It is not inessential for the success of a scientific investigation that questions which can be treated independent of others not be mixed with these. (Frege 1897/1983: 157)

‘But it must be like that’ is not a proposition of philosophy. (Wittgenstein, P.I.§599)

On 21 September 1931 Wittgenstein remarked to Waismann:

I believe that understanding is not at all a particular psychical process which remains external to, and attaches to, perception of sentential structure. If I hear, or read, a Satz, all sorts of different processes occur in me. … But such things are not what is of interest here. I understand the Satz in applying it. Understanding is thus not at all a particular occurrence, but rather it is so operating with the Satz. The Satz is that with which we operate. (1984: p. 167, italics Wittgenstein’s (or Waismann’s), bold mine.)

Inspiration for this remark is, plausibly, to be found in Frege 1903, notably §§91–92, which, on new year’s day of that year (by Waismann’s notes) the two were reading jointly.

The specific phenomenon Wittgenstein speaks of here is understanding what is said, either in the success sense (understanding tout court) or on its neutral reading, understanding as saying such-and-such). The object of the understanding thus belongs to the phenomenon of thought-expression. I have left ‘Satz’ untranslated. In philosophy, it is at least as waffly as the corresponding English ‘proposition’. In each case philosophical and ordinary use diverge considerably. (In a central ordinary use, ‘to proposition’, was once, is perhaps anew, a route to a Sittlichkeitsdelikt.) In philosophy the term needs dewaffling (soon to come). To adumbrate, to understand Wittgenstein, and the phenomena we will here need two different notions. On one, a Satz is a truth-bearer, but cannot bear content. On the other it does both.

The gist of the above remark, or that to be expanded here, is that, at least in philosophy, when it comes to understanding understanding, what matters is not by whatever inner workings such is enabled, but rather what it is that is enabled; in what ways understanding relates the understander to others and to the world (to our environment, more generally, the way things are). It recommends to us a perspective from which accounts of what goes on inside us when we understand are neither to be expected, nor their absence missed.

I will set Wittgenstein’s remark alongside a remark of John McDowell’s:

We can think of judgments as inner analogues to assertions. … In an assertion one makes something discursively explicit. And the idea of making things explicit extends without strain to judging. We can say that one makes what one judges explicit to oneself. (2008: 6)

For all the above Wittgenstein need not deny that there may sometimes be inner acts or events of making something explicit to oneself. Nor need we. But he does say of such things that they are not ‘what interests us’, not, specifically, where it is a matter of understanding understanding (what was said, or what was said as that such-and-such). Might understanding really be so different from ‘judging’-that?

The general phenomenon here is that of representing things as thus and so (‘things’ here mass, not plural, i.e., questions ‘Which things?’ out of order.) Such representing is thinker’s business on one aspect of the verb, that on which in representing things as thus and so the representer may commit. Given that aspect there is another in which a thinkable is what represents (without commitment, since only a thinker can commit). The first aspect is primary topic of the present investigation.

In this aspect representing comes in two forms: auto- and allo- (saying that, and taking that). Wittgenstein’s remark concerns understanding whose object is allorepresenting (what, if successful, bears an understanding). But it is meant to contribute to understanding understanding. And the object of this iteration is autorepresenting, whose objects are, inter alia, understandings to be borne. It is autorepresenting which invites the relevant psychologism malgré soi, and thus will be the central object of investigation here. Allorepresenting provides a model by which to deactivate the urge.

In autorepresenting one represents to oneself as to no other. One need do nothing to inform oneself that one so represents. To contrast with allorepresenting, one might call autorepresenting ‘inner’ in just this sense. But to be eligible so to engage there are prerequisites to be met. To so much as think lamprey to spawn in the Minho, one must be aware of there being such a way for things to be: of lamprey as potential spawners, of spawning as something for them to do, and so on. Here we have something which, on its face, needs to be enabled. And the enabling called for might seem to be inner in that second sense of the term. (As McDowell’s posit of inner ‘explicit-makings’ is inner on the first.) In both cases, a philosopher might be tempted to posit inner works or workings. Where these are uncalled for by the explanandum (or it is indeterminate what the explanandum is to be understood to be) I will speak of a myth which, borrowing from Jacques Bouveresse, I will call ‘the myth of interiority’.


1. The Minimax: The minimax arises in the business of representing-as as the original object of the verb. This business itself comes as a package with the notion true. The key feature distinguishing such representing from other things which bear the name is that it is non-factive. It thus contrasts with, e.g., that notion on which Sid’s shiner may represent a collision with a door. On this notion, the shiner does so only if there were such collision. By contrast, Pia may say that Sid collided with a door just as easily (in point of concept) if the truth is that for once Sid overdid the irony. Thus room for falsehood, thus room for truth.

Frege’s search for the variables over which laws of logic (for him laws of being true) ranged leads us to the notion minimax truth-bearer. (Though all we really need for this is the idea of a truth-or-falsehood: what, if true, is the truth, if false, the falsehood. E.g., if it is true that Sid sports a shiner, the truth is that Sid sports a shiner, if false then that Sid sports a shiner is a falsehood.)

Suppose we are in search of the laws of being true, that is, what holds of the minimax simply by virtue of what being true is as such. Plausibly, the object of our search must be both topic- and thinker-neutral. As Frege puts it, (1893: xvii) such law must govern all thought, no matter when, where, or by whom (or as to what). Thinker-neutrality is provided for in Frege’s first move, which is to detach the phenomenon of being true from that of holding true, that is, from any interaction with it in thinkers’ thought). To arrive at topic-neutrality, all proper thought-parts must be stripped of any topic-specific content, that is, what distinguishes any such part from any other of the same logico-syntactic type, e.g., a one-place predicative part. These two moves made, what we arrive at for logic’s laws to govern are specific forms for a thinkable (a given truth-bearer) to assume (logical forms).

The question now becomes in what ways such forms can be filled out into full-fledged truth-or-falsehoods; a question to which logic expressly does not speak. One requirement on filling-out, though, is this: What distinguishes any given such thing from any other can be nothing other than a feature which bears directly on just where this truth-bearer would be true (what would so count), where false. (Frege: A thought (Gedanke) is just that by which a question of truth arises; just that, no more, no less.

Thus, such a truth-bearer cannot also be a content-bearer. That is, nothing else can bear on what content is its. For what makes it that one is nothing other than how it represents things as being. So one cannot intelligibly first identify it and then ask what content it is to be understood to bear. Nor can a minimax truth-bearer have any distinguishing features which presuppose a spatiotemporal career, since it cannot have locations or velocities or trajectories. Hence (whether embarrassment or not) it cannot be intrinsic to a minimax truth-bearer, on any decomposition of it into proper parts, that it contains indexical or demonstrative elements. It is timeless and spaceless. Thus, it is a truth-or-falsehood in above sense.

Here, then, one notion of Satz. In German it occurs in the compound ‘Lehrsatz’, e.g., the Pythagorean theorem. It is contained in the notion truth-or-falsehood. Such Sätze are thereby the presumptive objects of such forms of representing-as take thinkers in subject position, also which occur as steps in proofs valid or not to logic’s eye. Thus the minimax truth-bearer.

It is worth noticing here that true is an identity under predication only over the domain of the minimax. Find a truth-bearer which is identified as the one it is by (inter alia) features other than what determine of what it would be true, and identity under predication fails to hold. For example, suppose that the sentence ‘Snow is white’, or some particular occurrence of it in some discourse or conversation, is either true or false. While, perhaps, expressing ‘the thought that snow is white’ (assuming determinate reference to have been made here) and expressing ‘the thought that snow is white is true’, thus predicating being true of first-mentioned are expressing the same thought (saying the same to be so), speaking the sentence ‘Snow is white’ and speaking ‘The sentence ‘Snow is white’ is true’ is not speaking the same truth-bearer twice. (If sentences are truth-bearers, then sometimes one can speak a truth-bearer.) It seems to me that a good deal of bad philosophy has resulted from insensitivity to this point. But this is not the occasion for demonstration.

The requirement on minimax status has now been set. What fulfils this requirement remains to be decided. In what follows I will speak loosely, perhaps misleadingly, of ‘the domain of the minimax’. I do not mean to rule out thereby the possibility that what satisfies the requirement, what not, may vary with the occasion for enquiring after this.


2. Proxies: Perhaps the most salient feature of allorepresenting, thought-expression, is that it must be recognisable what thought was thus expressed for such representing to have occurred at all. ‘Recognisable’ will do here as a placeholder for whatever form or sort of recognition is ultimately called for, though what is recognisable here is recognisable to one. And if a given case of thought-expression must bear an understanding, such must be an understanding of it for one to have, where for any given ones, perhaps others.

In particular, in an expression of a given thinkable, things are represented as being thus and so. In the sense of ‘one’ just outlined, such must be a way for one to take things to be, and then the way one who understood what was expressed would take them to have been represented as being. So the means employed by the relevant expression by which to achieve recognisability must, as so employed, in fact bear such an understanding.

Thought-expression is, grammatically, a performance; an episode of a sort I call ‘unfolding’. The distinctive feature of unfolding is that, divided into proper temporal intervals, the whole need not be in every part. Parts of the performance, that is, may occur in some such intervals and not in others. Such grammar is also found, e.g., in episodes of perceptual awareness (e.g., watching a scene unfold).

So the means employed for recognition occur in an unfolding episode, different bits of means liable to occur in different intervals (on some decomposings of the episode). For example, different words of a sentence spoken may occur in different temporal parts of the episode.

The crucial point brought out by the above is that the expression of a thinkable calls for a Satz such that, on the one hand, it can be recognised as present without appeal to the understanding it bears (how it represents things as being), but which also is to be understood as so representing things. The idea is: One can recognise how, in the given performance, things were represented as being by recognising how, in the circumstances of the performance, this Satz is to be understood. It would be both a truth-bearer and a content-bearer.

Such a Satz is thus non-minimax. Patently, true is no identity over it. It has both an environmental presence—something to be experienced perceptually—and a content, that of the thinkable it expressed. Given all this, for it thus to have represented things as being thus and so would also be for its author (speaker) so to have represented things. Because of its role in thought-expression I here dub this second sort of Satz a proxy.

A proxy bears an understanding, whereas a minimax simply is an understanding for a proxy to bear. It is not eligible to be to be understood in one way or another. What, now, might bear on how a proxy is to be understood, specifically, understood to have represented things as being? Let us confine attention to linguistic cases; ones in which a proxy is a string of words from the thought-expresser’s mouth/pen/keyboard. Let us further take it as given that the words are those of a given language, and being used for what they are used for in that language.

Of course, this is already for the proxy to be to be understood in a given way. And it has proven tempting to some philosophers to suppose that modulo linguistic ambiguity, deixis, etc., such already answers our question. For, modulo all that (the idea is) a declarative sentence of a language just is the expression of a given minimax thinkable. I take as point of departure from which such a picture begins to dissolve a remark of Frege:

That which contains what most clearly points to the essence of logic is assertive force. (1915: 272)

In asserting that P, a thinker presents himself as authoritative as to whether P, and as offering to transmit the awareness he thus enjoys to an audience, or, going on record, to a potential audience. He underwrites, vouches for, taking it that P, thus for treating things accordingly where, or insofar as, it mattered whether P. He thus signs on for the consequences of things being such that P—for praise if those consequences in fact obtain in things being as they are, for blame if they do not. Not blame for things not so being, but rather for his audience being led so to treat them.

At this point, then, the business of holding true acquires a moral dimension. The speaker used given words for what they are for in their language. He thus spoke of what those words do. He said, say, ‘Lamprey spawn in the Minho in winter months.’ If there is something which would obtain were things as he said, but does not—something he is rightly held to have vouched for—he is at fault, misleading others, misrepresenting himself. But is it fair to understand him as saying what would have this as a consequence? In matters of assigning responsibility such as this, doors are thus opened to what may be very complex issues. Frege is impressed above by logic’s role in drawing consequences. But, in search of logic, Frege prescinds from all thinkers’ truck with thinkables. Whereas proxies bring precisely such truck into question. Logic’s laws, of course, govern all thought (where they apply). But where the question is what understanding a given proxy bears, just what thinkable(s) it expresses, crucial questions take a different form. If one spoke as the speaker did in those circumstances, what would one properly appreciative of that performance expect to follow from what was thus expressed? (Of course, what one would rightly suppose to follow from things being a given way might on occasion not follow. The world is liable to hold surprises. Fair distribution of praise and blame, here as elsewhere, is no simple matter.)

A simple example. Sid tells Pia that her distinguished lecture is set for 8 pm. The lecture is to be in Berlin. Pia and Sid are, at the moment, in Porto. The time in Berlin is an hour later than in Porto. Sid means Berlin time. Pia understand Porto time. So she is late, in fact, too late, for her own lecture. Who is to blame? Nothing in the words used settles this. One could speak in either way, depending on the point in doing so. There may be enough in the circumstances of Sid’s speaking for the question to have a clear determinate answer. Or perhaps the question must just remain moot. Either way, a simple illustration of the sort of question arising in such questions of fairness and justice as for what a thought-expresser is rightly held responsible.

A more complex example. Pia faces a stack of copies of her latest book which must be autographed and in the bookstore by noon. She rummages throughout the house, but cannot find a pen. Sid, from the kitchen, (as though) helpfully, calls out, ‘There is a pen on the table.’ Take the table’s identity as read. Suppose that what is on the table is an antique pen, dry for decades, requiring an ink well, non-existent in the household. Or suppose that the pen has been bronzed, or set in the middle of a large block of plexiglas, or that it has been run over by a truck and now lies on the table in a collection of many parts, neatly assembled and placed in a ziplock bag. For each of these eventualities, one may ask in what circumstances Sid would fairly have been held to have said what would be rightly understood to have consequences inconsistent with that eventuality obtaining. I remark only that such questions of justice and fairness are not ones whose answers it is for a theory of language to supply.

With which I return to Wittgenstein’s 1931 slogan: “The Satz is that with which we operate.” Applied to a Satz, i.e., proxy, we can work the slogan, in one direction, from Satz to consequence. I dub this direction left–right. Here, note, the Satz is not a thinkable. It is rather an historical performance, subject to evaluation by those with a proper sense of fairness. Speaking in those words, in those circumstances, saddles one with commitments in re the consequences of things being as said. Seeing to just what one is thus committed is exercise of such a sense. Once the commitments thus made have been adequately fixed, one can work the slogan in reverse, as it were, right-to-left. Given the consequences thus rightly to be supposed to be, one need only find a Satz, i.e., minimax thinkable, whose consequences are (as near as matters) these. These two moves in succession give us a move from proxy to thinkable expressed; from the historical content-bearer to the content it is to be understood to bear.

This section and last thus de-waffle ‘Satz’ (equally ‘proposition) so far as needed for present purpose. There is Satz-the-content-bearer, and Satz-the-content-borne. One more remark before turning to McDowell’s analogy. Satz (the minimax) plays a central role in autorepresenting. It is the object of taking-that and related verbs. Taking it that things are thus and so presupposes a capacity, namely, an adequate ability to recognise what difference it would make whether things were or not the way in question; thus to recognise what would, what not, count as things so being. Allorepresenting works via a Satz with an environmental profile, in central cases, at least, a possible object of perceptual awareness. Correspondingly, engaging in allorepresenting presupposes a capacity to make oneself recognisable as so doing. Such draws on a very different set of skills. There is prima facie no reason to suppose that such skills are also drawn on in autorepresenting. If Sid thinks that lamprey spawn in the Minho, his so doing does not presuppose an ability on his part to make his so thinking recognisable to others. It does not turn on success in this field of play. Nor does his so thinking rest on his making this recognisable to himself, on pain of regress.


3. In Words Kant Used: I turn now to McDowell’s analogy. It occurs in the course of his introducing his 2008 proposal for bridging a certain (supposed) gap between perceptual and propositional awareness. In this context he quotes Kant approvingly:

Kant says: ‘The same function which gives unity to the various Vorstellungen in a judgment also gives unity to the mere synthesis of various Vorstellungen in an Anschauung ….’ (A79/B104-5,)

Kant continues:

Thus the same Understanding, and to be sure by the same operations (Handlungen) whereby, in the case of concepts, through the analytical unity, it brings into being the logical form of a judgement also, through the synthetic unity of the multiplicity in an Anschauung as such, brings into being a transcendental content …(B106/ A80)

The first thing I insist on here is that nothing I will say in this section should be read as ascribing any view to Kant. I am only interested in a use those words suggest, on which they would express a version of the Myth of Interiority.

Note: The German ‘Urteilen’, English ‘judge’, have each acquired a philosophical use quite different from their ordinary use. ‘Judge’, on ordinary use, speaks of a process, and its historical occurrence. A useful paradigm is the colloquial ‘to eyeball’, sizing up the looks of, e.g., the road ahead, making an estimate of time to next exit. A paradigm for Urteilen, I am told, is a judge rendering a judgement. In use such as Frege’s the object of judging is an outcome of the exercise; generalising, a case of taking-that full stop. The philosophical danger is conflating these two very different grammars.

I return to the text. Here there is talk of, for each thinker (of our sort) (at a time), a Funktion which generates a given stock of thinkables, each structured out of parts drawn from some given ‘basic’ vocabulary. Kant calls these parts ‘Vorstellungen’, a term which here could only speak (idiosyncratically) of a proper part of a thinkable (what does part of what a thinkable does the whole of). The generata of such ‘Funktion’ would then be just those thinkables thus made available to its owner (at a time) to engage with in his representing-as (notably, his takings-that).

So far, ‘Φ generates θ’ need only mean: θ is derivable in Φ (i.e., in some sequence of Φ’s operations). (Though the term ‘to bring into being’ (zu Stande bringen) does suggest something else.) So far, then, there need be no hint of enabling, nor explaining such. The Funktion identifies some possible object of a capacity (for thought), some domain of items then eligible to be objects of its owner’s takings-that. If one exchanged ‘some’ for ‘the’ in this last, the claim made would be (to adumbrate) by no means innocuous. But enabling is not per se yet in the picture.

Suppose that we now decided to assign the Funktion an enabling role. Here is one idea: It makes its generata available to its owner as thinkable by him. It does so by affording him awareness of these generata, thereby awareness of those things there are to think (by awareness of those ways there are for things to be). Now we can ask, first, just what such awareness of the generata is to consist in, and then how such awareness would help gain relevant awareness (in Frege’s metaphor ‘grasp’) of those ways for things to be.

Taking Kant’s words au pied de la lettre, the generata are to be, themselves, thinkables. But this cannot be quite right.

In thinking we do not produce thoughts, we grasp them. (1918 P 74)

A thinkable is the content of a taking-that. It cannot bear one (i.e., be to be understood in one way as opposed to others). And as minimax it has no spatiotemporal profile. It is not the sort of thing a Funktion, or anything, could manipulate or modify or ‘bring into being’. Nor, equally, is a proper thought-part.

A logical calculus provides means by which to operate on logical forms. One can do so in, or by, operating on the stand-ins (Vertreter) for these which the syntax of the calculus generates (thus provides). These Vertreter, potential objects of perceptual awareness, vertreten items which are not. They thus permit operating on those non-spatiotemporal items just where they are so understood (thus not merely in being those marks that form them).

A Funktion’s generata could not be objects of perceptual awarenesss. It is not as if its owner is to find the structures it provides in his ambient environment. But perhaps the awareness it offers could share grammar with the perceptual sort, in particular this: perceptual awareness of an object—say, the pen on my desk— does not entail awareness-that. (I need not, e.g., recognise that it is a pen I see before me.) And thus a calclulus. It is pedagogic commonplace that visual awareness of a generatum of a calculus’ syntax would enlighten no further than a wallpaper pattern might unless the viewer grasped (adequately) what that generatum, and its proper parts, vertreten.

The point applies to a Funktion. It is constrained ineluctably to operate on Vertreter for proper thought-parts. A generatum structures given such parts. Awareness of such a Vertreter can afford the awareness the Funktion is now meant to enable only if the enjoyer of this awareness grasps what the Vertreter (in part and in whole) vertreten. But if he does grasp this then the generatum is otiose. For to grasp it he must already enjoy just the awareness the Function is meant to offer by cited means. The explanandum is contained in the explanans.

A Funktion can thus not be assigned such enabling work in any such way. Awareness of a generatum is either of a sort which does not entail awareness of what it vertretet (namely, such-and-such), so does not enable, hence nor explain, what it was meant to, or it does, in which case, explanans contained in explanandum. A Funktion cannot work like that.

One might now try to go all subdoxastic. In broadest outline, a generatum would now be a structured signal (metaphorically, a ‘sentence’), destined for a role in given computations whose outcome would be (roughly) preparedness to see things to be treatable as they would be were things, casu quo not, that way for things to be which the signal vertretet, depending on which of two statuses it was assigned). But if the motivation for this is no more than the idea that ‘there must be something correctly held responsible for all this—it must be produced in some way’, what we have here is just the myth of interiority incarnate—and as Investigations §599 put it, not philosophy. So far, the lure of such idea may be merely a symptom of misconstruing the phenomenon (autorepresenting).

The next section elaborates the mythical. The remainder of this essay fills in, in one way, why this is misconstrual. To hint, the idea of the Funktion presupposes there to be just one way of counting objects of taking-that, or at least those available to a thinker at a time, or at least one way which is fundamental. Further reflection on the work the idea of autorepresenting, at least taking-that, is meant to do may—I think will—show this presupposition to be on quite the wrong track.

The True Analogy: Allo- and autorepresenting, both representing-as, are no doubt analogous in multiple ways. But which ones are here relevant? In thought-expression, in particular saying-that, the proxy is an object of perceptual awareness. An intention is to be understood to attach to it: that in it things be represented as being thus and so. Where successful, one who understood it would thus be able to see how it was to be understood, thus how it represented things as being, for what the author is held responsible. The proxy’s author contrives to make this so. His authority over what was said ends there. Thus the (often realised) possibility of not saying what one meant, or meant to, and of saying what one did not.

Capacities for understanding are presupposed here. Nothing has been said, nor, so far needs to be, as to how such might be enabled. Enabling them is not the proxy’s role. An understander must, presumably, be aware of the proxy. And in coming to a particular judgement about it—e.g., whether a nib-less pen would still count as pen in the meaning of relevant act—he might refer back to the proxy for better understanding of it. But such is not per se part of the bargain here. It is mere empirical hypothesis.

Thus the one term in the analogy. For the other, I first note three crucial disanalogies between allo and auto:

1. Taking-that is a condition for one to be in. So being may be said, like influenza, to come in episodes. But then these are not unfolding ones. The whole taking-that comes in each proper interval. Thus the irrelevance of events, e.g., making something explicit to oneself. What it is to take something to be so is what is present, a constant, throughout its history..

2. Taking-that is, so to speak, ‘lutheran’: a condition imposed on one, often, as Sextus suggests, an affliction, anyway something one is saddled with. Where there was room for choice in the matter, such would be, eo ipso, not so taking things. (The Pierre Ménard option, of course, is always open.) Such involuntariness of taking-that permeates the phenomenon.

3. Thus, the commitment which belongs to taking-that is not undertaken, or signed on for, as in saying-that, but rather suffered. It, too, is a constant throughout the duration of the episode (insofar as ‘episode’ is le mot juste). Commitment here is, anyway, precisely coeval with the taking-that, the same inability (seriously) to think (though not inability to imagine) otherwise. Pace McDowell, it cannot consist in makings-explicit.

Suppose we want an inner analogue to the proxy: a content-bearer, not itself a content. The story of the ‘Funktion’ provides what fits the bill. The Funktion generates Vertreter for thinkables. (Thinkables, unlike their Vertreter, are not literally to be operated on, nor brought into existence.) Such Vertreter might be thought of as objects of some sort of ‘inner’ awareness analogous to the outer awareness of a proxy in thought-expression. Or they might be thought of as subodoxastic.

One thing we know from the above is that Vertreter cannot explain, at least via a subject’s awareness of it, how this subject is able to take it that things are as per what that Vertreter vertretet. It cannot explain, that is, the subject’s grasp of how things would be in being as so represented (or of when things would so be, when not). It cannot explain this, or at least not in this way. Rather, as we have seen, such grasp is presupposed by the work thus assigned it. (Explandum in explanans.)

Here a genuine analogy between the allo and the auto. The proxy presupposes a capacity to see it for what it is. The Vertreter (so applied) presupposes a capacity for seeing it for what it is. But analogy stops there. As for the proxy, if it cannot explain the capacity it presupposes, what of it? Such was never in its remit. As for the Vertreter, if it cannot explain the capacity it presupposes, it fails at precisely what was its remit. Thus, disanalogy.

Not (so far) that no sense is to be made of the present idea of a Vertreter. One eligible to take things to be a given way—who, say, grasps what it would be for a pen to be on a table—might, for all said so far, appeal to a Vertreter as one might appeal to a proxy for better insight into whether the pen now in question (encased in plexiglas as it is) counts as a pen in the meaning of the (then-current) act. But such would be at best, contingent psychology, not part of what taking-that is as such.

Nor is it yet ruled out that Vertreter of present sort are at work subdoxastically, nor even that such subdoxastic work underpins a human capacity to think of pens and tables. But such, too, would be an empirical hypothesis. Moreover, for reasons to follow, so to hypothesise may already be to get the nature of autorepresenting very wrong.


5. In Third Person: The Funktion generates a set of items which (at a time) are those eligible objects of its owner’s takings-that. At least, the idea is, those generata (at a time) identify the things there are for that thinker to take so or not. One such thing is distinguished from another (so there are two things to be distinguished) just where the Funktion generates two distinct generata. What there is for its owner to take so or not is fixed accordingly (at least strictly speaking). Thus, the Funktion imposes a particular way of counting thinkables (when one twice, when two each once) in matters of attributing takings-that to its owner correctly (strictly speaking).

This contrasts with Frege’s idea that, in ontological order, whole thinkables (truth, or falsehoods, punkt) come first, decomposition (into logical forms) to be carved out of these. He thus writes,

I think that concepts arise through decomposing a judgeable. I do not think that for every judgeable content there is just one way in which it can be decomposed, or that some one of the possible ways may always claim objective priority. (1882: 118)

There is debate as to how far Frege envisions multiple decomposability extending. In any case, variation between decompositions runs along two dimension: a pair of decompositions of one thing may differ in the logical form each assigns a thought so decomposable; or they may differ in the fillings-in of blank space in a given logical form, e.g., a space for a one-place predicate. Or both.

Where Frege illustrates multiple decomposability, his examples vary along both dimensions. As for form,

[I]t is not impossible for the same Gedanke to be singular on one decomposition, particular [existential] on another, and on a third appear as universal. (1892a b: 200).

In any case, how the minimax (truth-or-falsehoods) are to be counted (where one thinkable occurs twice, where two each once) varies with when two decompositions are to be counted as of one thing. Were this to vary over occasions for decomposing, so, too, for ways of counting thoughts. A thinkable is known, in one way, by its decompositions.

For the rest, Frege on multiple decompositionality need only serve as inspiration. The inspiration wanted is just this: decomposition now appears as a parameter along which how thinkables are to be counted may be an occasion-sensitive matter. That is, what count as two decompositions of one thought for given purposes, or on a given occasion, may, anyway need, not do so on or for another.

So far this is only an hypothesis. But one misunderstanding is to be avoided from the start, namely, that somehow logic blocks such variation, demanding a fixed domain of thinkables to govern. What laws of logic govern in first instance are forms for a thought to take, their elements defined purely in terms of what being true is as such. Logic gives no advice on how to to fill out such a form into a full-fledged truth-or-falsehood. It only imposes criteria of adequacy on any filling to which the laws are to be applied. For example, where logic is to be applied, a given completion of a form must result in what can be counted (for relevant purposes) as a truth-or-falsehood. (‘Tertium non datur’.) For all of which, Frege tells us, we must not overlook “that a thought can be decomposed in multiple ways”, and “that various sentences may express the same thought.” (1892: 199) Correspondingly, which decompositions decompose some one thing may be liable to vary across occasions for the asking.

What might indicate exploiting this idea? For a clear view of this we need a certain distance from the phenomenon of taking-that. To achieve this I will focus on one special case of saying-that: ascribing takings-that to others, e.g., saying Pia to take goats to be nibbling at Sid’s lawn, or Benno to suppose that lamprey spawn in the Minho. The aim is to identify elements in the sort of understanding such ascriptions bear (or may need to).

I focus first on the dependent clause in such ascription. Its role is to speak of some given object of taking-that, thus some given (minimax) thinkable. If there are multiple ways for thinkables to be counted, variation across attributions in when two decompositions would count as decomposing some one object of taking-that, then an understanding of such a dependent clause, and work it does, may be called on to answer substantial questions as to when a decomposition of a thinkable would count as of the thinkable of which it is to be understood to speak. For such would then be one feature by which what it speaks of is distinguished from other items there are to be spoken of.

A decomposition, present notion, is of the minimax. So, to, what the thought-expression now in question ascribes. But in so doing it makes room for the notion paraphrase, which, though not itself belonging to the minimax, may run parallel close enough for us to bring out the point about interiority now aimed at. Caution is called for, as Frege warns:

As a means of thought-expression, language must resemble the thinkable. … [But] we must not overlook the deep gap which, for all that, separates the linguistic and the thinkable, and by which limits must be set to the mutual correspondence between the two domains. (1923(4?)/1897: 279)

The work of words in thought-expression is to make recognisable the thinkable(s) thus expressed. The words used have a syntactic structure. Which may, in turn, suggest a decomposition of the thought expressed, e.g., into a given denoting-predicating form. But we have no carte blanche to assume that each well-formed syntactic element in a thought-expressing sentence (or each at some level) is to be understood as dedicated to identifying a given element in ‘the’ decomposition of the thought expressed (in present matters) in a dependent clause. All we can ask of words per se is that the whole make recognisable a whole thinkable.

Just so that paraphrase has a role to play in identifying thought expressed. Frege, too, recognises this (Frege 1906, 1897: 213) in introducing what he calls ‘equipollence’. The same thing may be said in many ways. What it is that is said in any one way is identified (in one way) by what else would be a way of saying it. And this may depend on how a given way of saying something would be understood on the occasion of the saying. All of which translate to the proper understanding of a dependent clause in a given attribution of taking-that.

Pia (suppose) said, ‘Sid thinks lamprey are in season’, using those words for what they are for. Suppose she thus succeeds in saying Sid to take things to be a given way, one then identifiable in speaking of lamprey as ‘in season’. How else might she then have said Sid to think the same thing? Perhaps in speaking of lamprey as in the local rivers? Or swimming up them? Or spawning in them? Or turning up on local menus?

For each of these, perhaps so, perhaps not. Suppose that lamprey are manifestly in the rivers, but the word is that (global warming) they no longer spawn; or that, like ortalan, lamprey on the menu is now interdit. Or that, as the latest internet ‘debunking’ has it, lamprey swim down, not up, rivers, or perhaps do not swim at all, but really slither. One can imagine such things. Such need not require for a successful case of thought-expression, that the proxy which so expresses must choose some one among these as ‘the correct expression of what was thus said. Different circumstances may yield, on occasion, different same-sayings, correspondingly different classes of same-thinkers. Room is made for this by the that very idea, proxy, arrived at by dewaffling ‘Satz’ and applying Wittgenstein’s slogan accordingly. The point in a diagram:

$$ \begin{gathered} {\text{proxy}}\, \to \,{\text{consequences }}\left( {{\text{responsibility}}} \right)\, \to \,{\text{thinkable }}({\text{minimax)}} \hfill \\ \;\;\;\;\;\;\;\;\;\left[ {{\text{left}}\, \to \,{\text{right}}} \right] [{\text{right}}\, \to \,{\text{left}}]. \hfill \\ \end{gathered} $$

A proxy is not thinkable. It is rather a bit of history, visible, audible or etc. One works from it in two steps to a thinkable: first from left to right, from the performance to the treatment of things it would properly be held to underwrite, then from right to left to a generality, a way for things to be whose instancing by how things are would make things so treatable.

In such setting the idea of paraphrase comes to this: There need be no unique answer to the question how a give proxy is to be understood, or certainly not a formulation of an answer. Such might even be a rarity. There may be different ways of saying just how it would matter to how things were treatable were things as thus said. And where the proxy is to be understood as ascribing an instance of taking-that, such non-uniqueness may appear in the relevant dependent clause proper. Different ways of representing things may agree (so far as it mattters) in how things would be if as thus represented.

What thinkables there are depends, inter alia, on how thinkables are to be counted. Paraphrase points to at least one dimension along which counting thinkables can be an occasion-sensitive affair. In which case, mutatis mutandis for questions just what a given thinker (at a time) takes so, or not so. In given circumstances, saying Sid to think that lampreys are in season may be saying him to think the same as one would in thinking that lampreys are swimming up the local rivers, or on local menus. In other imaginable circumstances not.

To say this is to say that how a given thinker’s thought articulates into countable objects of his taking-that is determined not just by his thinking as he does, or by whatever enables this, but also on considerations extra to this. Paraphrase brings out one way in which this is so.

One more parameter along which articulation of a given thinker’s thought into countables is a candidate source of occasion-sensitivity. It starts from the truism that a thinkable is something for one to think, a possible object of one’s taking-so. So to attribute some given taking-so to a particular thinker is to represent him as belonging to a given indefinitely extendible class of same thinkers. As with any indefinitely extendible class, membership must be fixed, in part, at least, by a general requirement on such—a condition on so counting. What is now subject to being an occasion-sensitive matter is the question how, for relevant purposes, the class of thinkers full stop is to be articulated into classes of same-thinkers. Such is, accordingly, one element in an understanding some ascriptions of taking-that might bear (or call for).

I return briefly to the idea of an analog of a proxy with a role to play in autorepresenting. What role? Perhaps one of figuring in episodes, or events, of coming to see how to act accordingly, either through consultation by the thinker, as, chez McDowell, events of ‘making things explicit’ to himself, or subdoxastically. Such ‘pseudo-proxies’ (to name them) thus have an enabling role to play. An appendant idea might then be that these pseudo-proxies are supplied (as needed) by some sort of dedicated mental machinery, positioned at a time to generate some given stock of same, different generata for different things their owner takes so (or might) in thinking of things as he then does. So, on this view, what enables a thinker to take things to be the way he does, and to see them to be so treatable, auto-articulates that mass, his thought of things, into countable in a particular way, corresponding to the then current stock of pseudo-proxies.

Given the view of thought-ascription just sketched, such ideas of pseudo-proxies sound off-key. For on this view there is no unique way that a thinker’s thought articulates into thinkables. It is not as though articulating a thinker’s thought is a task assigned to, and reserved for, whatever ‘cognitive mechanics’ allows a thinker to think of things as he does. Rather, the point, on an occasion, of partitioning thought of things into same thinking—then relevant samenesses in the ways different thinkers are prepared to treat the world—imposes different things it would be for a thinker to count as thinking some given thing.

Since British empiricism, later abetted by Kant, philosophy has too often been in thrall to the idea that to understand a capacity is to understand how it is enabled. The above, I hope, suggests that it may sometimes be more enlightening to explore what it is that is accomplished by its exercise. Frege pioneered here, dismissing interest in aetiology in favour of interest in the grounds on which a capacity’s successes rest. (Cf. Frege, 1884: 3).


6. Why? What purposes are served by occasion-sensitivity in thought-ascription? Doubtless many. Here I illustrate just one. The illustration turns on the idea, just discussed, that to represent someone as taking it that such-and-such is to present him as same-thinking with a given indefinitely extensible class of others. It points to one reason for carving out classes of same-thinking in an occasion-sensitive way.

One source of such illustrations, I suggest, lies in a feature of such rational beasts as us by which we are meant to be distinguished (to our advantage) from the congenitally speechless. It is our ability to transmit and preserve information, inter alia, inter-generationally. An ability worth little without controls on the quality of what is thus transmitted. Controls in rapid decay since the advent of the smartphone. For all of which their exercise, I believe, remains part of what historians still do.

For illustration, a story, at least largely fictitious. Napoleon is said to have been an aficionado of the gâteau russe. As the story goes, he doted on these conspicuously enough to become their eponym. But might this story be pure hype, put out by some patissier to boost business? The kind of question historians deal in. Suppose, now, that in the attic of an old house in Toulon, a box was discovered containing a number of old diaries. Several of these mentioned a certain short man with a bicorne hat who was known around the mid-1790s for slipping into a patisserie near closing time, ordering multiple slices of gâteau russe, seeking a secluded corner and wolfing them down. The historian’s curiosity is whetted. Further accounts of this man occur in the diaries, referring to him as (translating freely from the French) ‘the gâteau russe slut’. These and further features of the diaries make for a fairly conclusive case that the man in question was the Corsican, Napoléon. Legend is thus confirmed as truth.

Through the diaries we have the presumably authoritative word of a few Toulonais that Napoléon doted on gâteau russe. Unlike us, they had the best grounds for saying so. They witnessed this. Of course, they did not refer to him as Napoléon, nor, when hearing of his later feats, trials and tribulations, would they have recognised the reports as about the old gâteau russe slut. Still, though, it was he who they took to be wolfing around closing time. They recognised the truth of just that about which we wondered—the truth, that is, that Napoleon doted on gâteau russe. As thinkables are being counted here, they, us, those historians, all belong to one extendible class of same-thinkers.

But for the historian to complete his research, and ensuing publication, he must carry out a certain inference. From evidence in the diaries he must be able to conclude that the ‘gâteau russe slut’ was Napoleon. Proof (pleonastically non-degenerate) is called for. One must be able to conclude from, e.g., The thought that the gâteau russe slut slipped into the corner of the patisserie near closing time to the thought that Napoléon slipped into the corner of the patisserie near closing time. Some substantial intermediate steps are called for in order to make this inference into a proof. The essential and crucial point here is entirely general: a non-degenerate proof is, by definition, from one thing to something else, from one thinkable to another. No such proof could be produced in a case like the present one unless there were a way of counting thinkables on which difference in denoting thought-element of the kind the example calls for (or some equivalent phenomenon elsewhere in a decomposition) could make for two such different thinkables.

What the example thus illustrates is one reason why thinkables must sometimes be so counted, and sometimes not. Which shows in one way why same thinkable needs to be an occasion-sensitive notion. The ascriber, in his dependent clause, mentions a thinkable in expressing it in one particular way in which it is expressible. If truly, there are the ways in which the subject of the attribution might give voice to, or otherwise express, the relevant allegiance. What these might be belongs to a proper understanding of what the ascriber said (for what he is thus rightly held responsible). These belong to a range of ways of expressing thoughts of which the ascriber’s dependent clause is but an example.

In Sum: More than one philosopher has tried to capture taking-that (autorepresenting) by analogy with saying-that (allorepresenting). There is an important analogy. But the disanalogies matter at least as much. The analogy, put in present terms, is this: The crucial move, in auto- as in allo-, in taking as in saying, is from applications to thinkable applied. (Where there is a match-up between these is open to being an occasion-sensitive matter.) The crucial disanalogies lie in the different grammars of the verbs. Most basically, saying is a performance, taking-that a steady-state condition (i.e., not an unfolding episode in present sense of ‘unfolding’). Which shows up in the very different natures of commitment made in saying-so, and commitment saddled with in taking-so.

Ascribing taking-so may be (to be understood) as in service of any of a number of goals. Prominent among these (notably, giving proof, answering questions who thinks such-and-such—where there is same-thinking) are ones which require the object of the taking (a thinkable) to be counted in different ways for different ascriptions. Consequently, what count as expressions of the same thinkable varies from occasion to occasion for the counting. So, too, one thinkable may be subject to different assignments of logical form, and different fillings out of a given form.

The above (among other things) gives reason not to try to associate a taking-that with some identifiable bit of psychomechanics purveyed as just that by which so thinking is/was enabled. Given the different things taking things to be some given way may may properly be understood to come to on an occasion, there is no reason to expect there to be any such enabling explanation which in fact does the work required.

The fundamental fact here is that of the thinker in question being responsive as he is to how things are, prepared as he is to treat things. This can be articulated in many ways into preparedness to treat things (or see them to be treatable) as they would be if things were thus and so. How he may be said truly to take things is fixed accordingly. One cannot deduce the enabling from the success enabled. Nor need there be any general story as to how such successes are achieved. Same-thinking lies, not in aetiology, but in common responsiveness to how things are; in what a thinker is prepared to recognise. Such, as a rule, leaves it largely indeterminate what aetiology might be, or even was where it came to cases. The myth of interiority arrives in missing, or ignoring, precisely this point.