A debate between Schrödinger and Carnap on the problem of Other Minds

Michel Bitbol


First Published (In French) in: Philosophia Scientiae, 3 (cahier 2), 203-213, 1999 (with the title: "L'alter-ego et les sciences de la nature, Autour d'un débat entre Schrödinger et Carnap"), and also as chapter 2-13 of: Michel Bitbol, Physique et philosophie de l'esprit, Flammarion, 2000.

Recently revised English version:

M. Bitbol, « The problems of other minds: a debate between Schrödinger and Carnap », Phenomenology and the Cognitive Science, 3 (1), 115-123, 2004

Copyright: Michel Bitbol. Translation: Paul Tappenden.

"All the premises on which science depends, when they are not of a purely conventional nature, rest on experience". It is with this programmatic sentence that Carnap concludes an article published in French, in 1936, in the journal Scientia. This article was entitled "Existe-t-il des prémisses de la science qui soient incontrôlables?" (Are there premises of science which are beyond control?).
If Carnap believed himself obliged, in 1936, to reaffirm a strict partitioning of analytic and synthetic propositions, and if he regarded himself as having a duty to indicate again the impossibility of a synthetic proposition being beyond empirical test, thus blocking any return to some form of synthetic a priori, he took the opportunity first of all by responding to an article of Schrödinger's which appeared a year earlier in the same journal, and was originally written in French. The response to Schrödinger is nevertheless undoubtedly aimed, by way of Schrödinger himself, at other philosophers such as Neurath, Schlick or Popper, whom Carnap does not name but with whom a debate is underway in the mid-1930s .
Schrödinger's article, modestly titled "Quelques remarques au sujet des bases de la connaissance scientifique" (Some remarks on the bases of scientific knowledge), contains a thesis which is outrageous for Carnap, explaining his prompt and vigorous response. That thesis, all the more provocative to Carnap because it had been formulated by one of the greatest physicists of the time, is as follows: "Science is not self-sufficient; it needs a fundamental axiom, a basic axiom from without". An "axiom" which is radically outside the system of science because it is neither empirically testable nor assimilable to a convention. A basic axiom which Carnap thereby identifies as a trans-empirical premise which should be called metaphysical.
But what is, then, this "fundamental axiom"? Schrödinger avoids stating the content of it straight away, as if he sensed that even giving expression to it was going too far, and as if he anticipated an argumentative strategy of Carnap's wherein the most force rested, in the end, on what was naïve in the explicit formulation of such an "axiom".
The beginning of Schrödinger's article consequently seems to be a long preamble whose sole clear objective is to play for time. Then, after having underlined the impossibility of a researcher himself rethinking every theory and, above all, of himself alone repeating all the experiments which have led to the present state of his science, he starts to let his fundamental hypothesis or axiom be glimpsed. He writes: "The sensory perceptions of another human being are something which I have never experienced myself. Nevertheless, I do not hesitate to interpret them by calling up the memory of what I call my own similar perceptions". To be able to use a colleague's account of an experiment as if I myself had made and recorded all the observations, continues Schrödinger, I must accord him "sensory perceptions" like my own. The work of a scientist rests in the end on an anti-solipsistic hypothesis which Schrödinger calls "Hypothesis P" (for the "Personality" of other human beings) but of which Carnap alone, it must be emphasised, gives a synthetic version of in his reply to Schrödinger:
"Hypothesis P: It is no only I who have sensations (and consequently thoughts, feelings, memories, etc.); other human beings have them too".
That this hypothesis cannot be "verified by exact scientific method" (or rather that it is beyond "empirical control", as Carnap preferred to say following his then recent critique of verificationism in Testability and Meaning) is certainly not a circumstance which is accidental or auxiliary so far as Schrödinger is concerned. It is not just that all the observations which I am capable of making of the behaviour of another person are equally interpretable either in terms of neurophysiological processes or in terms of sensations and thoughts, but that the second type of interpretation (which would tend to confirm Hypothesis P) must necessarily, methodologically, be rejected in a particular framework of scientific explanation. To demonstrate this, Schrödinger chooses a very simple illustration: I pinch another man; the man will perhaps make an exclamation, but wouldn't that be the way in which an automaton reacts to being pinched? On one hand, says Schrödinger, we indignantly reject this degrading suggestion according to which the alter-ego could not be anything but an automaton, but on the other hand we must insist that the physiologist gives a systematic treatment. If we ask why the man makes an exclamation, Schrödinger continues, would you believe that a physiologist (in the rôle of physiologist) would be inclined to reply that the man made an exclamation because he felt a pain? Certainly not, because in replying in that way the physiologist would be ignoring the genuine scientific problem.
The physiologist who would insert into his account of neuronal processes a reference to what is felt by the owner of the nervous system would thereby merely show his partial ignorance of the class of phenomena he is responsible for elucidating. Even more seriously, if he regularised the surreptitious involvement of Hypothesis P in his explanations, he would be in great danger of falling back into some sort of vitalism, or of returning to the invocation of entelechies.
The physiologist is in fact bound to conform to a way of thinking which no-one has ever accepted overtly, but which characterises the scientific enterprise from start to finish. That way of thinking, which Schrödinger puts into a dramatic setting at the end of his Nature and the Greeks (1948) , and then in his 1956 book Mind and matter , is that which moves towards objectivising. In striving to construct our representation of the external world, he writes, we have used a simplifying device which involves the exclusion of ourselves: "The scientist subconsciously, almost inadvertently, simplifies the problem of understanding Nature by disregarding or cutting out of the picture to be constructed, himself, his own personality, the subject of cognizance" . That is exactly the reason why the scientific picture of the world does not itself contain any ethical values; nothing about esthetics, no word of our ultimate aims and destiny . Nevertheless, this melancholic observation does not amount to an invitation to regress. Once his work is determined by opting for objectivising, that is to say, by opting to subtract from our representation of the world everything which is proper to "the subject of cognizance", the scientist must follow through. He must never again head back towards a lost paradise which Schrödinger refers to, following his reading of The Advaita Vedanta and then Schopenhauer, as the domain of a mystic unity between minds and between the One-Mind and the world. Objectivising, for Schrödinger, thus involves neither the contingency of fact nor the indifference of convention; it has the imperative character of a commandment to which Heraclitus gives voice in his fragment 2: "It is therefore necessary to follow the common. But while reason is common, the majority live as though they had a private insight of their own" .
For Schrödinger, the seemingly paradoxical reason why Hypothesis P has no place in scientific discourse is that the mode of scientific discourse is wholly supported by it; it is the background of science (in a sense close to that intended by John Searle ); it will not be seen amongst the resolutions put forward by science. The elements of Hypothesis P cannot count amongst the properties of the ob-jects of scientific knowledge, not even when those objects are human beings. Schrödinger points this out in a sentence of his 1935 article. Hypothesis P is not scientific, he insists, and that "(...) prohibits mentioning it in the very dealing with a scientific problem, in spite of , and perhaps because of the fact that science in its totality depends on that hypothesis". This exterior character of the foundation of the sciences would undoubtedly seem unacceptable to a philosopher for whom the scientific method constitutes the exclusive test of the validity of propositions, and for whom, consequently, no proposition must in principle be placed beyond the reach of that method, even, and perhaps above all, where its own premises are concerned. But Schrödinger, who holds for a long time that "physics does not reduce to atomic physics, nor science to physics, nor life to science" , is not affected by that exterior character at all. On the contrary, according to him the scientific ex-territoriality of Hypothesis P enables it, to its credit, to provide one of those signs, at once ambiguous and undeniable, of the rootedness of the sciences in a lebenswelt which always-already came before it . As Schrödinger writes in his 1935 article, he does not even think that science needs to regret the fact that one of its principal pillars rests on non-scientific ground; for it is thus that science links itself more directly with other human thoughts and aims than if it existed in and of itself.
In making his reply, Carnap does not radically doubt Schrödinger's Hypothesis P. He shares ("obviously", he would say) the anti-solipsism of his interlocutor. But his way of defending this position in 1936 shows signs of an evolution in his thinking which had taken place since the appearance in 1928 of Der logische Aufbau der Welt .
In the Aufbau, Carnap chose to adopt a position which he qualified as "methodological solipsism"; that is to say, to proceed to the construction of the world from "auto-psychological" material. Methodological solipsism was clearly distinguished throughout the work from some sort of metaphysical solipsism. One of the most acute remarks which serves to establish this distinction is that the characterisation of the basic elements of the constructive system as "auto-psychological" and as "mine" does acquire a meaning only after the domain of the non-psychological (and, to begin with, of the physical), as well as of the "you", have been constructed . To put it another way, the characterisation of elements as "mine" has no meaning other than in opposition to the things which are "theirs" and "yours" and which constitute at once one of the products of the construction and its point of departure.
In his 1936 text, on the other hand, Carnap bases his metaphysical anti-solipsism on an argument which is borrowed from the article "Physicalismus" published by Neurath in 1931, and which arises from what the latter called "social behaviourism". According to Carnap, it is legitimate to infer the possession of feelings, thoughts, memories and perceptions by someone on the basis of a "determinate exterior behaviour". That inference is just as legitimate as that which allows the derivation of the intensity of an electric current in a wire on the basis of quantities measured such as the rise in temperature of the wire or the deviation of a magnetised needle placed in the vicinity of the wire. Just as the physicist would never think of doubting the existence of the current in the wire because he could not himself be the wire and establish from the inside that by which the current manifests itself experimentally, so the psychologist must not doubt the existence of the feelings, thoughts and sensations of others because he cannot merge with others and establish from the inside that which motivates their behaviours. The proposition according to which other human beings have feelings, thoughts and sensations is in the end experimentally testable on condition that we allow only exact relational laws linking mental states with behaviours. Hypothesis P, in this modest version which Carnap favoured and labelled P1, thus does not elude empirical control. Only a strong and metaphysical version of Hypothesis P, labelled P2, which would return to postulating the impossibility of establishing a law-like relation between mental states and behaviours would remain beyond the reach of control by experience. But, Carnap adds, only the modest version P1 of Hypothesis P counts amongst the premises of scientific work. And consequently, he believes himself warranted to conclude, the type of anti-solipsistic premise demanded by science is empirically testable.
What are we to make of this debate and the arguments put forward by the two protagonists? I find it quite striking that throughout their discussion both Schrödinger and Carnap, without wishing to, occupy ground which gives too much purchase for the others' arguments; and that they thereby weaken their ability to defend their real theses.
Let us see first of all how Schrödinger puts himself in a position which gives an advantage to Carnap. As we have seen, Schrödinger shows at the beginning of his article a certain reluctance to say just what is this extra-scientific background of the scientific enterprise, the existence of which he invokes without spelling it out. But he ends by presenting the elements of a description which Carnap can afterwards rearrange, and he goes so far as to endow it with the status of an "hypothesis" or "axiom". Schrödinger, despite himself, thus puts his "fundamental hypothesis" within the sphere of the propositions of empirical science; a sphere where some propositions implicitly have the status of intangible premises and where others are in the position of being experimentally tested, but where none is intrinsically sheltered from empirical testing. Carnap is right under these circumstances to emphasise that there is no reason why Hypothesis P should in principle be exempted from empirical control. The expression of Hypothesis P, however much it is shielded from doubt, like a "hinge" on which doubt turns , cannot take advantage of any other justification but this function of being a fixed point in the network of reasoning in order to escape de facto (and not de jure) from a process of experimental testing. Even if we urge that Schrödinger has good reasons to believe that that proposition, or at least what underlies it, is truly outside the ordinary domain, the fact of having included it amongst the propositions serving as explicit premises of empirical sciences cannot but expose it to being subjected to ordinary treatment.
Reciprocally, Carnap makes enough concessions to Schrödinger to weaken his own arguments. Carnap concedes, for example, that it is necessary to rest the scientific enterprise on a belief in the perceptions and thoughts of other human beings. From then on, all his efforts consist in demonstrating the validity of an inference from that which is experimentally accessible to that which is not; from behaviours to perceptions and thoughts. Unfortunately, the inductive character of this inference leads Carnap to loose sight of the under-determination of his mentalistic explanation by the behaviours explained; that is to say, the fact that even if the mentalistic account is sufficiently corroborated it is undoubtedly not the only acceptable one. From then on he cannot see the subtlety in Schrödinger's line of thinking. The latter did not, in effect, propose to separate the mentalistic account from the domain of possible explanations of behaviour other than by comparison with a purely physiological type of explanation. Schrödinger did not deny the plausibility of the mentalistic account; he even went so far as to admit that the physiological account is less satisfactory because generally incomplete. But he did not hesitate, despite that, to exclude the mentalistic account from the class of scientifically acceptable explanations and to himself prefer the physiological account. And if he excluded mentalism, it is that he thought that the mentalistic account of behaviours was profoundly at odds with the founding aim of science, namely the objectivisation. We would say nowadays that Schrödinger had rested his affirmation of the intangibility of Hypothesis P on the pragmatico-transcendental principle which consists in making manifest the performative contradiction which arises in the practice of a science conditioned by objectivisation whenever it attempts to make use of a type of explanation of which, until better informed, all the elements cannot be considered as objectivised.
Let us now imagine that neither Carnap nor Schrödinger had made such advances into each others' territory. Suppose that Schrödinger had not given Carnap to understand that the background of science could be expressed in the form of a proposition labelled 'hypothesis' or 'axiom', and that Carnap, for his part, had not conceded to Schrödinger that science requires amongst its premises the proposition according to which other human beings have perceptions and thoughts similar to mine. Schrödinger and Carnap would then have been able to agree on this: the point of support of science is not an explicit assertion concerning what other human beings see and think; it is simply a practice of communication which anticipates or presupposes the prefect interchangeability of positions amongst the members of the linguistic community. The belief in thoughts analogous to my own for other human beings, the mentalistic vocabulary of folk-psychology, used by Carnap, as by Schrödinger, do not take first but last place in this perspective; because these devices do nothing but express after the event the confidence to which the disputants bear witness regarding a generally successful practice of communication. The difficulties only arise when we accord to this retrospective verbal expression of mutual understanding the position of the process of mutual understanding. But, we will ask, how can we do otherwise; how are we to express the primeval understanding without giving it the form of propositions by which we come to interpret it once its first step has already been taken?
Maybe in asking a poet, in this case Paul Eluard, to effect the desired mediation without compromising us with a psychologising language which comes, inevitably, too late.
"Ce n'est pas plus difficile de parler avec les oiseaux qu'avec n'importe qui sur terre: tu parles, l'oiseau fait celui qui a compris, il te répond et tu fais celle qui a compris; et tu réponds à ton tour" (It is no more difficult to speak with the birds than with anyone on Earth: you speak, the bird acts the one who has understood, it replies to you and you act the one who has understood; and you reply in turn) .
It is not here a question of comprehension, of perceptions or of thoughts, but of acting the one who has understood, who sees or who thinks. No inference from behaviour to "interior" representations, but nothing either to favour a reductionist behaviourism which would not see anything but isolated gesticulations there, instead of the play of echoes and reciprocities by which an understanding manifests itself. Picking up on a remark of J. Bouveresse's concerning the Wittgensteinian concept of a rule , we should say this: it is only for someone who takes communication as a form of recognition which is purely external, reduced to behaviour as a simple anthropological fact (human beings emitting sounds and making movements) that the ideas of significance and expression disappear. From the point of view of internal recognition, that is to say, for someone who effectively participates in verbal and gestural exchange, utterances signify and gestures express, without its being at any time indispensable to support the signification and expression on the explicit belief in their mental counterparts. In place of the difference between mentalistic and neurophysiological accounts of communication there is substituted a much more decisive difference between the engaged conduct of the speaker and a multitude of disengaged accounts variously produced by the ethnologist (behaviours), or the physiologist (neuronal processes) or even the psychologist (mental processes).
If they had recognised that, Schrödinger and Carnap would have had no difficulty in being in agreement. Schrödinger would have admitted that as soon as the dynamic of understanding translated into the form of a proposition, it finds itself projected onto the same plane of disengagement as that of the propositions of natural science, none of which is intrinsically beyond empirical control. He would have also recognised that the use of the terms 'hypothesis' or 'axiom', including when qualified as "fundamental" and when expressed with the help of a psychologising vocabulary which would make claim to being closest to the point of view of the speaker, really does nothing but manifest the partially disengaged position of he who speaks of it. What other word than 'hypothesis' must we use under these conditions? Surely not 'convention', which encompasses a dimension of intersubjective agreement, and which thus presupposes that which it is meant to express. Perhaps, then, it would be more suitable to invoke some "form of life", or better still, as Wittgenstein proposed in On Certainty, a Weltbild which might be thought of as the integrated and active totality of forms of life:
"I say Weltbild and not hypothesis, because it is the matter-of-course foundation for his research and as such also goes unmentioned"
Once this move is made, Carnap could have for his part understood without difficulty Schrödinger's insistence on placing his Hypothesis P beyond the reach of the empirical control applied to propositions of a factual form; because what Hypothesis P tries, a little clumsily, to translate is not really an hypothesis, nor an axiom, nor a proposition, but rather the pre-condition of a linguistic practice which puts into play hypotheses, axioms and every other sort of proposition.
One can sometimes get the impression, on reading the articles of Schrödinger and Carnap, that these two authors came within an inch of the pragmatic turn of which I have been making myself an advocate. Consider Schrödinger: when he says of his hypothesis not that it is simply evident or commonly accepted, but that it is "very, very, very evident"; or when he admits that if he is absolutely sure that one hypothesis is correct then he has only to build on it without concerning himself with the reasons for his certitude. He then recognises that the explicit acceptance of his Hypothesis P is indispensable neither in the course of communication nor in the practice of science. As for Carnap, he goes even further in this direction in writing in his response to Schrödinger that, without doubt, in everyday life the relational laws between behaviours and mental events are not explicitly formulated but tacitly presupposed and applied.
Carnap and Schrödinger did not, however, draw all the consequences from their remarks. It is easy to understand why. Carnap was committed to the Weltbild of logical positivism, which drove him to ignore the transcendental dimension of the pragmatic and to not attribute to it anything more than the status of empirical knowledge . Schrödinger, for his part, was a late representative of the Weltbild of German post-Kantian idealism, which included amongst its tasks the metaphysical hypostatisation of performative backgrounds. The core, at once minimal, global and universally shared, of the Weltbild which pre-conditions simple communication could not be sufficient in itself to reduce the divergence which was in place between the two historic Weltbilder in which Carnap and Schrödinger enlisted themselves.