Karl Popper was an important critic of materialist theories of the mind. His most significant and original criticism is an argument against the possibility of a causal theory of intentionality -- an argument I discuss at length in my recent paper “Hayek, Popper, and the Causal Theory of the Mind.” But Popper also put forward, albeit sketchily, an argument that implies the impossibility of a computational theory of the mind in particular. The argument is presented in The Self and Its Brain, a book he co-wrote with neuroscientist John Eccles. It foreshadows arguments later presented by John Searle and by proponents of what has come to be known as the “argument from reason,” such as Victor Reppert and William Hasker.
As I note in my recent paper (and had reason to note in an earlier post), Popper distinguishes four major functions of language. There is, first of all, the expressive function, which involves “an outward expression of an inner state” (The Self and Its Brain, p. 58). Here language operates in a way comparable to the sound an engine makes when it is revved up, or an animal’s cry when in pain. The second, signaling function adds to the expressive function the generation of a reaction in others. Popper compares it to the danger signals an animal might send out in order to alert other animals, and to the way a traffic light signals the possible presence of cars even when there are none about. The difference between the expressive and signaling functions would seem to parallel Fred Dretske’s distinction between “natural meaning” (or meaningn) and “functional meaning” (or meaningf), which I discussed some time back in a post on Dretske. Meaningn or “natural meaning,” it will be recalled, amounts to nothing more than an effect’s indicating the presence of its cause, as spots on the face indicate the presence of measles. There is no possibility of misrepresentation here, since an effect will meann whatever it is that happens to cause it. Hence if the spots on someone’s face were caused, not by measles but instead by an allergic reaction of some sort, then that, rather than measles, is what they will meann. Popper’s “expressive function” seems more or less the same insofar as he appears to think that an effect (the sound of the engine, the animal’s cry of pain, or someone’s angry and spontaneous utterance of the appropriate expletive when stepping in something at the dog park) will “express” whatever inward state it is that happens to cause it. The possibility of misrepresentation only clearly enters the picture with the “signaling function,” just as it does (at least if Dretske’s account succeeds) with meaningf or “functional meaning.” An internal state or utterance might meanf that such-and-such is present even when it is not; similarly, it might in Popper’s sense “signal” the presence of something (predators, cars, or the headache your wife claims she is having) even when that something is not really there.
Popper allows that these two elementary functions of language might be explicable in causal terms. What he regards as inexplicable in such terms are the remaining two functions. The descriptive function of language involves the expression of a proposition, something that can be either true or false. The paradigm here would be the utterance of a declarative sentence, such as “Roses are red,” “Two and two make four,” or “There is a predator in the area.” Notice that the latter example differs from an animal’s cry of warning in having a conceptual structure. A bird’s squawk might cause another bird to feel fear and take flight. What it does not do is convey an abstract concept like eagle, predator, or danger, and thus it does not convey the sort of propositional content that presupposes such concepts. (Popper tentatively allows at p. 58 of The Self and its Brain that at least some animal behavior “may perhaps” involve a descriptive component and not mere signaling, giving the bee’s dance as a possible example. I don’t find this plausible myself, but nothing in what follows rides on the issue.) Finally, the argumentative function of language involves the expression of an inference from one or more propositions to another in a manner than can be said to be either valid or invalid, as when we reason from All men are mortal and Socrates is a man to the conclusion that Socrates is mortal.
It is Popper’s treatment of the “descriptive function” of language that indicates what he takes to be problematic about the notion of a causal theory of intentionality. Again, I examine his argument against the possibility of such a theory in detail in “Hayek, Popper, and the Causal Theory of the Mind.” It is in Popper’s treatment of the “argumentative function” that we find his implicit objection to computational theories of the mind. (See The Self and Its Brain, pp. 75-81.) Like the descriptive function, the argumentative function is something that in Popper’s view cannot be accounted for in causal terms, and he gives a separate argument to this effect. Though he does not claim that this argument strictly refutes materialism, he says that it shows “that materialism has no right to claim that it can be supported by rational argument”; in particular, it shows that materialism, even if it were true, “is incompatible with… the acceptance of the standards of critical argument” insofar as “these standards appear from the materialist point of view as an illusion, or at least as an ideology” (p. 81). The nerve of Popper’s argument is contained in the following passage:
The property of a brain mechanism or a computer mechanism which makes it work according to the standards of logic is not a purely physical property, although I am very ready to admit that it is in some sense connected with, or based upon, physical properties. For two computers may physically differ as much as you like, yet they may both operate according to the same standards of logic. And vice versa; they may differ physically as little as you may specify, yet this difference may be so amplified that the one may operate according to the standards of logic, but not the other. This seems to show that the standards of logic are not physical properties. (The same holds, incidentally, for practically all relevant properties of a computer qua computer.) (p. 79)
Unfortunately, while this is suggestive, Popper does not develop the argument in a formal way; the passage quoted is taken from an imagined dialogue between a “Physicalist” and an “Interactionist,” and Popper lets the responses of the latter stand in for an explicit formulation. But the overall thrust of the argument can be reconstructed by comparison with some clearly related ideas to be found in the work of John Searle, on the one hand, and the work of proponents of what has been called the “argument from reason” on the other.
The context makes it evident that Popper intends to make both a narrow point against any attempt to explain human rationality specifically on the model of the modern digital computer, and a more general but related point against any materialist attempt to explain rationality in causal terms. It is with respect to the former point that we find a clear parallel with Searle. In the passage quoted, Popper says that “practically all relevant properties of a computer qua computer… are not physical properties.” This may seem odd given that he also allows that “the property of… a computer mechanism which makes it work according to the standards of logic is… in some sense connected with, or based upon, physical properties.” But Popper also points out that the reason a computer operates according to logical principles is that it “has been designed by us – by human minds – to work like this” (p. 76). Its operations mirror the semantic features of linguistic symbols and their logical relationships, just as the words written in ink on a piece of paper do; but the semantics and the logical relationships are no more inherent to the physical properties in the case of the computer than they are in the case of the ink marks. In both cases they are imparted to the physical phenomena by us – by programmers and users in the case of computers, and by writers and readers in the case of written words – rather than derived from the physical phenomena. Hence they can hardly provide a model of how rational thought processes might be explained in purely physical terms.
Searle’s version of this line of argument emphasizes that the key notions of the modern theory of computation – “symbol manipulation,” “syntactical rules,” “information processing,” and the like – are not definable in terms of the properties attributed to material systems by physical science, but are observer-relative, existing in a physical system only insofar as some interpreting mind attributes computational properties to it. Hence the very idea that the mind might be explained in terms of computation is incoherent. The argument can be summarized as follows:
1. Computation involves symbol manipulation according to syntactical rules.
2. But syntax and symbols are not definable in terms of the physics of a system.
3. So computation is not intrinsic to the physics of a system, but assigned to it by an observer.
4. So the brain cannot coherently be said to be intrinsically a digital computer.
(Searle develops this argument in his paper “Is the Brain a Digital Computer?” and in chapter 9 of his book The Rediscovery of the Mind. Note that this argument is different from Searle’s better known “Chinese Room” argument.)
There is a clear parallel between this Popper-Searle argument against a computational theory of rationality and Popper’s argument against causal theories of intentionality (which I examine in the paper linked to above). In both cases, the materialist or physicalist is accused of making use of notions (certain causal notions in the one case, computational ones in the other) to which he is not entitled given his working assumption that the only genuine features of reality are those describable in the language of physical science.
It is in his application of this basic idea to a critique of any possible causal account of rationality that Popper’s position resembles the anti-materialist “argument from reason.” This is a label that has recently come to be applied to a family of related arguments to be found in the work of thinkers as diverse as Popper, J. B. S. Haldane (whom Popper cites as an influence), C. S. Lewis, Alvin Plantinga, Victor Reppert, and William Hasker. (For a useful overview, see Reppert’s article “The Argument from Reason” in William Lane Craig and J. P. Moreland, eds., The Blackwell Companion to Natural Theology. I discuss and defend the argument in chapter 6 of Philosophy of Mind.) There are significant differences between these writers’ respective statements of the argument, but a “generic” version might go as follows:
1. Materialism holds that thinking consists of nothing more than the transition from one material process in the brain to another in accordance with causal laws (whether these transitions are conceived of in terms of the processing of symbols according to the rules of an algorithm à la computationalism, or on some other model).
2. Material processes have their causal efficacy, including their ability to generate other material processes, only by virtue of their physical properties (i.e. those described by physical science), and not by virtue of any meaning or semantic content that might be associated with them. (For example, punching the symbols “1,” “+,” “1,” and “=” into a calculator will generate the further symbol “2” whether or not we associate the standard arithmetical meanings with these symbols or instead assign to them some eccentric meanings, because the electronic properties of the calculator alone are what determine what symbols get displayed. Similarly, neural processes that are in fact associated with the thought that all men are mortal and the thought that Socrates is a man would still generate the neural process that is in fact associated with the thought that Socrates is mortal even if these neural processes had all been associated with some other meanings instead, because the neurophysiological properties of the processes alone are what determine which further processes get generated.)
3. But one thought can serve as a rational justification of another thought only by virtue of the meaning or semantic content of the thoughts. (For example, it is only because we associate the symbols “1,” “+,” “1,” “=,” and “2” with the standard meanings that “1 + 1 = 2” expresses an arithmetical truth. Similarly, it is only because “All men are mortal,” “Socrates is a man,” and “Socrates is mortal” have the meanings they do that the first two sentences logically entail the third, and only when the neural processes in question are associated with the corresponding thoughts that the first two provide a rational justification for believing the third.)
4. So if materialism is true, then there is nothing about our thought processes that can make one thought a rational justification of another; for their physical and causal relations alone, and not their semantic and logical relations, determine which thought follows which.
5. So if materialism is true, none of our thoughts ever is rationally justified.
6. But this includes the thoughts of materialists themselves.
7. So if materialism is true, then it cannot be rationally justified; the theory undermines itself.
The upshot of this argument is that instantiating causal relations, of whatever sort, does not by itself amount to instantiating logical relations; and this is precisely what Popper is getting at in the passage above when he says that “brain mechanisms” or “computer mechanisms” may “differ physically as little as you may specify, yet this difference may be so amplified that the one may operate according to the standards of logic, but not the other.” Hence even if we concede that certain causal processes are necessary conditions for our reasoning logically (which Popper allows insofar as he says that our ability to follow standards of logic is “in some sense connected with, or based upon, physical properties”), they are not sufficient conditions – in which case there can be no (purely) causal explanation of our ability to reason logically.
Step 2 of the argument seems to follow from the standard materialist assumption that whatever happens in the natural world supervenes on what happens at the microphysical level of nature – the level of the basic particles described by physics and the laws governing them – together with the further materialist assumption that meaning or semantic content is not a microphysical property, whatever else the materialist wants to say about it. That this appears to make the meanings of our thoughts “epiphenomenal” or causally irrelevant to what happens in the world is known as “the problem of mental causation.” Of course, the meanings of our thoughts seem to have an effect on what we say and do; in particular, it certainly seems to us that we judge an inference like All men are mortal and Socrates is a man, therefore Socrates is mortal to be rational because of the meanings associated with these words, and would not judge it to be rational if they had some different content. But Popper’s point is that, if materialism is true, then we can have no grounds for believing that what seems to be the case really is the case. Perhaps the inference in question is in fact irrational, while an inference that seems irrational to us, like All men are mortal, and Grandma drives a Buick, therefore robots are stealing my luggage is a paradigm of rational thinking. Perhaps we don’t see this for the same reason the calculator would spit back “2” in response to the sequence “1 +1 =” even if the latter set of symbols expressed the question Does Grandma drive a Buick? and the former expressed the bizarre answer No, robots are stealing my luggage – namely for the reason that only the physical properties of events occurring in both calculators and brains, and not any semantic or logical properties associated with them, determine what effects they will generate.
For this reason Popper claims that materialism tends to reduce the argumentative function of language no less than the descriptive function to the sub-rational expressive and signaling functions, and thereby tends also to “make us blind to the difference between propaganda, verbal intimidation, and rational argument” (The Self and Its Brain, p. 59). Now Popper presumably thought that his friend F. A. Hayek’s account of the mind was open to this sort of criticism, just as it was open to the criticism I discuss in “Hayek, Popper, and the Causal Theory of the Mind.” There is irony in this, for Hayek himself accuses others of what Popper implicitly accuses him. Hayek argued in The Counter-Revolution of Science, an important critique of scientism, that “the ground for a thorough irrationalism” lay implicit in any view of human beings aimed at “uncovering hidden causes which, unknown to the thinker, have determined his conclusions.” (p. 159). His target was the relativist idea that a person’s race or class situation determines what he thinks. Popper’s claim is that the materialist view that our thoughts are determined by the hidden causal processes uncovered by physical science is no less implicitly irrationalist.