What is idealism

Let me begin with an outline of the idealist position Ajdukiewicz targets in his paper. Ajdukiewicz addresses neo-Kantian versions, but for the contemporary reader it will be more convenient to focus on the better known Husserlian formulation. Beginning with the publication of the Ideas I Footnote 1 Husserl sought to persuade us that the so-called ‘objective’ world is in point of fact a product of cognitive activity taking place in the transcendental consciousness. This cognitive activity Husserl calls ‘constitution’, and so the claim is that the world is constituted by the pure transcendental consciousness. What exactly this pure transcendental consciousness is was never entirely clear, but in any event it is supposed to be a kind of residuum obtained by a sophisticated process of ‘purification’, starting with reflection on the totality of our mental acts and involving Husserl’s (in)famous transcendental reduction. The ‘content’ of the transcendental consciousness corresponds thus to the phenomena we find in our ‘impure’ mentality. We learn about presentations, judgments, sense-data, etc. The difference lies rather in how this content is to be interpreted: not as a ‘natural’, ‘objectified’ process taking place in the natural world, but rather as something transcendental, responsible for all objectification and a fortiori for all we can find in the natural world.

Now what does this constitution look like? Imagine you are perceiving a white horse on a green meadow. The first important observation, heard in every introductory course to phenomenology, is that what we ‘really see’ is only one side of the horse. Well, in a mundane sense of the word ‘see’, we of course see the whole horse, but in a more sophisticated, philosophically strict, phenomenological sense, its back side is always hidden from our view. What we ‘really’ see is only the front side of the animal in question. That we believe we perceive the horse in its entirety and not just one side has to do with nature of perception: in every perception quite complex operations are tacitly performed. Usually, we can move around and look at the horse from different angles. We remember what we have seen, ‘put together’ the horse’s many sides, and treat them as belonging to a single object. Further, assuming that we have a sufficiently rich fund of experience, we even don’t need to move, for we remember that other horses have many visible sides, and we assume that the horse on the meadow we are actually looking at is in all general respects very similar to all the others we have experienced.

All these operations—putting together successively inspected sides, treating remembered views as belonging to the same object, assuming that the general structure of the actually perceived object is in general very similar to the structure of the previously perceived ones—Husserl calls syntheses; and so his theory says that all composed objects of our cognition are synthesized from simpler data.Footnote 2

As all adepts of phenomenology know, at the end of the day the theory of transcendental constitution gets very complicated, but happily there is no need to disentangle all its facets in this paper. All we must know is that constitution consists in the application of certain transcendental norms,Footnote 3 which are to be understood as epistemic rules stipulating that we are justified in believing certain contents on the basis of determinate data. For example, if we experience the front side of an object, we can apply a certain rule (let’s call it ‘the three-dimensionality rule’) and obtain the belief that the object in question has other sides as well. And if we observe that the appearance of this object doesn’t change very rapidly, we can apply another rule (let’s call it ‘the substance rule’) and obtain the belief that the object in question has a certain ‘stable nature’ that doesn’t change when the object endures in time, moves around, etc. Now the idealist’s claim is that object A exists if and only if we are able to justify in this way the belief that A exists. The very existence of objects becomes in this sense a product of a coherent, systematic constitution.Footnote 4

Ajdukiewicz’s semantic reformulation

In his paper, Ajdukiewicz observes that this doctrine can be reformulated in semantic terms.Footnote 5 Idealists prefer to use mentalist language. We hear about sense-data, presentations, judgments and beliefs. But it is possible—or so at least was Ajdukiewicz’s claim—to treat all this psychological talk as referring rather to expressions of a certain ideal language than to mental episodes. And the idealists’ insistence that their analysis doesn’t concern a ‘natural’, individual mentality, but rather a ‘pure’, ‘ideal’ or ‘transcendental’ consciousness, makes this conceptual shift even easier. What is essential is that we have here a certain system of representations that behaves in many important respects like a language. The most important point is that it must involve a kind of syntax that makes it possible to formulate and apply certain epistemic principles (corresponding to the norms of constitution from the last section). If we accept the idea of such a reformulation, the idealist position begins to look like this:

  1. 1.

    Our entire system of representations (all the content of our pure consciousness) can be treated as a kind of language. In particular it involves syntax defining what it means to be a well formed formula of this language.

  2. 2.

    There are some sentences of the language in question that we accept axiomatically. They correspond to the ‘data’ which build the starting point of the whole transcendental constitution.

  3. 3.

    There are certain purely syntactic rules by means of which we can prove sentences on the basis of other sentences.

  4. 4.

    The process of constitution consists basically in applying the rules (3) to the data (2) and to other sentences, provided they have been already proved.Footnote 6

  5. 5.

    For object A to exist means that the corresponding existence claim—‘A exists’—is demonstrable in the process (4).

To avoid misunderstandings a few comments concerning the notion of ‘proof’ involved in this reformulation will be in order. It goes without saying that, in order to obtain any plausible formulation, the rules (3) cannot be restricted to (i) purely deductive principles. In addition we must have also (ii) empirical, sometimes called ‘inductive’ principles, and probably also (iii) a set of very rudimentary perception-rules of the kind ‘if it seems to me that I see that p, then I am (prima facie) justified in believing that p’. It should be also clear that the justificatory procedures employing rules (ii) and (iii) are fallible.Footnote 7 So maybe it would be more appropriate to talk here about justifiability instead of demonstrability.

Another important point is that the mentioned demonstrability or justifiability is to be construed not as a contingent epistemic position that a particular subject can have vis à vis a particular proposition at a particular time, but rather as something like an ‘ideal’ justifiability: justifiability for an absolutely rational subject in an ideal epistemic situation ‘in the long run’. In fact, idealists often referred explicitly to an infinite chain of coherent constitution, which had surely to do with the fallibility of many of the principles involved. In particular, Husserl stressed that the existence of a physical thing can be only understood as a Kantian idea—a correlate of an infinite process of coherent constitution.Footnote 8 It seems thus that an Ajdukiewicz-style reformulation of the real idealist doctrines should in any event allow for infinitist methods of proof.Footnote 9

Application of Tarski’s results

Rephrasing traditional philosophical puzzles in linguistic terms is a technique that has often been used in the tradition of analytic philosophy. In general, analytic philosophers tend to believe that such a reformulation is a good way to gain clarity.Footnote 10 But for Ajdukiewicz the reformulation of the idealist doctrine as a thesis about a language-like representational system was of a special importance. His idea was to directly apply the results obtained by Alfred Tarski, and, as is well known, they were formulated for languages of a certain kind. An important result of Tarski’s analysis of the concept of truth was that, for any language that is rich enough to contain elementary arithmetic, the concept of truth cannot be equated with the concept of provability (see Tarski 1933 and, less technical, Tarski 1969).

The easiest way to see how this works is to refer to Gödel’s famous example used in his demonstration of the incompleteness of arithmetic (see Gödel 1931). Put informally, what Gödel did was to show that in the language of arithmetic one is able to construct a sentence saying of itself that it cannot be proved. The crucial point is that the concept of proof as a purely syntactical conceptFootnote 11 can be expressed in the same language provided the language in question is rich enough to contain elementary arithmetic.

So assume we have before us a sentence saying of itself that it cannot be proved. What should we do with it? Its falsity would mean that it is provable (since it claims precisely that it isn’t), and it is not hard to see that this possibility is not very tempting. The consequence being that we would have in our system provable falsities and this is definitely the last thing we wish. It seems thus that the only sensible option is to claim that Gödel’s sentence is true, which in turn means that there are certain non-provable truths and therefore the concepts of truth and that of provability are two quite different concepts—two concepts that are not even materially equivalent (i.e. not even co-extensional).

Now let’s go back to Tarski. His most famous thesis was that the concept of truth for a (reasonably rich) language L cannot be expressed in the language L itself. What we need is a meta-language L’ with considerably richer resources. This meta-language should contain three parts:

  1. (i)

    a ‘syntactic’ part that allows us to refer to the expressions of L (involving e.g. structural names of L’s sentences);

  2. (ii)

    a part by means of which we can speak about the objects belonging to L’s universe of discourse (in the simplest case L’ can be thought of as containing the whole of L as its part); and finally

  3. (iii)

    the semantic vocabulary expressing concepts like truth, reference etc.

With these resources at hand we can build ‘a formally correct and materially adequate definition of truth’. According to Tarski, a materially adequate definition of truth—i.e. a definition that does justice to the broadly Aristotelian, realist intuition that the truth of a sentence consists in a certain relation to reality—must, for any sentence ‘p’ generate a corresponding T-sentenceFootnote 12 of the following form:

(T):

p’ is true if and only if p.Footnote 13

And the most important condition of its formal correctness is the requirement that the semantic predicates like ‘is true’, that can be applied to the sentences of L, cannot belong to the same language L. Otherwise we would be able to construct the (in)famous liar sentence:

(*):

The sentence (*) is not true,

and our theory would end in paradoxes.

Now what are consequences of this for the problem of transcendental idealism? As we have seen in section“Ajdukiewicz’s semantic reformulation” an idealist claims that the whole content of the real world is defined by the rules of epistemic justification. We know that the set of rules she takes into consideration goes far beyond the deductive rules of proof Tarski and Gödel were concerned with, but the crucial point is that all these rules are—exactly like the deductive ones—purely syntactic. They refer only to the syntactic structure of representations, not to their semantic properties. In this sense an idealist is trying to define truth in purely syntactic terms which, as Tarski has shown, is bound to fail.Footnote 14

In a later paper from 1948, entitled “Epistemology and Semiotics,” Ajdukiewicz takes a slightly different route. In this paper he targets Berkeley’s claim ‘esse est percipi’ and argues that it amounts to operating with a meta-language of syntax deprived of any means of referring to the objects belonging to the universe of the object-language. Also here Ajdukiewicz begins with translating the original mentalist language into the statements about a language-like representational system. An idealist who, like Berkley, defines things as ‘clusters of ideas’ equates them in fact with complexes of representations described purely syntactically. Recalling the three parts that have to be contained in Tarski’s meta-language, namely (i) a ‘syntactic’ part referring to the expressions of the object-language L, (ii) a part by means of which we can speak about the objects belonging to L’s universe of discourse, and (iii) the semantic vocabulary expressing concepts like truth, reference, etc., we see that Berkeley’s conceptual resources are clearly restricted to (i). The outcome of this analysis is similar to the conclusion of Ajdukiewicz’s earlier paper. An idealist who restricts herself to a meta-language of syntax is bound to understand truth and reference in purely syntactic terms, which—as we have seen—is not a very good idea.Footnote 15

Epistemic theory of truth and conceptual empiricism

Well, this result is doubtless interesting on its own, but does it also have any systematic value beyond just one more, even if particularly impressive, refutation of idealism? If it were no more than this, then Ajdukiewicz’s paper would be at most historically interesting, for, to be honest, when was the last time you came across an idealist? If we look at a philosopher’s explicit declarations, then idealism as a philosophical option doesn’t seem to be very popular.

But fortunately for the actuality of Ajdukiewicz’s analysis (although maybe unfortunately for the reputation of philosophy as a serious intellectual activity), it turns out that there are still quite a few active idealists among us. The trick is that they seldom call themselves by their true name, using instead various misleading labels, such as ‘antirealism’ or ‘pragmatism’. What they all have in common is a strong tendency to define truth in epistemic terms and this is, as we have seen, the main disease of idealism as Ajdukiewicz saw it.

Of course the sole fact that a philosopher has some problems with the Aristotelian understanding of truth doesn’t by itself imply that he or she is an idealist. In fact, there have been many philosophers who expressed reservations about this construal of truth and some, like e.g. Franz Brentano, did not evince the slightest idealist tendency. It must also be conceded that before Tarski’s works excellent reasons were available for regarding the Aristotelian idea with suspicion. Before Tarski succeeded in demonstrating how the realist, Aristotelian concept of truth could be coherently explicated it was very easy to think that it is irreparably paradoxical and should be abandoned.Footnote 16

Nonetheless typical arguments against the Aristotelian view involved something more than a hint of the logical predicament illustrated by the liar sentences. Normally they have revolved around the idea that a realist concept of truth unsupplemented by some epistemic rules of application, simply couldn’t function as an element of our cognitive system. Franz Brentano also argued in this way (cf. Chrudzimski 2001a: 71ff), but here let me cite two contemporary examples. In his book Pragmatism: An Open Question Hilary Putnam writes:

To say that truth is ‘correspondence to reality’ is not false but empty, as long as nothing is said about what the ‘correspondence’ is. If the ‘correspondence’ is supposed to be utterly independent of the ways in which we confirm the assertions we make (so that it is conceived to be possible that what is true is utterly different from what we are warranted in taking to be true, not just in some cases but in all cases), then the ‘correspondence’ is an occult one, and our supposed grasp of it is also occult. (Putnam 1995: 10)

In a similar vein Michael Dummett argues that:

the notion of truth, when it is introduced, must be explained, in some manner, in terms of our capacity to recognize statements as true, and not in terms of a condition which transcends human capacities. (Dummett 1976: 116)

The assumption behind these declarations seems to be that a grasp of any concept essentially involves a grasp of some criteria of its application. Consequently a concept totally devoid of such criteria would be essentially ungraspable, which is rather bad news, since it seems that a concept that is essentially ungraspable doesn’t deserve the name of concept in the first place.

But what could be the reason to think that every concept indeed necessarily involves such epistemic criteria of application? It can be conceded that this claim—let’s call it the criterial theory of concepts—has a certain initial plausibility. For many concepts like ‘being a horse’, ‘having a headache’, or ‘enjoying football’ there are without doubt such criteria, and we tend to assume that for all other concepts the situation is similar. But when confronted with the realist concept of truth, doubts arise. After all, Aristotle and Tarski seem to have been quite sensible and reflective human beings; both believed themselves to understand this concept perfectly. Is it really wise to think of them as being deeply confused—to treat them as victims of a sophisticated conceptual illusion, instead of simply restricting the scope of the criterial theory of concepts? Why can we not say that certain concepts (amongst them the realist concept of truth) just don’t involve any criteria of application? Is there any independent justification for an unrestricted reading of the criterial theory of concepts?

It seems that the further assumption, one that plays an important role in generating the epistemic approach to truth we are talking about, is something we can call an unrestricted concept empiricism. The claim is that all our concepts ‘derive from experience’, a claim we find both in Aristotle and in the works of British empiricists like Locke and Hume. Franz Brentano defended it vigorously against Kant’s apriorism, and also Husserl’s theory of constitution is a (rather sophisticated) version of this view (see Chrudzimski 1999).

According to the traditional version of this view our concepts can be acquired only in the following ways: (i) a concept can be either ‘abstracted’ from the direct data of experience or (ii) it can be ‘constructed’ from the elements previously abstracted in the above mentioned way. Sometimes it is also assumed that (iii) conceptual elements can be further idealized in order to obtain such concepts as material point, frictionless movement or ideal vacuum. Beyond these procedures of abstraction, combination, and idealization there is no other way by means of which a human being can acquire a concept.

This position seems to be perfectly consistent for many of our ordinary concepts. The catastrophe happens when a concept empiricist wants to be so consistent as to extend his empirical treatment to the concept of truth as well; and this is precisely what many philosophers do.

Imagine that we acquire our concept of truth in one of the aforementioned ways. In this case there must be some kind of experience in which truth presents itself and from which the concept of truth can be abstracted; or there must be at least some experiences from which we can abstract the essential elements of this concept that could be thereafter (possibly after some process of idealization) mentally ‘put together’ to build our concept of truth.

The reason why a concept empiricist typically finds the realist definition of truth non-intelligible becomes clear when we ask ourselves what kinds of experiences must be postulated to satisfy this requirement. Recall that the realist definition introduces a relation between a judgement (or belief) and some parts or aspects of reality. If my judgement that a cat is on the mat is true, then there must be in the external mind-independent world a cat on the mat. We have here a kind of comparison between a mental state (judgement, belief) and some pieces of reality (structured objects, tropes, state of affairs or something else, depending on our favourite theory of truthmaking).

Now according to the thesis of concept empiricism we could conceptually grasp such a relation only if we claim to have an experience from which we are able to abstract both of its terms. With the cognitive access to our mental states there will be in general no problem. Most idealists have been Cartesians in that they believed that we are able to reach our mental states cognitively in a particularly direct and epistemically secure way. But what about the ‘pieces of reality’ that are supposed to constitute the second term of the adaequatio relation? The sad truth is that we can access them cognitively only ‘through’ a mental state, namely through a corresponding justified belief. Indeed, to have an experience that a cat is on the mat is nothing over and above having a justified belief that a cat is on the mat (where the justification involved has a particular ‘perceptual’ character).

The crucial point here is that a justified belief is essentially (and not merely contingently) the only way in which we can be cognitively acquainted with a piece of the mind-independent world; and this means that within the framework of concept empiricism we are principally unable to separate the relevant pieces of reality from the mental states in which they are given to us. What we are able to abstract from our experiences in which ‘the truth reveals itself to us’ is thus only the concept of a justified belief. The best we can do is therefore to compare our beliefs with other properly justified beliefs and the only concept of truth we can have is consequently the concept of an idealized justifiability. This is, in a nutshell, a typical argument leading to the epistemological construal of truth, and we end with the claim that:

truth is an idealization of rational acceptability. We speak as if there were such things as epistemically ideal conditions, and we call a statement ‘true’ if it would be justified under such conditions (Putnam 1981: 55).

There are various possible reactions to this kind of argument:

  1. 1.

    We can accept it as it stands and abandon the very idea of the realist concept of truth, as something that is ‘for conceptual reasons’ principally ungraspable. The natural outcome of this reaction will be some version of the epistemic definition of truth. This was e.g. the way of Franz Brentano.

  2. 2.

    We can try to reconcile the realist construal of truth with the thesis of concept empiricism by finding (or inventing) a kind of experience on the basis of which the concept of adaequatio can be abstracted or constructed. This was Anton Marty’s reaction (cf. Chrudzimski 2001c).

  3. 3.

    Finally, we can interpret the above argumentation as a reductio ad absurdum of its premises and claim that the premise which has to be rejected is the very thesis of concept empiricism (at least in its general, unrestricted form).

Tarski and Kant

I happen to think that the last reaction is the right one. It leads us to a position that I want to term concept Kantianism. A partisan of this position is committed to the claim that not all of our concepts ‘derive from experience’ and the most important of those non-empirical concepts is the realist concept of truth (cf. also Chrudzimski 2001b).

As we remember, one of the central features of Tarski’s analysis was that a formally correct definition of truth cannot be expressed in the object language L. What we need is an essentially stronger meta-language L’. An important consequence of this treatment is that there could be no definition of truth simpliciter. We can speak only of a definition of truth for a given language L, which can only be formulated within the framework of a stronger meta-language L’. In order to formulate a corresponding definition of truth for L’ we have to move one level higher, to a still stronger meta-meta-language L’’, to speak of truth in L’’ we have to introduce L’’’, etc.

These limitations can be regarded as a very good explication of the difficulties that all concept empiricists have with the classical concept of truth. What is to be expected is namely that the very idea of splitting our representational scheme into an open-ended hierarchy of meta-languages, which is required by any Tarski-style explication, is something which no concept empiricist could be very happy with. If all our concepts are abstracted from some homogenous level of the basic data, then it seems to be principally inexplicable how a meta-language (which according to concept empiricist principles probably has to be construed as a product of a certain higher-order abstraction) can ever acquire conceptual resources that are significantly stronger than its abstraction basis. What we would expect is rather that a concept empiricist will tend to understand the totality of our conceptual resources as one homogenous ‘language of thought’.

Nonetheless, the work of Tarski seems to show that our conceptual scheme at least can be regimented this way, so that the Aristotelian idea can be expressed in a perfectly intelligible form. Since Tarski, there is therefore strong evidence for the concept Kantianism relative to the concept of truth, and the burden of proof falls definitely on the partisans of the unrestricted concept empiricism.

It seems therefore that we have good reasons to regard the realist concept of truth as a ‘Kantian’ concept in the sense that it doesn’t seem to be derivable from experience. If we grasp this concept at all – and Tarski’s brilliant analysis suggests strongly that we indeed do—then it must be a priori in Kant’s sense. But this is not the end of the story, because the realist concept of truth deserves the name of a ‘Kantian concept’ also in another, possibly even more important, sense. It looks namely as if a grasp of the realist concept of truth constitutes—to speak with Kant—a kind of a formal condition of any objectual reference.

Imagine for a moment that we are happy with an epistemic construal of truth. In this case the whole business of ‘representing the world’ would consist in having justified beliefs; and as justification is a purely syntactic notion it would be purely intra-linguistic (or intra-conceptual) affair. There would be no room for the idea that our representations can be ‘not self-sufficient’, that they can ‘point outside’, ‘transcend themselves’—in a word: there would be no hint that our representations are intentional. To get intentionality, we must grasp the idea that our representations are ‘about something’, that they are meant to represent something beyond themselves—and this is nothing other than a rudimentary grasp of the realist concept of truth.Footnote 17

To be sure, even then having justified beliefs will remain the only attainable epistemological goal. Empiricists are surely right when they stress that beyond trying our best in justifying our beliefs we have no other way to compare them with reality. But the difference is that, after having grasped the idea of such a comparison, we understand why we are interested in epistemic justification. We seek it not for its own sake, but because it is the only sign of truth we can have. Justification has only instrumental value. What we are really pursuing is truth.