent to physical theory or evolutionary theories. The closest we have come to an overall theory is Freud’s theory of childhood sexuality as an explanation for human behavior and motivation. And, of course, that theory remains controversial and, for practical not metaphysical reasons, difficult to verify/falsify. The lack of a theory may not be just a historical contingency that will eventually be overcome. There may be structural reasons why scientific theories as we have come to know them don’t work for psychology. It may also be that this lack of theory is responsible for the plague of crackpot pop psychologists that degrade contemporary society. There is no such thing as pop physics or pop biology (Well, I guess Xtian biology, aka creationism, is the exception that proves the rule) and the reason is the presence of a unifying explanatory theory in these disciplines). All he actually defends, however, is physicalistic monism. This is another illustration of where Carnap confuses ontological reduction with unified methodology. There is a difference between the (in my opinion true) doctrine that mental states are reflections of neurophysiological structures and the quite different doctrine that psychological events can adequately be explained by the laws of physics alone. There is a difference between the complex reflection of what we regard as mental and psychological states, dispositions and activities by neuronal and synaptic networks, on the one hand, and the complete explanatory reduction theories of thought and behavior to the laws of chemistry.
At the sociological stage even the notion of a unified methodology is subject to serious doubt. Hayek’s classic argument to the effect that the overwhelming multiplicity of individual economic decisions renders centralized economic planning impossible also, in my opinion, has serious consequences for the very idea of a social science. A logical positivist might argue that, since economic and sociological data are in principle finite, an ideal researcher with divine capacities for observation could record all of them even though real scientists cannot. I don’t know. That idea sounds pretty damned metaphysical to me.
Carnap saw himself as attacking the metaphysical (and genuinely non explanatory) theory of vitalism. But the faults of vitalism do not justify adopting Carnap’s contrapositive and equally metaphysical (or ‘pataphysical) doctrine of physicalist mechanism. In his intellectual context arguments for the non-reducibility of biological theories to physical theories was equivalent to espousing vitalism. Mach and Schrödinger, for example, were sympathetic to vitalism. But the two are not equivalent. You can reject vitalism (and incidentally reject classical mechanism and its troubled notion of causality) without embracing any specific alternative theory. What motivates the biological anti-reductionists in the current atmosphere is the same intriguing notion that motivates anti-reductionists in physics: complexity. (Cf. e.g. Sandra Mitchell’s arguments in Biological Complexity and Integrative Pluralism. Robert Laughlin devoted an entire book to the analogous notion of emergence. It would be interesting to work out the relations between complexity and emergence.)
10. A Rose By Any Other Name. Carnap's distinction between formal and material modes of speech is one of the less successful justifications for the so called linguistic turn in philosophy. His argument is that the distinction between personal sensations and objective events in the material mode leads to an unbridgeable gap between talk of mental life and physiological observation. But Carnap's formal mode entails the same contradictions as the material mode of speech. If sensations are essentially personal so are protocol statements using the first person singular. The speaker may make use of rules and conventions that allow him to refer to his own private experience, but his use of these conventions about indexical expressions involves referring to inaccessible sensations assuming these sensations are indeed inaccessible. Reciprocally there is no reason why an isomorphism between statements about a person's body and statements about that person's sensations cannot be established in the material mode of speech. Or that inferential relations cannot be established between statements in the material mode. (It is the possibility of inferential relations that Carnap considers the particular virtue of formal speech.) The conclusion is not that solipsism is unavoidable. On the contrary, adherence to physical monism is not guaranteed by adopting a formal mode of speech. Carnap never goes beyond asserting the primacy of the formal mode. He doesn't show why it doesn't lead to the same problems as the material mode. His reasoning may be that by adopting the formal mode we commit ourselves to the objectivity of language (Supposedly – and incredibly - unlike other objects language is part of the world and not in the mind). Protocol sentences are supposedly objective things and not internal experiences. But, if we assume the truth of that dubious assumption, uttering any sentence at all commits us to the objectivity of language. Speaking in the material mode about internal sensations makes those sensations public because the material mode statements are as public as formal mode statements.
Cognitive psychology in practice has partially rejected Carnap’s unidimensional distinction between “internal” sensations and physiological facts. It completely ignores the distinction between formal and material modes of speech. It simply employs different concepts. In the first place introspection and methodological phenomenology are working tools of the psychologist because research and experiment cannot get off the ground without some basic hypotheses and sets of concepts. Often the only source of these hypotheses and concepts is introspection. Secondly psychologists use working definitions to identify “inaccessible” mental states and processes. Working definitions consist of observable physical and behavioral changes – the dependent variables of psychological experiments – to signal the occurrence of a hypostatized mental event. There is no question of there being a neurophysiological isomorph to a mental event. It is just that the identification and description of that isomorph (the state of the synaptic network that corresponds to the mental state) is not the primary goal of psychological research. Rather changes in the observable referents of the working definitions are recorded as the basis for the hypotheses of laws and model construction.
In (1) I argued that although Carnap rejects metaphysics, he espouses a ’pataphysical ontology. That is, he makes universal claims and assumptions about the nature of things. In (2) I argued that Carnap's criterion for a truly scientific statement presupposes a ’pataphysical proposition about the nature of all scientific statements. In (3) I argued that Carnap’s idea of reduction to physicalist language is hopelessly vague and belied by much of current scientific practice. In (4) I argued that there is no clear distinction between formal and observational sciences. Even the basic concepts of logic may be derived from observation. In (5) I argued that Carnap's idea of formal language shares drawbacks with coherence theories of truth and with formalisms such as Hilbert's view of mathematics. Furthermore, Carnap confuses the distinct philosophical projects of ontological reductionism, epistemological foundationalism and a physicalist interpretation of the social sciences. The reductionist project to which Carnap subscribes is fraught with difficulties. In (6) I argued that the logical positivist attitude toward ethics can have unhappy practical consequences. In (7) I argued that the notion of a protocol sentence is insufficiently clear to function as a means for verifying scientific laws. Protocols are certainly not indisputable. In (8) I argued that the virtues of intersubjectivity on the part of physical language cannot account for meta-skeptical doubts or for conflicting perceptions and conflicting testimonies. In (9) I argued that Carnap's test case of reducing biology to chemistry is oversimplified where it is not just mistaken. In (10) I argued that adopting the formal mode of scientific speech does not reach the goal of guaranteeing a working ontology of physicalist monism.