Seizing the World: From Concepts to Reality

1 Prologue: Argumentum ad Lapidem

In a famous anecdote, Boswell recounts an incident that took place in 1763, when he and Dr. Johnson stayed at Harwich and visited the local church:

After we came out of the church, we stood talking for some time together of Bishop Berkeley’s ingenious sophistry to prove the non-existence of matter, and that everything in the universe is merely ideal. I observed that though we are satisfied his doctrine is not true, it is impossible to refute it. I never shall forget the alacrity with which Johnson answered, striking his foot with mighty force against a large stone, till he rebounded from it, ‘I refute it thus.’ (Boswell 1998, 333)

There has been a broad consensus among philosophers that Dr. Johnson’s experiment failed miserably as a disproof of Berkeley’s idealism; for all it showed was that a certain sequence of sensations or sense-data followed one another in Johnson’s consciousness—the sensation of his conversing with Boswell, the sensation of his hitting the stone and finally the sensation of pain in his toe—all of which Berkeley would not have denied. Yet, Berkeley would have asked what evidence there is for assuming that Johnson’s sensations refer to something real—something that exists independently of his experiences. Indeed, why should the stone that Johnson perceived be any more objective and public than the pain he felt?

I daresay, however, that Dr. Johnson was on the track of something. Perhaps, his point was not so much that his sensations of striking the stone and of his foot’s rebounding from it prove the stone’s external existence, but rather that the striking and the rebounding themselves reveal an independent reality of the stone (cf. Hallett 1947, 138). Johnson’s intuition might have been that we do not just passively witness the phenomena of our experiential life, but actively engage with them; that those phenomena, in turn, do not just conform to our actions, but counter them in unforeseeable and uncontrollable ways—and, in this interplay, manifest their intrinsic agency and reality. This, at any rate, is roughly the story I would like to tell in this essay.

Idealism takes many shapes (cf. Guyer and Horstmann 2020). Metaphysical idealism, pinnacled in Berkeley’s immaterialism, claims that all the things there are depend for their existence and nature on the cognitive activities of minds. Epistemological idealism, epitomized in Kant’s critical philosophy, holds that we cannot know anything about things as they are in themselves, independent of our cognition of them. Epistemological idealism often motivates metaphysical idealism. But it does not entail the latter. Kant’s transcendental idealism, which alleges that there exist things-in-themselves although they cannot be known, stops halfway between epistemological and metaphysical idealism.

The kind of realism I wish to defend in this essay is epistemological realism. Epistemological realism implies metaphysical realism: it contends that there exist entities whose being and character is independent of all minds but nevertheless cognitively accessible to at least some of them. My argument for epistemological realism will be a transcendental one. It starts with the truism that we are conceptual creatures—holders and users of concepts—and reasons from there to the existence of a knowable outer world as a condition of the possibility of such creatures. In Section 2 I lay down the general approach to concepts and conceptualization which will be at the heart of my argument. According to this approach, concepts are rules that cognitive agents follow in unifying experiences, making judgments, or using words. In Section 3 I explicate the peculiar normative nature of rules and rule-following, from which I extract, in Section 4, following considerations by the later Wittgenstein, what I take to be an intractable problem for any idealist account of concept possession. In Sections 5 and 6 I elaborate and defend the view that the very possibility of conceptualizing requires the existence of external objects that enable cognizers to regulate their use of concepts. I close in Section 7 with some remarks on the nature and scope of the kind of metaphysical knowledge that emerges from the preceding insights.

2 Concepts: Representations versus Rules

There have been two broad approaches to concepts in the history of philosophy. According to the first one, whose roots can be traced back at least to Plato and Aristotle, concepts are representations. More precisely, concepts are supposed to be things which are, in some sense, about or directed towards other things and—by way of being present in or to the mind—convey knowledge of these things. In Aristotle, e.g., concepts are identified with “affections of the soul,” which are “likenesses” of “actual things” in the minds of epistemic subjects, i.e., mental representations (De Int.16a3–9). Actual things are seen as composites of form and matter—the former of which makes them things of certain kinds, while the latter makes them particular things. According to Aristotle, when an epistemic subject thinks of a particular thing as being of a certain kind, its intellectual faculty receives the intelligible form of that thing (abstracted from its matter) in much the same way in which a piece of wax takes on the imprint of a signet ring (De An. 424a20).

Representational views of concepts are open to different ontological interpretations. Conceptual representations might be deemed, e.g., abstract entities (as in Plato and many formal sciences) or mental particulars (as in Aristotle and most contemporary cognitive science). Representationalism also does not necessarily lead to epistemological or metaphysical idealism. According to Aristotle, and later Aquinas, e.g., the things we conceive are entirely independent ontologically of our conceptions of them. At the same time, our conceptions are epistemically transparent, as the very same form that is ‘really’ present within the things conceived is ‘spiritually’ present in the conceiving mind. Hence, there is a formal identity or isomorphism of our concepts and the realities they represent, which makes our cognitive access to the former necessarily an epistemic access to the latter (cf. De An. 432b21–23).

At least since the advent of Cartesianism, however, representationalism has become associated with skepticism and idealism (cf. Tegtmeier 2005, 77–9). In Descartes, concepts are assimilated to ideas, i.e., simple modes of thinking, which are “as it were the images of things” (AT VII 37). As such, ideas are “whatever is immediately perceived by the mind” (AT VII 181). Now, Descartes is not an idealist—he believes that at least some ideas originate from things outside the thinker. Yet, he also holds that “when a thing exists in the intellect by means of an idea, it is not an actual entity, that is, it is not a being located outside the intellect” (AT VII 103). For Descartes, then, what the mind immediately perceives when it thinks of a certain thing is not the actual thing but a purely mental surrogate.

Descartes’s theory of ideas has been crucial for engendering skepticism, among philosophers, towards the knowability of the external world. Cartesian ideas are epistemically opaque: they are tertium quids which potentially screen off the mind cognitively from the realities ‘behind’ them. Inevitably, the question arises whether we can know the world as it is in itself at all. While empiricists in the aftermath of Descartes, such as Locke and Hume, tried to avoid this question by focusing on the given—the mind and its operations—idealists like Berkeley decided to grab the bull by the horns and endorse metaphysical idealism to lay skepticism to rest (cf. Guyer and Horstmann 2020, Section 3).

Representationalism, however, has a more fundamental flaw than its susceptibility to idealism. Representational views of concepts typically come along with passivist accounts of concept possession. On these accounts, possessing (knowing, understanding) a concept is a condition of the epistemic subject that more or less automatically ensues as a result of certain cognitive processes, often modeled on sense perception, which the subject undergoes. Passivist accounts accommodate what is imparted into our pretheoretic notion of conceptualization by the etymology of the term ‘concept’ (which derives from Latin ‘concipere’ = ‘to take in, become pregnant’). At the same time, they appear to ignore an essential aspect of our intuitive idea of concept possession and of the role concepts play therein. It does not seem to be enough for an epistemic subject to have a concept of something that it is receptive to certain things and affected by them in a certain manner. Such a subject no more conceives of things than thermometers conceive of temperature and a rusting piece of iron conceives of moisture. What is additionally required for possessing a concept, it seems, is that the epistemic subject assume a certain attitude towards the things that act on it and treat them in a certain manner.

Intuitions of this kind thus have motivated activist accounts of concept possession, according to which conceptual knowledge or understanding manifests itself by a special conduct of the epistemic subject. Such accounts take up what is preserved in the term ‘Begriff,’ whose etymology (harking back to Old High German ‘bigrīfan’ = ‘to catch hold of, seize’) more accurately corresponds to the etymology of ‘comprehension’ than to that of ‘concept,’ although it usually translates as the latter. The metaphor of prehension at work in the verbs ‘comprehend’ and ‘begreifen’ emphasizes the active character of conceptualization. Conceiving is something done rather than suffered. It is something that is attributable to the epistemic subject, and not merely explicable in terms of the things that inform the subject’s thoughts.

Philosophers have deemed several types of action fundamental to the kind of conduct that goes along with concept possession. One accomplishment of conceptualizing things is to sort them into classes. This may be done by words or deeds (both in- and outwardly). Just as calling things dangerous is a way of classifying them as dangers, eating things is a way of classifying them as food. Another feat of conceptualization is to draw inferences from one’s classifications. Again, this may be done verbally (as when one says, “Fido is not a cat,” upon telling that Fido is a dog) or practically (as when one looks for the fox upon finding its trail). Whatever the relevant activities of conceptualization are in detail: what is crucial about them, according to activist accounts, is that they are endogenously controlled, rule-governed performances. Concept possessors do not just causally interact with their environment—they do so in a principled manner which is both goal-directed and adaptive. Classifications, e.g., serve epistemic subjects to reduce the amount of information they need to process so that they can treat many things equally, and are adjusted to strike an optimal balance between cognitive informativity and economy. Inferences, in turn, serve epistemic subjects to activate background information that they can use to explain and predict how things behave, and are attuned to achieve the most successful interaction with those things.

Insisting on the regulative significance of concept possession, activist accounts suggest a novel view of the function of concepts in conceptualization. From their perspective, concepts emerge, not as representations an epistemic subject acquires by the end of the cognitive process, but rather as norms that govern this process from the outset. Thus, activist accounts of concept possession have led to the second main philosophical approach to concepts. According to that approach, championed by philosophers like Kant, Frege and Wittgenstein, concepts are rules. More precisely, concepts are supposed to be the rules a subject uses and follows in its epistemic conduct—the rules that guide its conceptual agency. Kant, e.g., identifies concepts with rules for synthesizing the empirical manifold, i.e., for sorting and structuring one’s perceptions of objects (cf. A 106). Frege, in turn, equates concepts with functions that map objects to one of the two truth-values, i.e., with rules for making judgments about these objects (cf. Frege 1960, 31). Wittgenstein, finally, likens concepts to “technique[s] of using a word” (LPP, 50), i.e., rules for the application of words to objects, for their employment in declarative, interrogative, and imperative sentences, and for their use in definitions and arguments.

Like representational views, regulistic views of concepts are amenable to different ontologies. Conceptual rules might, e.g., be taken to be mentally codified (as in Kant), abstractly reified (as in Frege) or practically enacted (as in Wittgenstein). Moreover, regulism is not necessarily incompatible with representationalism (cf. Boniolo 2001, 100–1). Kant already takes concepts also to be representations by virtue of being rules; and, in fact, almost all contemporary versions of representationalism—which mostly revert to teleological (cf. Millikan 1984) or functional (cf. Block 1987) theories of mental content—explain the representational qualities of concepts at least in part with recourse to their regulative role in the production of cognitive behavior and higher-order thought processes.

Such assimilations are legit as far as they go. But they must not mislead into underestimating the significance of the paradigm shift from the representational to the regulistic approach. For, where the former resorts to a descriptive notion of concepts and concept possession, the latter advances a normative one (cf. Brandom 1994, 9–10). The far-reaching insight of regulism is that concepts specify criteria for the correct implementation of conceptualizations to which they bind their possessors. Thus, an epistemic subject’s classifications, inferences, etc., become liable to evaluations according to the concepts they involve. It is this peculiar normative dimension of concepts and conceptualization that demarcates activist accounts of concept possession from passivist ones, and regulistic views of concepts from representational ones.

3 Normativity: Freedom and Force

To appreciate the philosophical implications of the regulistic approach to concepts and conceptualization it is advisable to look more closely at rules and the nature of rule-following. Rules are distinguished from laws of nature by the way they relate to individual objects or agents. Formally, this relation—the rule-following relation—is characterized by the way it partitions the domain of individuals. While laws divide individuals into those which do fall under them and those which do not, rules divide individuals into (i) those which follow the rules correctly, (ii) those which do so incorrectly and (iii) those which do not follow them at all. This threefold division mirrors two essential (and interconnected) facts about rules and rule-followers (cf. Railton 1999, 322–4). On the one hand, rule-followers retain a certain freedom vis-à-vis the rules they follow: if an individual x follows a rule R, then x will still be able to act contrary to R without thereby losing its status of following R (cf. Tolley 2006, 375). On the other hand, rules hold a certain force vis-à-vis the individuals who follow them: if an individual x follows a rule R, then x will be appropriately assessable in light of R, i.e., it will be legitimate to appraise and criticize x for acting in accordance or disagreement with R (cf. Kiesselbach 2014, 430).

Both facts set rules apart from natural laws. First, laws do not allow for violations. Some laws, it is true, allow for exceptions (in the sense that they allow for individuals which satisfy the antecedent of the law and yet fail to satisfy its consequent). However, behaving exceptionally according to such laws is not violating them. For example, if a certain substance falls under a ceteris paribus law to the effect that, ceteris paribus, this substance will dissolve if it is put in water (where the ceteris paribus-clause provides some nontrivial explanation for deviations from this regularity), then it will be possible for this substance not to dissolve when being put in water to the extent that its failing to dissolve can be explained by the ceteris paribus-clause of that law. But then, realizations of this possibility are perfectly consistent with that law; hence, there can be no talk of violation. Nor can one speak of violation if it is possible for the same substance not to dissolve in water to an extent that is not covered by the relevant ceteris paribus-clause, since in that case the substance simply does not fall under the law in question.^[1]

Second, and relatedly, natural laws do not afford appropriate assessments. They do afford assessments, but only in the sense that one is, of course, free to assess individuals in light of them—e.g., as being correct when behaving ordinarily according to an applicable (non-strict) law and incorrect when behaving exceptionally according to that law. Yet, such assessments would not seem to be appropriate. At least they will beg question why the individual should be measured against that law rather than against others. After all, if ‘exceptionally’ just means that an individual fails (within the bounds of possibility) to satisfy the consequent of a given law, then there will automatically be, for any law L such that an individual x behaves exceptionally according to L, a complementary law L* such that x behaves unexceptionally according to L*. For example, if a match behaves exceptionally according to the law “A match will light if struck (unless it is wet),” then it will behave ordinarily according to the complementary law “A match will not light if struck (unless it is dry).”^[2] Choosing one of these laws for assessing the individual’s behavior would be baselessly presuming that the individual ought to be evaluated in terms of that law (cf. Brandom 1994, 28). The upshot is that critical estimations of the behavior of individuals are themselves in need of justification—which, it seems, can only be given by the rules these individuals follow.

Materially, the formal properties of the rule-following relation can be explained by the way rules and followers participate in it. Kant was the first to point out that individuals which fall under a law of nature enter into that relation passively, on account of their nature, whereas individuals who follow a rule enter into that relation actively, by virtue of their assent: “Everything in nature works in accordance with laws. Only a rational being has the capacity to act in accordance with the representation of laws, that is, in accordance with principles, or has a will” (AA 4:413). Kant’s insight is that an individual x follows a rule R if and only if x accepts R, where x’s acceptance of R involves (i) an understanding of R on the part of x (“the representation of laws”) and (ii) an intention of x to act according to its understanding of R (“a will”).

The Kantian analysis accounts for both the peculiar force of rules and the characteristic freedom of their followers. First, rule-followers are appropriately assessable in light of the rules they follow because they accept these rules: it is precisely because individuals commit themselves to certain rules why one is entitled to judge on their actions in light of those rules. Second, rule-followers are capable of acting contrary to the rules they follow because those rules affect their actions only indirectly by affecting their intentions, which in turn affect their actions, with subjective and random factors possibly interfering at any point. (Thus, an individual’s intention to act according to a rule may be misinformed by false beliefs, superseded by other inclinations or suspended by free choice; cf. AA 4:414; AA 17:465–6.) Therefore, a rule’s grip on its followers will remain loose and provisional, even if the latter accept the rule as compulsory and unconditionally valid. Accordingly, rules can at best dispose but never determine their followers to act in accordance with them (cf. AA 4:413).

Kant’s account of rule-following is as ingenious as his regulistic approach to concepts. Yet, one might feel a certain tension between the two. In Kant’s theory, for an individual to possess a concept is for it to follow a rule. For an individual to follow a rule, in turn, is for it to accept that rule, which, again, is for it to have an intention to act according to its understanding of that rule. Now, having this intention and understanding seems to presuppose the possession of a concept of that rule on the part of the individual. But then, the individual’s possession of one concept seems to presuppose its possession of another concept, hence its following another rule, and so on. Thus, Kant’s theory seems to fall prey to an infinite regress.

This, of course, is nothing but Wittgenstein’s famous rule-following paradox (cf. PI, § 201). Wittgenstein’s own solution to this paradox is to assume that “there is a way of grasping a rule which is not an interpretation [i.e., a conception], but which, from case to case of application, is exhibited in what we call ‘following the rule’ and ‘going against it’” (ibid.). Hence, Wittgenstein claims that an individual can have practical knowledge of rules without needing to have intellectual knowledge of them—just as a child can practically know how to ride a bicycle without having to know it intellectually (cf. PI, § 202).^[3] To follow a rule, then, an individual does not have to accept that rule explicitly, i.e., it need not reflectively understand the rule and intend to act according to that understanding. All that is required for its rule-following is that the individual demonstrate its acceptance of the rule, i.e., its relevant understanding and intention, in the very way it acts.

Wittgenstein’s solution to the paradox of rule-following does not commit one to a Wittgensteinian ontology of rules—not without further ado, at least. With some charity, one may extend Wittgenstein’s idea of the practical enactment of rules to the realm of mental actions. Thus, one may think of inner manifestations of an individual’s acceptance of a rule on the model of agents’ overt manifestations of that acceptance to embrace the possibility of purely mental acts of rule-following, without having to specify what the covert counterparts of such outer manifestations in fact consist of (cf. Rosenberg 1980, 120). There is, however, another problem with the Kantian account of rule-following in the vicinity, whose ramifications Kant does not seem to have acknowledged sufficiently.

4 Rule-Following: Autonomy and Authority

Kant’s account of rule-following is, in essence, his idea of autonomy. For an individual to follow a rule is for it to submit itself to that rule; rule-followers are self-legislators. Basically, the problem with the Kantian account of rule-following is that the idea of autonomy it promotes threatens to be paradoxical (cf. Pippin 2000). Completely autonomous rule-followers would have full authority over the rules they follow; for they alone could choose the rules and determine what counts as following them. But then, these rules could not have any authority over them: after all, one could not even criticize the followers, should they not conform to a certain course of action. Hence, completely autonomous rule-followers would not be following rules, but behaving haphazardly and unblameably. There would be no nomos and, hence, no autonomy at all, just capricious antics.

The issue at stake becomes clear once one explicates the autonomist principle that drives the Kantian account of rule-following. According to the Kantian account, an individual x follows a rule R if and only if x accepts R. This is Kant’s autonomist idea. Furthermore, x accepts R if and only if x has an intention to act according to its understanding of R. By implication, x will follow R only if x has an understanding of R. Call this the cognitive requirement for following a rule. Now, what does it mean for an individual x to have an understanding of a rule R? It means, among other things, that x has a distinction between correct and incorrect implementations of R at its disposal. Roughly, for x to have an understanding of R, there must be two sets of possible actions the elements of which x takes to be correct and incorrect instances, respectively, of following R (cf. Rosenberg 1980, 99).

Now, the crux with the cognitive requirement is not that one must not think of x’s ‘taking’ actions to be correct or incorrect instances of following R as x’s consciously conceiving of these actions as being such, on pain of infinite regression. As explained before, an individual’s ‘taking’ things to be a certain way may be thought of as manifesting itself by the individual’s (physical or mental) conduct, i.e., by its practically treating these things that way. The trouble is rather in the idea that x’s distinction between correct and incorrect implementations of R is what x takes that distinction to be. For, if whatever x takes to be correct (or incorrect) actually is what is (in)correct for x, there will be, for x, no difference between real and apparent (in)correctness. Lacking the difference between genuine and seeming (in)correctness, however, x will in fact not have a distinction between correct and incorrect implementations of R—and lacking the distinction, x cannot have an understanding of R, let alone an intention to act according to that understanding. In the end, x cannot even be said to accept, i.e., follow R. The upshot of these considerations is that as long as an individual alone determines what counts as following a rule (correctly and incorrectly), no sense can be made of its following rules at all. To make it possible for an individual to follow a rule, there must be a standard for correct implementations which the individual can adopt but which at the same time is independent of its own judgment.

Due to Kant’s neglect of the need for objective criteria for one’s correctly following a rule, regulism has often been associated with idealism. The cognitive requirement of the Kantian account of rule-following appears to entail that rule-followers are completely unconstrained both in picking the rules they follow and in deciding how to follow them. In conjunction with the regulistic approach to concepts, according to which possessing concepts is (a case of) following rules, it seems to follow that an epistemic subject is entirely free in choosing its concepts as well as its preferred manner of applying them. Even if one must assume the existence of noumenal things as that to which the subject applies its concepts, it would seem to be totally up to the subject how it conceives of them. Indeed, there would seem to be nothing more to the fact that these things fall under certain concepts than that the subject takes them to fall under those concepts.

It belongs to Wittgenstein’s credit that he expounded the necessity of external criteria for the proper ability to distinguish between correct and incorrect implementations of rules as a condition for the possibility of rule-following and, hence, concept possession. In his famous private language argument (cf. PI, §§ 243–68), Wittgenstein contemplates the possibility of an individual which contrives its concepts (the rules for the use of its words) all by its own, just to disprove this possibility on account of such an individual’s unavoidable lack of independent criteria for the correct application of its concepts (cf. PI, § 258). Wittgenstein’s argument has often been regarded as making a case for behaviorism, or as being intended as a solution to the problem of other minds, because Wittgenstein mostly discusses the intelligibility of private languages using the example of private mental concepts, i.e., concepts for interoceptive sensations such as pain (cf., e.g., PI, § 257). But his point applies to concepts for exteroceptive impressions as well. Regardless of whether one is, say, having a toothache or seeing something red, the criticism of the private language argument is that an individual’s cognition of things will be frustrated if the criteria for their correct conceptualization are not independently fixed (cf. PO, 244).

The picture that emerges from these considerations is that an individual x will possess a concept C only if x can avail itself of an external, independent criterion for C’s correct application, y. For x to avail itself of y is for x to “calibrate” (Pears 1988, 334) its use of C on y—in the minimum sense that x is (or becomes) capable of recognizing a sufficient number of correct and incorrect applications of C. Correlatively, for x to calibrate its use on y is for y to act on x as a “stabilizing resource” (ibid., 368) in the sense that y has an intrinsic propensity to make x recognize sufficiently many (mis)applications.^[4]

The presented argument against an unfettered conceptual autonomy of the epistemic subject raises a question and a worry. The question is what, if not the whims of the cognizer, constitutes the criteria for the correct implementation of conceptual rules. The adjacent worry is that whatever the answer to the first question may be, it will still be possible to reduce the actual criteria to purely internal phenomena.

5 Beyond the Private Realm: People and Things

In the long-running debate that has been sparked by Wittgenstein’s examination of private languages, two types of candidates have been cast for the role of external criteria for the correct application of concepts: other people (minds, agents) and things (nonmental, non-agential objects). According to the first proposal, individuals must check their execution of conceptual rules against the (cognitive) behavior of fellow rule-followers (cf. Kripke 1982, 86–92). According to the second proposal, they have to check it against the (nomological) behavior of ordinary objects that may be available whether or not other rule-followers are present (cf. Baker and Hacker 1984, 38–42).

One ramification of the first proposal is that it makes an individual’s possession of concepts—and indeed all its cognitive abilities based on this possession—dependent on an agreement between the individual’s own rule-following and the practices of other rule-followers. Such a dependence obstructs the possibility of a cognitive ‘Robinson Crusoe’ who follows conceptual rules all by himself, which is why this proposal has come to be known as communitarianism. According to the second proposal, by contrast, an individual’s possession of concepts depends on a correspondence between its own rule-following and the behavior of ordinary objects in its environment. Since such a dependence is compatible with the hypothesis of a cognitive Robinson Crusoe, this proposal has been labelled individualism.

Another ramification of the communitarian proposal is that it leads to a social constructivism which—in contrast to solipsistic idealism—acknowledges the existence of a community of minds (or minded agents) but maintains that the natures of all nonmental things depend on the collective cognitive activities of these minds (or else cannot be known). Which concepts apply to these things is not determined by the properties that the things have per se but rather by the conceptual rules that the members of the community of minds inflict on themselves for their common handling of those things. The natures of things are not (known to be) what these things are in themselves, but what they are for those minds. Only the individualist proposal promises to yield some form of epistemological realism in that it holds that the concepts applying to nonmental things are at least partly determined by the properties of these things, so that their nature is revealed, at least to some degree, to the conceptual agent.^[5]

On the face of it, checking one’s applications of concepts against the behavior of things should be no less possible than checking one’s applications against the practices of other people. Communitarians typically argue against the individualist proposal, however, that checking one’s conceptualizations against the behavior of things (rather than against the practices of other people) is impossible, as ordinary things lack the requisite power to impart a difference between correct and incorrect conceptualizations to an individual:

My sequence of [applications of a concept] cannot be consistent or inconsistent with the world, in brief, simply because any sequence of [applications] which I produced … is a sequence of responses evoked in me by the action of the world on me, so any such sequence bears the same relation to that world as any other. The world is not an object of comparison for me … The only consistency available to the individual here is the synchronic consistency of present impression and simultaneous ostensible memory, but this becomes the sought diachronic (semantic) consistency only if impression and ostensible memory are themselves correct. The only place which remains to seek the fact of correctness, then, is … in the community. The correctness of an individual’s [application of a concept] can only consist in this: that others agree. (Rosenberg 1980, 101, 105)

The communitarian argument seems to run as follows. Recall that an individual x will possess a concept C only if x can calibrate its use of C on some criterion y that acts on x as a stabilizing resource. Now, what the communitarian apparently wants to claim is that, where y is an ordinary object, x’s calibrating its use of C on y is itself a conceptual affair; for, it involves x’s application of concepts in certain perceptual and memorial judgments about the behavior of y (i.e., that object). In the absence of such prior conceptual activities on the part of x, y would always bear the same undiscriminating relation to x and could not impose any bivalence on x’s conduct (cf. Rosenberg 1980, 103). In other words, y can only act on x through x’s conceptualizations of y—and how y acts on x depends on how x conceives of y on different occasions. But then, y will in effect be incapable of acting on x as a stabilizing resource, i.e., it will be impossible for x to calibrate its use of C on y. Hence, x will ultimately fall short of possessing C.

The communitarian argument prompts both internal and external criticism. The internal criticism is that the argument’s central premise, if true, will backfire on the communitarian. If an individual’s calibrating its use of a concept really requires its applying further concepts to the criteria for the correct application of the first concept, then the individual’s calibrating should require such applications regardless of whether the relevant criteria are ordinary objects or fellow rule-followers. If an individual can do no more than check its perceptual and memorial judgments about objects for consistency with its use of a concept, then it should likewise be unable to do more than check its judgments about other rule-followers for such a consistency. But then, the communitarian’s conclusion to the effect that the criterion for an individual’s correct application of a concept must be the agreement with other people will be a non sequitur.

The ensuing external criticism of the communitarian’s argument is that its central premise cannot be true, on pain of infinite regression. If an individual’s possession of one concept required that the individual apply other concepts to whatever is the individual’s criterion for the correct application of the first concept, then the individual would never acquire any concepts. (This, again, is the gist of Wittgenstein’s rule-following paradox.) As mentioned above, the communitarian’s appeal to people will not stop the regress. The only way to cut it off is to assume that there is a way of calibrating one’s use of a concept which itself is not mediated by prior conceptual activities of the individual. But this is to assume that there are entities which possess an intrinsic capacity of acting on an epistemic subject as stabilizing resources. And there is no reason to think that these entities could not be things just as well as people.

In fact, on closer inspection, the possibility of checking one’s conceptualizations against the practices of other people turns out to presuppose the possibility of checking those conceptualizations against the behavior of things (cf. Pears 1988, 368–71). For, people can only agree in their applications of concepts if they apply these concepts, everyone for themselves, constantly in the same way. Yet, people will apply their concepts in the same way only if they apply them to the same persistent kinds of things; and that they apply them to such things is a fact (if it is a fact) that is determined by how the things are—how they behave, not the people. For example, if it were not for the fact that certain things are red and keep their color, there could be no fact of the matter as to whether people apply the concept of redness to the same kind of things, thus, apply the concept in the same way and, hence, as to whether they agree in their application of the concept. Accordingly, it will only be possible for an individual to verify whether its own conceptual activities agree with the practices of other people if it is possible for the individual to verify whether it, for itself, applies its concepts to the same persistent kinds of things. And that, in turn, is possible for the individual only if it is possible for it to check its conceptualizations against the way those things behave.

So, one cannot but acknowledge the existence of nonmental, nonagential things with an intrinsic capacity of stabilizing an individual’s use of concepts. Acknowledging the existence of such things, however, amounts to assuming metaphysical realism about them—precisely in the sense that those things have an existence and character independently of their conceptualization by the individual. It also amounts to epistemological realism with respect to the individual’s knowledge of them; for the fact that such independently existing and charactered things can stabilize an individual’s use of concepts means nothing other than that the individual can calibrate its application of concepts on them. Now, as stated earlier, an individual’s ability to calibrate the application of its concepts involves, at least, the ability to recognize the correctness and incorrectness of its use of concepts in a sufficient number of instances. But this presupposes that the individual be able to recognize with fair reliability when the external criteria of the correct use of its concepts, namely, the objects to which those concepts are properly applied, are present or absent—which in turn requires that the individual have some knowledge of the intrinsic natures of those objects.

It is, however, important to understand the specific nature of this foundational knowledge that an individual has about the (kinds of) things on which it calibrates its concepts. First, as already noted, this knowledge cannot be conceptually mediated knowledge, because such knowledge would obviously presuppose a possession of the concepts by which it is mediated, and therefore require prior knowledge about the application criteria of those concepts. But second, it cannot be perceptual knowledge either, if by this is meant a cognition that is mediated by internal (nonconceptual) acts of perception. Such perceptions would equally lack the objectivity required for the criteria on which the individual calibrates its use of concepts and would themselves stand in need of external conditions of adequacy that determine their perceptual content and degree of veridicality (cf. PG, § 54). The specific knowledge that grounds an individual’s calibration of concepts cannot, therefore, be understood other than as a direct acquaintance with the external criteria of the correct application of the relevant concepts—as a knowledge, that is, that cognitively connects the individual to the referents of these concepts themselves, not merely to some subjective (conceptual or perceptual) representations of them. Such knowledge cannot consist in any kind of intermediary awareness or imagination (although it may certainly be accompanied by such phenomena), but only in the individual’s interactive rapport with the objects of its cognition, which recurrently manifests itself in the way the individual practically responds to those objects in the course of calibrating its application of concepts on them (when, e.g., it shows some proficiency or sense of accomplishment in applying a concept correctly, or expresses astonishment or frustration at a misapplication).^[6]

This need not mean that such practical acquaintance provides the epistemic subject with an infallible positive knowledge about the objects of its cognition. Depending on exactly what an individual’s ability to calibrate its conceptual activities entails, the knowledge manifested in the course of such calibration may be of a rather negative kind. If the ability to calibrate merely includes that an individual can (learn to) distinguish correct from incorrect applications of concepts (as has been assumed so far), then it may even be that the individual recognizes nothing but that its concepts are constantly misapplied. Only if the ability to calibrate further implies the ability to (learn to) fix at least some misapplications, an individual will come to know the objects of its cognition as they positively are in the course of calibrating its concepts.

I will come back to the issue of what exactly should be taken to be involved in the cognitive faculty of calibrating concepts in Section 7. Until then, just note that, either way, the very fact that an individual can or cannot refer to things by means of certain concepts tacitly shows the individual something about the things as they exist in themselves—namely, that they are such that it can, respectively, cannot refer to them by means of those concepts. The degree of conformity or discrepancy between an individual’s conceptions of things, on the one hand, and the practical results obtained in the course of its interaction with those things under the presupposition of these conceptions, on the other, reveals to the individual the existence and—in one way or another—also the essence of a reality that defies its control while at the same time enabling its conceptual-epistemic endeavors.

6 Back to the Subject? Phenomena and Phenomenal Laws

The private language argument, according to the proposed interpretation, licenses a transcendental argument for epistemological realism. From the evident fact of (our) conceptual agency it derives the existence of a knowable external world as a condition of the possibility of such agency. Now, there is a familiar objection to transcendental arguments, according to which, whenever the truth of a particular proposition—e.g., the proposition that there is an external world or that there are other minds—seems to be a necessary condition for there being any meaningful language, “the skeptic can always very plausibly insist that it is enough to make language possible if we believe that [this proposition] is true, or if it looks for all the world as if it is, but that [the proposition] needn’t actually be true” (Stroud 2000, 24). This critique seems to hold, e.g., with respect to ‘hinge propositions’ (cf. OC, § 341), like the proposition that the earth existed long before one’s birth (cf. OC, § 84). Wittgenstein thinks such propositions are exempt from doubt, as they form the “background against which I distinguish between true and false” (OC, § 94). Yet, the skeptic may urge that their truth “does not follow from everyone in the world’s believing [them], even fully reasonably believing [them], or their being completely unable to avoid believing [them]” (Stroud 2000, 196).

The present deduction, however, is immune to this objection. The thrust of Wittgenstein’s remarks, as I read them, is that the possibility of conceptualization, and thus the very possibility of forming beliefs, presupposes the existence of an external world. Beliefs may be true or false; but what determines the contents of belief—concepts—cannot be systematically out of touch with reality, on pain of losing all sense of right and wrong (cf. O’Donovan-Anderson 1997, 120–1). All concepts, and all thoughts composed of them, turn on a distinction between ‘being’ and ‘seeming’ that is borne by a friction with an external world. Consequently, any belief will presuppose the existence of such a world—including the belief that there is an external world and the belief that there is none. Thus, from the perspective of the private language argument, it is entirely orthogonal to the meaningfulness of language and thought whether everyone believes (accidentally, reasonably, or compulsively) that there is an external world. Even if everybody doubted that fact, their doubt would only be possible because of that fact. In other words, the skeptical doubt must be false in order to be meaningful (cf. Davies 2018, 184–5).

Nevertheless, the worry that the proffered argument leaps to its realist conclusion is not removed. An idealist might still contest the original claim of the private language argument, namely, that the stabilizing resources for an individual’s calibration of concepts—whether they are thought to be things or people—must be external to the epistemic subject. The idealist might insist that whatever is supposed by the realist to act on an epistemic subject as an external stabilizer may be phenomenologically reduced to sets of possible sense-data (sensations, perceptions)—just as its behavior (on which the individual is supposed to calibrate its conceptual activities) may be reduced to certain regularities (generalizations, laws) among those sense-data:

If the phenomenalist is right, we have pre-analytically grouped qualia into a schema to which we refer as material objects. Some qualia of the schema are reliable indicators of the presence of others, and the entire schema serves as a construct by means of which various qualia can be interrelated by laws. (Hardin 1959, 527)

I may use my system of laws to decide that I have made a memory mistake, and withdraw a previously accepted statement, just as Robinson Crusoe might correct an error in his diary by discovering it to be out of line with several other observations or regularities. (ibid., 524)

Thus, according to the idealist, an individual needs to check its applications of concepts, not necessarily for external correspondence, but rather for internal coherence. Where the realist posits, e.g., that, for an individual to be able to calibrate its concept of redness, there must be things in the real world that are red and keep their color, the idealist alleges that all there needs to be is certain patterns in the epistemic subject’s applications of the concept of ‘red,’ e.g.,:

1. For any temporally ordered set of sense-data S that falls under the concept of ‘blood’ there will be a sense-datum s ∈ S that falls under the concept of ‘red.’

2. For any temporally ordered set of sense-data S that falls under the concept of ‘blood’ and any sense-datum s ∈ S that falls under the concept of ‘red’ there will be a sense-datum s* ∈ S earlier or later than s that falls under the concept of ‘red.’

Given such regularities, the epistemic subject may well be in a position to verify single applications of the concept of ‘red,’ namely, by checking whether they fit with other applications of concepts according to those regularities—and this, says the idealist, is all that is required for an individual to master the concept of redness.

The cogency of this proposal hinges, however, on the status of the supposed regularities. Two options must be considered here. Option one is that the individual first applies its concepts and then induces the regularities from those applications. In this case, however, the subject will not be able to check its applications against the regularities, because the regularities will themselves be determined by the applications and will, therefore, be disqualified as standards for the latter (cf. Nuttycombe 1965, 38).

Against this it might be objected that induced regularities are perhaps unsuitable as standards for the applications from which they have been induced; yet this in no way precludes consulting them as standards for future applications. This boils down to the second option: the individual first determines the regularities and then deduces applications of its concepts from those regularities. Essentially, this means that the individual treats the regularities as conceptually necessary and implicitly defines its concepts in terms of them. If, e.g., an individual imposes patterns (R1) and (R2) on its applications of the concept of ‘red,’ it will thereby define ‘red’ as the X such that (i) any temporally ordered set of sense-data S will fall under the concept of ‘blood’ only if there is a sense-datum s ∈ S that falls under X and (ii), for any temporally ordered set of sense-data S that falls under the concept of ‘blood,’ any sense-datum s ∈ S will fall under X only if there is a sense-datum s* ∈ S earlier or later than s that falls under X.

In this scenario, however, it will be impossible for the subject to verify the correctness of any single application of a concept. This is because the correctness of any separate application will be conditional on the correctness of other applications (cf. Pears 1988, 395). Having preset patterns (R1) and (R2), e.g., the correctness of an individual’s application of the concept of ‘blood’ to a given set of sense-data will depend on the correctness of its application of the concept of ‘red’ to some element in that set. Yet, the correctness of that application will, in turn, depend on the correctness of further applications of ‘red’ to other elements in the same set, and so on.

Against this it might be alleged that, in the present case, the individual will perhaps be unable to apply its concepts one by one; yet still, it should be able to apply them in a row. After all, even if an individual cannot apply the concepts of ‘blood’ and ‘red’ separately to (the elements in) a given set of sense-data, it should still be able to apply them to the set (respectively, its elements) simultaneously. In other words, the individual might be able to subsume, not sense-data under concepts, but sequences of sense-data under regularities. Alas, in the present scenario this is not a genuine possibility; for, there is for the individual no way of verifying the correctness of applications of such regularities. These applications could not be checked against the regularities themselves (as they are applications of those regularities); nor could they be checked against applications of single concepts (as there are, in the present scenario, no such applications).

Phenomenal regularities, then, will not enable the individual to check its applications of concepts. The only alternative would be to assume metaphysical laws among sense-data, against which an individual’s applications could be pitted. Yet, this option is not open to the idealist, as it is tantamount to realism. Assuming that an epistemic subject should calibrate its conceptual activities on sense-data which come into existence and behave according to laws that hold independently of the subject’s cognitive activities is, in effect, assuming that these sense-data are things (‘tropes’), or modes of things, that have an intrinsic capacity of stabilizing the subject’s use of concepts. It is assuming that the cognizer’s objects of experience confront the cognizer—indeed, any cognizer—as (modes of) intrinsically natured, mind-independent existences.

One challenge remains to be met, though. The skeptic might press the point that, even if our possessing concepts requires checking our applications of those concepts against real objects (and not merely subjective appearances), we will not be able to verify whether the objects against which we check our applications are real (or merely apparent). Therefore, we will not be able to know whether we truly possess any concepts:

[A]lthough Wittgenstein’s argument does show that we can never obey a rule privately, it leaves us in the uncomfortable position of never being able to tell whether we are really obeying rules or merely think we are doing so. For seeming to obey a rule would, for all an agent could tell, be indistinguishable from actually obeying one … And in such a situation we would not be able to know whether we really were acquainted with an external world or only with our own ideas. For someone to know that he really was following a rule, he would have to know that the framework that gave the rules life was real—but this involves a verificationistic assumption. (Swoyer 1977, 48)

Of course, for the realist, knowledge of the realness of things would follow from knowledge about our own possession of concepts together with the Wittgensteinian insight (which the skeptic meanwhile concedes) that our possessing concepts requires our calibrating their use on real objects. Thus, the realist cannot but reject the conclusion of the skeptical argument and cleave to the claim that we know for sure about our concept possession—at least in the sense of a first-order knowledge of the concepts we possess, if not in the sense of a higher-order knowledge that we possess those concepts. Nonetheless, the skeptic’s argument proposes a conundrum: Just how do we know of our concept possession—considering (as we must) that everything would appear the same to us phenomenally even if there was, in fact, no external world of independent objects, but just our internal microcosm of subjective experiences?

The realist may remind the skeptic at this point, however, that in the Wittgensteinian picture, knowledge of our own concepts—and of the realness of the framework against which we check our application of these concepts—is not phenomenally constituted at all, but is of an entirely distinct kind. It has already been shown to be a consequence of the private language argument that an individual’s first-order conceptual knowledge cannot be conceptually or perceptually mediated, but must be grounded in a direct acquaintance of the individual with the objects of its cognition. Such direct acquaintance, however, precisely because it is supposed to be direct, cannot be construed as phenomenal knowledge (or any kind of knowledge that places appearances before reality), but must be taken as the individual’s practical knowledge of how to interact with the relevant objects, which in turn consists in nothing other than its knowledgeable practical interaction with these objects itself. Now, such knowledge could not be acquired by the individual if the objects of its interaction did not really exist. Thus, even if its entire lifeworld were to appear phenomenally the same to the individual in the idealist scenario, it would still cognize that world in a strikingly different way than it would in the realist scenario (if it were capable of any cognition at all under those circumstances). Hence, the putative possibility of total self-deception with respect to one’s own concept possession, as alleged by the skeptic, does not turn out to be a genuine possibility at all.

7 Epilogue: Veritas Vincit

In Section 5 I insinuated a question about the verisimilitude of cognition: Is it enough for conceptual agents to calibrate their use of concepts that they (learn to) recognize sufficiently many misapplications of concepts, or is it furthermore necessary that they (learn to) rectify sufficiently many of such misapplications? Put differently, is there any guarantee that conceptual agents develop correct conceptions of what they conceive—that they form true thoughts about the things they think about? In short, is there for them not only an abstract possibility, but also a real prospect of knowing the world?

This question harks back to a question about the force of rules. In Section 3 it was argued that normative rules are distinguished from natural laws by the fact that rule-followers are free to violate the rules they follow while retaining their identity as followers of those rules. However, this does not necessarily mean that the freedom of rule-followers is unlimited, and the force of rules exhausted in making the behavior of their followers appropriately assessable. Indeed, it is usually maintained that rules guide their followers in the sense that, if an individual x follows a rule R, x will tend to act in a way conducive to accordance with R which is at least partially explained by R (cf. Railton 2006, 8). The claim is not that rules compel their followers (in the way laws compel their subjects), but rather that there is a considerable (although indeterminable) likelihood that rule-followers will act according to the rules they follow. The crucial question now is whether it is essential for rules to guide their followers in this sense.

The Wittgensteinian answer to this question is that an individual must follow a rule by and large correctly in order to follow that rule at all. If, hypothetically, the individual followed a rule incorrectly for the most part, or correctly and incorrectly in almost equal parts, it would, as a matter of fact, be following another rule or none at all: “if rule became exception, and exception rule; or if both became phenomena of roughly equal frequency—our normal language-games would thereby lose their point” (PI, § 142). To put the point in terms of the use of concepts: if an individual constantly applied a concept to the wrong kinds of things, it would in fact be calibrating its use of this concept on those things. That is, it would be these things that stabilize the individual’s use. But then, the individual would simply apply that concept correctly to them—it would just be the individual’s concept for those things (rather than the others). Then again, if the individual were to apply a concept in a totally erratic way to the most different kinds of things, it would in fact not be calibrating at all. In the end, the individual could not even be said to have a concept. Hence, the possession of concepts is logically tied to the prevailing correctness of their use, with the result that judgments about whether something falls under a given concept cannot randomly or predominantly go wrong.

An elaborate answer would supplement these considerations with a holistic-dynamic epistemic theory, according to which conceptual schemes and systems of belief develop over time and converge in the limit to a state of true knowledge. For reasons of space, however, let me defer this task to another occasion and just summarize here the results of the previous investigation. In this essay I have proposed a transcendental argument for epistemological realism. I started out from the assumption that concepts are rules that agents use and follow in their cognitive (classificatory, inferential) activities and argued that this requires cognizers to be capable of practically calibrating their application of concepts on external objects that serve as independent standards for the correctness of their applications. Thus, it has been shown that the possibility of concept possession presupposes the existence of entities with an intrinsic capacity to stabilize an individual’s use of concepts. At the same time, the regulatory interplay of calibration and stabilization establishes an epistemic nexus between cognizer and reality that, although fallible and limited, is invulnerable to radical skepticism.

The upshot is an epistemological realism which ensures both the mind-independence and cognitive availability of the outer world and thus steers clear of the dilemma between metaphysical and epistemological idealism. In our very existence as conceptual creatures struggling to synthesize their experiences and integrate their actions, the world emerges as the antagonistic force that confers normative significance on our concepts and whose alternating resistance and compliance account for the varying success of our conceptual coping strategies. As I speculated at the beginning, something like this might have been what Dr. Johnson really wanted to show when he kicked the stone and refuted idealism.^[7]