We meet here a false doctrine largely due to a misleading metaphor. My known world is taken to be a construction built upon such and such foundations. It is argued, therefore, to be in principle a superstructure which rests upon these supports. You can go on adding to it no doubt, but only so long as the supports remain; and, unless they remain, the whole building comes down. But the doctrine, I have to contend, is untenable, and the metaphor ruinously inapplicable. The foundation in truth is provisional merely. In order to begin my construction I take the foundation as absolute – so much certainly is true. But that my construction continues to rest on the beginnings of my knowledge is a conclusion which does not follow. It does not follow that, if these are allowed to be fallible, the whole building collapses. For it is in another sense that my world rests upon the data of perception.
My experience is solid, not so far as it is a superstructure but so far as in short it is a system.
– F.H. Bradley, ‘On Truth and Coherence’, Mind, New Series, Vol.18, No.71 (July 1909), 329–342 (335).
Bradley and the analytic tradition
Within Anglophone philosophy, Bradley is probably best remembered as the indirect impetus behind the “analytic break”, the rebellion initiated by Bertrand Russell and G.E. Moore against British Idealism, of which he was the principal exponent. It’s likely, in fact, that Bradley is more frequently read by scholars of literature than by modern analytics: he was a significant influence upon T.S. Eliot, whose undefended doctoral thesis was dedicated to the philosopher. Indeed, if English-speaking philosophers have read anything of Bradley at all, it is the “meaningless” sentence about the Absolute exhibited by Ayer in Language, Truth and Logic (1936).
Bradley’s reputation as an obscurantist pariah is, however, profoundly unmerited. Not only is his prose markedly English – concise, ordered, and a model of clarity compared to that of Hegel – but his philosophy holds up today far better than his early analytic detractors’ attempts at establishing a foundationalist epistemology. This short article from Mind is a fascinating document of the debate that took place during analytic philosophy’s formative years. Bradley defends a coherence theory of truth whilst criticizing the assumptions underlying foundationalism put forward in essays by Russell and George Stout, then-editor of the famous journal. What’s astonishing is the irony with which the wheel has come full circle: coherentism has been thoroughly rehabilitated by Davidson, and the contradictions of foundationalism have been made apparent by Quine and Sellars. Indeed, Bradley’s article contains fore-echoes both of Quine’s second dogma, and of the later analytic pragmatists under Rorty’s sway, Robert Brandom and John McDowell.
In this post, I will read through Bradley’s essay and discuss some ways in which his arguments predict, but also fall short of, views expounded in this later tradition. I will pay particular attention to Sellars. The profound yet difficult arguments of ‘Empiricism and the Philosophy of Mind’ (1956) provide much of the impetus for subsequent analytic pragmatism: Rorty, Brandom and McDowell owe more to Sellars than to Quine or Davidson, though I shall try, also, to make some points of comparison with their work. Bradley’s argument is nonetheless quite distinct from this tradition because it focuses on the issue of fallibility in sense experience. The later philosophers are interested in the logic of empirical knowledge. Their concerns are hence more fundamental, and the issue of fallibility is relegated to a secondary concern.
Foundationalism vs. coherentism
Bradley’s discussion is limited to “a test of truth in the case of facts due to sensible perception and memory” (329). Both Russell and Stout deny that coherence provides such a test; that is, they deny that a fact is true by virtue of its concordance with other beliefs in a system of knowledge. Their opposing view is that certain atomistic facts – in this case, those of sense and perception – are true independently of their role within a system of knowledge; rather, these facts provide the objective foundations of such a system. This argument depends on the notion that certain judgements of sense are in principle infallible. Bradley, on the other hand, maintains that coherence within a comprehensive system of knowledge is the only possible test of truth: there are no independently true facts and no infallible judgements of sense.
Bradley makes clear that he believes sensation to furnish the basic materials for knowledge and that the facts of perception are, in part, non-rational (that is, independent of the mind). We hence cannot make ourselves independent of certain non-rational data. Yet, he argues, this does not mean that we have any access to independent facts or to judgements free from error. In the rest of his article, Bradley attempts to demonstrate this by rebutting the two principle arguments in favour of independent facts and infallible judgements, namely:
(i) Such data can be shown.
(ii) Our intelligence cannot function without them. Even if (i) is not true, we are bound to assume certain independent facts, for in their absence, the system of our knowledge would have no foundation and we would descend into scepticism.
The first group of arguments pertain to individual sense judgements: how can a judgement be (a) infallible and (b) an ‘objective’ report of an independent fact? The second group of arguments oppose foundationalism to coherentism, and thus are interested in the way that individual judgements relate to each other within a system.
Bradley deals with (i) by arguing that we would need particular judgements of perception if these were to count as infallible truths: the simple fact of fundamental sensations and their combinations would be insufficient for a foundationalist programme; yet, as soon as judgements are particularized, we encounter problems of interpretation that render them fallible.
The trajectory of Bradley’s argument here is quite clear, but it is curious that he focuses primarily on fallibility, and only briefly elaborates the intervening step in his argument – that is, the point about the insufficiency of fundamental sensations. Indeed, it is from an investigation of this point that the most profound insights of analytic pragmatism arise. I will now summarize Bradley’s argument, whilst contrasting it with the more elaborate discussion of sensations’ insufficiency offered by Sellars.
The Myth of the Given
Bradley first claims that the initial proposition (that certain sensible data can be indubitably shown) requires some clarification. What does it actually mean for such data to be shown? He writes: “In the case of any datum of sensation or feeling to prove that we have this wholly unmodified by what is called ‘apperception’ seems a hopeless undertaking” (331). Having asserted this, he then concludes that (i) is in fact claiming that we, within our particular mode of apperception, have access to verifiable facts of perception and memory, and judgements free from error. This is the claim Bradley will then go on to refute.
What’s interesting, however, is just how much is at stake in Bradley’s assertion. It effectively implies a version of Sellars’ Myth of the Given, although in Bradley’s case it is not at all worked out. Sellars aims to show the impossibility of claiming that knowledge is given in perception apart from conceptual activity, which is to say, apart from inference: in order to function as knowledge, basic sensations must be subsumed within a system of inferences, which in turn deprives them of their epistemic independence. If this holds, it is sufficient to deal a significant blow to foundationalism.
Let’s consider Sellars’ discussion in ‘Empiricism and the Philosophy of Mind’ (1956) in more detail, before returning to Bradley. It is a sophisticated argument against foundationalism that deals with the way empirical knowledge is logically structured, rather than with the idea of fallibility.
A foundationalist system infers knowledge from a set of atomic facts. Sellars names such facts “the given”, a notion which he then shows to be fraught with paradox. Given facts must fulfil two conditions. Firstly, if they are to count as a foundation, they must be independent, ie. not inferred from other facts. Secondly, in order to be combined into other judgements, it must be possible to draw inferences from them, ie. they must be propositionally structured. These two conditions appear to be incompatible. The logic of propositional statements means applying concepts to objects, but for this reason such statements can never be truly independent: concepts only make sense within an inferential network of other concepts. “The given” thus falls short of either the first criterion – because its propositionality, entailing the presupposition of acquired knowledge, precludes its independence – or of the second, because if it renounces its propositionality, it is no longer possible to infer anything useful from it.
Sellars rejects sense data as a candidate “given” because they are not propositionally structured. Importantly, he does not claim that concepts are at work in basic sense reception; rather, the process by which such data can meaningfully be said to contribute to knowledge necessitates the acquisition of concepts. I will discuss this aspect of his argument (how knowledge proper can arise from a preconceptual ability to make differentiated responses to sense contents) in the second half of this post. For now, it suffices to grasp the big picture: that attempts to render sense data propositional involve the assumption of further epistemic statements. Sellars writes, for example, that
if a sense-datum philosopher takes the ability to sense sense contents to be unacquired, he is clearly precluded from offering an analysis of x senses a sense content which presupposes acquired abilities. It follows that he could analyze x senses red sense content s as x non-inferentially knows that s is red only if he is prepared to admit that the ability to have such non-inferential knowledge as that, for example, a red sense content is red, is itself unacquired. And this brings us face to face with the fact that most empirically minded philosophers are strongly inclined to think all classificatory consciousness, all knowledge that something is thus-and-so, or, in logicians’ jargon, all subsumption of particulars under universals, involves learning, concept formation, even the use of symbols. (‘Empiricism and the Philosophy of Mind’, §6)
Foundationalism runs into problems, then, not because of fallibility but because of the logic of empirical knowledge, which demands an inferential, propositional form. Sellars expresses this later in his essay with reference to the “space of reasons”:
The essential point is that in characterizing an episode or a state as that of knowing, we are not giving an empirical description of that episode or state; we are placing it in the logical space of reasons, of justifying and being able to justify what one says.” (§36)
The world can thus be divided into two distinct logical spaces: reasons and causes. Davidson echoes this when he writes that “nothing can count as a reason for holding a belief except another belief” (‘A Coherence Theory of Truth and Knowledge’). A sense datum can stand in a causal relation to beliefs, but this says nothing about whether the belief is justified. Likewise, a belief can stand in a causal relation to an action (see Davidson’s early paper ‘Actions, Reasons and Causes’), but this is just one description of that belief: it can also be described within a different, normative vocabulary, dealing entirely with the process of justification. As Sellars points out, knowledge must be understood within this second, normative domain. Foundationalism, wishing to derive complex beliefs from a limited set of atomic facts, must also subscribe to this view of knowledge. Yet, as soon as the leap to this domain is made, simple causation gives way to the web of justification.
Fallibility: a mistake
We can contrast Sellars’ Myth with Bradley’s attack on foundationalism. Recall that Bradley seeks to deny that we have access to verifiable facts of perception and memory, and to infallible judgements. He thus wishes to find a source of fallibility. In doing so, he points to the fact that sense data alone are insufficient to count as a foundation for knowledge, meaning to show that particularized judgements, strong enough to count as knowledge, can in fact be fallible.
In making the point that ‘bare’ sense data cannot count as knowledge, Bradley does something far simpler than Sellars: he reminds us that we are talking about human perception, from which he makes further arguments about how we relate sensation to knowledge. Bradley argues briefly that for sensory experience to become a candidate for knowledge we must be dealing with facts on the level of “I am here and now having a sensation or complex of sensations of such and such a kind” (332). The bulk of his treatment of (i) then involves unpacking the inadequacies of such facts, which take the form of propositions on a higher level than mere reports of sensation: if “I” refers to a self, memory is involved and judgement hence becomes fallible; the deictics of “here” and “now” pose additional problems; sense hallucination always remains a possibility.
This section of Bradley’s argument strikes me as particularly off-target: his primary concern here is with fallibility in reports of judgement, rather than the logical structure of empirical judgement itself. And yet, to get to the position where he’s criticizing statements of the kind “I am here and now, &c.”, Bradley has to justify some much more fundamental propositions about just how a sensation becomes a candidate for knowledge, and why we are only permitted to talk of knowledge on this higher propositional level. That he cannot see the trees for the wood, as it were, testifies to the surprising weight afforded to notions of the (in)fallibility of sense experience at this stage in analytic philosophy’s development. It also shows exactly where Sellars’ insights fit into the discussion.
The important (and prescient) points in Bradley’s treatment of (i) essentially remain in the background, both in his introduction to this section of his essay, or where he insists, for example, that “[e]verywhere such fact depends upon construction” (332). It’s exactly these arguments that need working out, and from which the more profound (Sellarsian) conclusions follow. Bradley is worth quoting in full here:
(a) If we take the instance of simple unrelated sensations or feelings, a, b, c – supposing that there are such things – what judgement would such a fact enable us to deny? We could on the strength of this fact deny the denial that a, b or c exist in any way, manner or sense. But surely this is not the kind of independent fact of which we are in search.
(b) From this let us pass to the case of a complex feeling containing, at once and together, both a and b. On the ground of this we can deny the statement that a and b cannot or do not ever anyhow co-exist in feeling. This is an advance, but it surely leaves us far short of our goal.
(c) What we want, I presume, is something that at once is infallible and that also can be called a particular fact of perception or memory. And we want, in the case of perception, something that would be called a fact for observation. We do not seem to reach this fact until we arrive somewhere about the level of “I am here and now having a sensation or complex of sensations of such or such a kind.” The goal is reached, but at this point, unfortunately, the judgement has become fallible, so far at least as it really states particular truth. (331-2)
All Bradley really provides here is a basic picture of the inadequacy of sense impressions in themselves: if we attempt to think of an “atomic” sensory experience, unrelated to others (and to other concepts) – that is, exactly the kind of irreducible, epistemically-independent datum upon which foundationalism relies – then this does not let us assert anything other than the fact that such sensations occur; likewise, the coincidence of two such sensations allows us only to assert that two such sensations can occur together. Bradley does not press the question as to why these two assertions do not provide the kind of atomic facts adequate to found a system of knowledge; rather, he takes this as self-evident, and so his transition into (c), along with its central assertion of just what kinds of facts we are looking for, is somewhat murky.
Nonetheless, so much is present here in nuce, because it’s exactly by pressing this question that we arrive at the idea of the Myth of the Given – a myth, because the “given” fails to overcome the paradox of being both epistemically independent and a meaningful candidate for knowledge via its propositional nature. Bradley’s (a) and (b) suggest the emptiness of immediate sensations if they are not (in Kantian terms) subject to determination by a specific concept, which would entail their dependence and hence non-candidacy as atomic facts. I don’t take this to be his point, but I find it an interesting feature of the argument; rather, Bradley wants to show that the experiences of (a) and (b) are not propositional in the right way. And this is the real crux of the matter: what does being propositional in the right way have to do with being knowledge, and why does this counteract the possibility of epistemic independence?
Again, Bradley’s answer seems to miss the more fundamental issue at stake, and in a rather particular fashion. For him, the propositionality of knowledge is entailed by the fact that “the perceived truth, to be of any use, must be particularised” (333). By this, he does not demand the conceptual specification of ur-sensations, but instead that a (potentially) infallible judgement have the form “I am here and now, &c.” This does, of course, bring with it a sufficient degree of conceptual specification, but Bradley is not interested in this half of the dilemma: he wants to show the fallibility of “I”, “here”, and “now”. This, I think, is peculiar, as is it not so much the structure of empirical knowledge that is being analyzed here, but infallible judgement. Bradley’s reasoning is clearly motivated by a kind of Cartesian scepticism. Pace Sellars, Bradley does not claim that knowledge should have a propositional structure as it belongs to a logical space of reasons; his point is not that conceptuality is inferential in a way that raw sense experience is not. For Bradley, the question is not how the report “this is green” of a foundational sensation could function as knowledge, but rather under which conditions it could count as infallible.
Upon Sellars’ close examination, the propositional structure of empirical knowledge seems to offer a sufficient challenge to foundationalism in itself, quite apart from the question of the fallibility of propositional reports, which Bradley places centre stage. Nonetheless, Sellars’ arguments for the insufficiency of sense data bolster Bradley’s point that reports of such data (“I think this is green”) are in principle fallible, and thus are not the facts that Russell and Stout are looking for.
Towards a coherentist epistemology
Having concluded his argument against (i) by dismissing the reliability of atomic facts sought after by foundationalists, Bradley is still left with a broader challenge: to describe a system of knowledge that does not need such foundations – that is, a properly coherentist epistemology. The second part of Bradley’s essay is devoted to arguing against (ii): that our intelligence relies upon atomic facts, lest it fall into an infinite regress. I will now review his arguments against this proposition, whilst noting further points of continuity with both Quine and Sellars, and exploring in further detail Sellars’ account of concept acquisition. Finally, I will argue that Bradley’s approach, concerned with fallibility and not the causal/normative distinction, does not allow him to give a satisfactory answer to the question of sense reports’ authority in empirical judgements. He attempts to answer an objection along these lines, but does not formulate the issue adequately. It is, however, a question that Sellars formulates and fails to answer clearly himself, the most sustained attempt to do so having been made by John McDowell in Mind and World (1994).
Bradley’s fundamental thesis is given in the paragraph cited at the beginning of this post. He rejects both the conclusion that we must assume atomic facts, and the premise that our knowledge otherwise has nothing to stand on. Instead, Bradley proposes that knowledge is secure in so far as it builds a system; it must not repose upon a certain base of facts.
Bradley begins his discussion, intriguingly, with the notion of error. Even though he has already shown that there are, in principle, no infallible judgements, Bradley now asks whether we need a certain set of infallible judgements to go about the business of building a system of knowledge. If we did, it would follow that knowledge as such would be impossible. What does it mean, then, to say that our knowledge is secure, even if all facts are in principle capable of being “relegated to the world of error”?
Bradley takes pains to point out that he relies upon sense experience, that he must return to the world, both to learn anew and to confirm what knowledge he has. His position is not a rejection of sense judgements per se. Even though sense judgements are all in principle fallible, it is not possible to imagine correcting all such judgements. Bradley writes that “I cannot […] imagine the world of my experience to be so modified, that in the end none of these accepted facts should be left standing.” Yet, crucially, he goes on: “There is still a chasm between such admissions and the conclusion that there are judgments of sense which possess truth absolute and infallible” (335). That we cannot imagine overturning all sense judgments is no argument for the infallible truth of some such judgements, or even the necessity of assuming such a set. Consider that it is possible to imagine a certain number of sense judgements to be modified (just not all of them), and for the facts to remain standing. Thus, we can say that all facts may in principle be erroneous, because in practice, facts tend to be accepted or rejected on an individual basis.
On Bradley’s conception, then, facts are true “just so far as they work, just so far as they contribute to the order of experience” (336). Likewise, other facts are to be classified as errors if they disrupt the order of experience. Again, Bradley admits his reliance on sense experience, but does not see this as a sufficient defence of foundationlism:
Certainly there are truths with which I begin and which I personally never have to discard, and which therefore remain in fact as members of my known world. And of some of these certainly it may be said that without them I should not know how to order my knowledge. But it is quite another thing to maintain that all and every single one of these judgements is in principle infallible. The absolute indispensable fact is in my view the mere creature of false theory. […] A foundation used at the beginning does not in short mean something fundamental at the end, and there is no single “fact” which in the end can be called fundamental absolutely. It is all a question of relative contribution to my known world-order. (336)
Note that there is an important continuity between foundationalism and coherentism in the way that they assess a novel proposition’s veracity. Neither refers this proposition ‘directly to the world’, so to say, in order to assess it; rather, both check the new proposition against beliefs that we already hold to be true. The two approaches differ in that foundationalism holds a certain set of beliefs to be infallible: they are the ultimate destination of the referral. Coherentism, on the other hand, privileges no set of beliefs as infallibile. A proposition is referred to the world-order as a whole.
Bradley’s case against the common foundationalist critique of coherentism – it reposes on an infinite regress – is that this simply is not how knowledge is structured. To say that knowledge does not require foundations of this sort is not simply question begging, in order to avoid the idea of a regress. Rather, it is an important thesis about the way justification occurs holistically and not merely sequentially. The foundationalist enterprise involves tracing a fact’s justification back linearly, along a chain of other facts, until a certain ground is reached. Applying this model to coherentism, it seems that the chain either runs in a circle or falls eternally into a void. The point is, however, that justification should not be understood like a chain; rather, a belief is referred to the system itself. The regress objection is a misunderstanding of coherentism, and question-begging in its own right, as it effectively assumes the foundationalist approach to be correct.
Concept acquisition
Nonetheless, it is tempting to ask how, and whence, a coherent world-order arose in the first place. This is not a subject that Bradley broaches, but it is hard to avoiding posing the question to Sellars. As we saw earlier, Sellars’ key insight is that knowledge, properly speaking, belongs to the justificatory space of reasons. A single proposition cannot be independent because it takes part in conceptual networks of inference. As Sellars himself puts it, “one couldn’t have observational knowledge of any fact unless one knew many other things as well” (§36). The point is not that sense data themselves are inferentially structured, but that for them to count as knowledge, they have to be inferentially articulated. The passage from a causal, non-inferential domain to a conceptual domain in which knowledge claims can be disputed involves moving from one logical space to another. We thus might reasonably ask just how this movement takes place. The question can be understood in two ways. Firstly, there is a developmental perspective: how do we go about acquiring concepts? Can we give an account of how our conceptual apparatus might arise from mere sense impressions without falling for some version of the Myth of the Given? Sellars implies such an account, but still leaves open a second, more fundamental question: can we ever close the logical gap between our beliefs about the world and our sense impressions of the world? Since belief is entirely a matter of rational justification (only a belief may justify another belief), what gives sense experience its authority in supporting beliefs?
I will now attempt to summarize Sellars’ account of concept acquisition. ‘Empiricism and the Philosophy of Mind’ does not present it in this way; rather, Sellars provides us with a persuasive critique of foundationalism, and an equally persuasive account of how “knowledge” involves a linguistic, social space that leads to a coherentist perspective. He does not set out a theory of concept acquisition tout court, and focuses explicitly only on the acquisition of concepts of inner episodes in the second half of his essay. My reading of his more general understanding of concept acquisition is indebted to Robert Brandom’s remarks in his study guide, as well as his essay on Sellars in Tales of the Mighty Dead (2002). The second problem posed by coherentism – the logical gap between the domain of belief and that of experience – I will return to at the end of this post. Sellars does not deal with it satisfactorily, but it also proves problematic for Bradley.
The opening sections of Sellars’ essay constitute a critique of two forms of foundationalism: traditional empiricism and rationalism. Traditional empiricism is criticized for employing versions of the Myth of the Given. Rationalism, on the other hand, is deflated. Sellars construes a version of rationalism that sees judgements of the form “x looks φ” as prior to “x is φ”: in this way, knowledge about the world is to be founded on the ways in which it appears to subjects. Sellars shows that such rationalists are confused, and that all looks judgements presuppose is judgements: we can imagine a situation in which a certain object looks blue, but when held up in a different light, it looks green. When we say it “looks” a certain way, we are in fact just failing to endorse an is claim (and in doing so, presupposing a notion of something being a certain way, not just looking that way).
Sellars’ argument against this kind of rationalism is a key step towards his notion of the “space of reasons” that defines knowledge. He goes on to extrapolate that an is judgement involves a level of confidence absent from a looks judgement: I am confident that my perception can be correlated with that of other subjects in optimal observing conditions. In making this point, Sellars implies what Brandom terms a “two-ply” account of observation. The first level involves the simple ability to discriminate between stimuli and respond to them. This does not constitute any kind of awareness that we might term “knowledge”. Rather, knowledge (and conceptuality) emerges only on the second level, in which responses take on inferential roles. Thus, Sellars suggests that a necessary first step towards concept formation is an ability to differentiate and emit responses to sense contents – a simple kind of learned response mechanism, such as a parrot saying “this is red”. Such a notion is implicit in passages like this:
Now, it just won’t do to reply that to have the concept of green, to know what it is for something to be green, it is sufficient to respond when one is in point of fact in standard conditions, to green objects with the vocable “This is green.” Not only must the conditions be of a sort that is appropriate for determining the color of an object by looking, the subject must know that conditions of this sort are appropriate. And while this does not imply that one must have concepts before one has them, it does imply that one can have the concept of green only by having a whole battery of concepts of which it is one element. It implies that while the process of acquiring the concept of green may – indeed does – involve a long history of acquiring piecemeal habits of response to various objects in various circumstances, there is an important sense in which one has no concept pertaining to the observable properties of physical objects in Space and Time unless one has them all — and, indeed, as we shall see, a great deal more besides. (§19)
This “two-ply” account of observation thus involves a distinction between sentience and sapience (to use Brandom’s terms). Sellars admits the possibility of organisms’ differentiating between different sense contents, but imposes important strictures on our referring to these responses as concept-use (and hence knowledge). Ultimately, his aim is not to provide an account of how concepts emerge from differential responses (hence the implicit nature of his model of concept acquisition); rather, Sellars’ goal is to underline the gap between the level of sentience and that of sapience, thus confirming his coherentist perspective. He is very clear about this gap, ultimately pointing out that sapience is entirely “a linguistic affair”:
It clearly makes all the difference in the world how [an association between similar sense contents] is conceived. For if the formation of the association involves not only the occurrence of resembling particulars, but also the occurrence of the awareness that they are resembling particulars, then the givenness of determinate kinds or repeatables, say crimson, is merely being replaced by the givenness of facts of the form x resembles y, and we are back with an unacquired ability to be aware of repeatables, in this case the repeatable resemblance. Even more obviously, if the formation of the association involves not only the occurrence of red particulars, but the awareness that they are red, then the conceptualistic form of the myth has merely been replaced by a realistic version, as in the classical sense-datum theory.
If, however, the association is not mediated by the awareness of facts either of the form x resembles y, or of the form x is f, then we have a view of the general type which I will call psychological nominalism, according to which all awareness of sorts, resemblances, facts, etc., in short, all awareness of abstract entities – indeed, all awareness even of particulars – is a linguistic affair. According to it, not even the awareness of such sorts, resemblances, and facts as pertain to so-called immediate experience is presupposed by the process of acquiring the use of a language. (§29)
Sellars’ basic point here is that when traditional empiricism attempts to account for concepts via a process of abstraction from particulars, it implies an awareness of their similarities that is itself not preconceptual. The Myth of the Given returns as empiricism lets sapience seep back into sentience. Sellars, on the other hand, maintains that sapience simply cannot travel that way: facts are not given in perception, but only emerge along with the language that articulates them.
This has an important consequence for Sellars’ understanding of knowledge: his picture is not just coherentist, but social. Not only is language acquired socially, but the “space of reasons” in which facts are disputed is also a social space. As we have seen, judgements that employ “is”, rather than “looks”, assume a number of facts about the conditions of observation that relate one observer to others: to endorse such a statement fully, I have to be aware of the conditions in which I make it (if the conditions are abnormal, I could be mistaken). Likewise, in inferring “there’s something green over there” from another observer’s reports, I have to assume her competence in making such a statement – something I could presumably verify by checking for green objects myself. The point is that, for us to say that a certain subject has observational knowledge, two factors must obtain: (a) their reports are reliable; (b) they are aware that they are using them reliably. Or, as Sellars puts it:
for a Konstatierung “This is green” to “express observational knowledge,” not only must it be a symptom or sign of the presence of a green object in standard conditions, but the perceiver must know that tokens of “This is green” are symptoms of the presence of green objects in conditions which are standard for visual perception. (§35)
We can quite easily imagine (a) to be true, for example, of a parrot, but would not go so far as to attribute it awareness of its reliability. Thus, Sellars places an important stricture on knowledge: it is not just about naming concepts, but using them and knowing how to use them.
Sellars presents the acquisition of concept use as something that emerges gradually, and, because of its linguistic nature, holistically from the basic level of making differentiated responses to stimuli. An observer can begin to make an association before being aware of this association in a way that would constitute knowledge; they may even report it and remember it. But this was, in Sellars’ view, only sentience and not sapience. The latter, he portrays as a phenomenon emergent from the former:
Thus, all that the view I am defending requires is that no tokening by S now of “This is green” is to count as “expressing observational knowledge” unless it is also correct to say of S that he now knows the appropriate fact of the form X is a reliable symptom of Y, namely that (and again I oversimplify) utterances of “This is green” are reliable indicators of the presence of green objects in standard conditions of perception. And while the correctness of this statement about Jones requires that Jones could now cite prior particular facts as evidence for the idea that these utterances are reliable indicators, it requires only that it is correct to say that Jones now knows, thus remembers, that these particular facts did obtain. It does not require that it be correct to say that at the time these facts did obtain he then knew them to obtain. And the regress disappears. (§37)
Unlike the traditional empiricism Sellars criticizes, he does not allow anything from this sapient level to have a role in merely sentient sense impressions. A gap remains between the two, and if it is mysterious, then its mysteries are those of language. Concept-use emerges as a coherent system and cannot be understood as composed of independent elements. As Wittgenstein writes, “Das Licht geht nach und nach über das Ganze auf” (Über die Gewissheit, 141).
Bradley, Sellars and Quine
This account of conceptuality undermines a simplistically foundationalist picture of knowledge. Sellars is happy to concede that, in a certain sense, our knowledge rests on observation reports which are not inferred from other propositions. Nonetheless, he disputes the notion of foundation, for “if there is a logical dimension in which other empirical propositions rest on observation reports, there is another logical dimension in which the latter rest on the former” (§38). Sellars’ thus sees our knowledge as a system. But, if he rejects the idea that it stands on a simple “foundation” of some kind, he also rejects the idea that it is wholly self-supporting (regressive or circular):
One seems forced to choose between the picture of an elephant which rests on a tortoise (What supports the tortoise?) and the picture of a great Hegelian serpent of knowledge with its tail in its mouth (Where does it begin?). Neither will do. For empirical knowledge, like its sophisticated extension, science, is rational, not because it has a foundation but because it is a self-correcting enterprise which can put any claim in jeopardy, though not all at once. (§38)
This long Sellarsian excursus has, then, led us right back to Bradley, whose picture of the system of knowledge is remarkably similar. Compare the previous paragraph from Sellars with this from Bradley:
the observed fact must agree with our world as already arranged, or at least must not upset this. If the fact is too much contrary to our arranged world we provisionally reject it. We eventually accept the fact only when after confirmation the hypothesis of its error becomes still more ruinous. We are forced then more or less to rearrange our world, and more or less perhaps to reject some previous “facts”. The question throughout is as to what is better or worse for our order as a whole. (337)
Thus, the key interaction is not between a fact and the foundations of a system, but between a new fact and the system as a whole. This new fact thus can be rejected, accepted, or accepted whilst inducing a modification of the system. Bradley points out that the authority of memory does not arise from its infallibility, but from the disorder that would be introduced into my world were I systematically to distrust my own memories. Nonetheless, in the right conditions, I could admit that I was mistaken. A similar process, he argues, is at work in the facts of history.
Bradley’s focus on the interaction of a new fact with a whole system also brings him very close to Quine’s famous “forcefield” metaphor:
The totality of our so-called knowledge or beliefs, from the most casual matters of geography and history to the profoundest laws of atomic physics or even of pure mathematics and logic, is a man-made fabric which impinges on experience only along the edges. Or, to change the figure, total science is like a field of force whose boundary conditions are experience. A conflict with experience at the periphery occasions readjustments in the interior of the field. Truth values have to be redistributed over some of our statements. Re-evaluation of some statements entails re-evaluation of others, because of their logical interconnections – the logical laws being in turn simply certain further statements of the system, certain further elements of the field. Having re-evaluated one statement we must re-evaluate some others, whether they be statements logically connected with the first or whether they be the statements of logical connections themselves. But the total field is so undetermined by its boundary conditions, experience, that there is much latitude of choice as to what statements to re-evaluate in the light of any single contrary experience. No particular experiences are linked with any particular statements in the interior of the field, except indirectly through considerations of equilibrium affecting the field as a whole. (‘Two Dogmas of Empiricism’)
Whilst Bradley may still be thinking in terms of fallibility, which makes him look outdated next to Quine and Sellars, the fundamental model of knowledge that he offers is extremely similar. Whence the irony that was the impetus for this post: analytic philosophy has essentially come to recapitulate the views against which it, incipiently, rebelled.
An objection
The interest of Bradley’s essay lies not so much in the strength of its arguments as in the moments that it prefigures later analytic philosophy. Indeed, there are some passages that are quite poorly reasoned and lead me into a rare sympathy with the likes of Ayer. One of these comes towards the end, where Bradley deals with a potential objection:
“But,” it may still be objected, “my fancy is unlimited. I can therefore invent an imaginary world even more orderly than my known world. And further this fanciful arrangement might possibly be made so wide that the world of perception would become for me in comparison small and inconsiderable. Hence, my perceived world, so far as not supporting my fancied arrangement, might be included within it as error. Such a consequence would or might lead to confusion in theory and to disaster in practice. And yet the result follows from your view inevitably, unless after all you fall back upon the certainty of perception.” (338)
Bradley’s response to this follows a tidy internal logic, but is ultimately a bizarre piece of philosophy. Both he and his imagined interlocutor play fast and loose with numerous categories. Bradley writes that a system of knowledge includes “all possible material.” Hence, the imagined world is to be weighed not just against the perceived world, but also against all other imaginable worlds: “Not only must you include everything to be gained from immediate experience and perception, but you must also be ready to act on the same principle with regard to fancy” (339). If this is the case, then every imaginary modification is to be weighed against its imaginary opposite, and the sum of all these imagined worlds comes to zero. Thus, only the data of immediate experience remain.
Bradley says that this objection is a misunderstanding of coherence and that “[t]he aspect of comprehensiveness has not received here its due emphasis” (339). I presume this means that a subject should weigh all the data available to them when accepting and rejecting judgements (so, immediate sense data are weighed against judgements I also hold to be true). But how do we meaningfully interpolate imagination into this? Bradley seems to think that his highly imaginative subject is at all times aware of what belongs to fancy and what belongs to perception, and chooses to believe in fancy because of its greater order. He thus adopts a specific set of fanciful beliefs. Yet, Bradley implies, if he is aware of the distinction between fancy and perception, then he should know that his fanciful set of beliefs are fanciful and, for all their order, have no priority over an equally fanciful set of totally opposing beliefs. (Of course, the criterion for selection of beliefs is coherence, and who is to say that the totally opposing beliefs would be equally orderly? The relation of these terms is here too ill-defined to warrant a proof of the contrary, but Bradley misses this point.) Still, Bradley begs the question by reinforcing a distinction between fancy and reality, in which reality is already assigned a justificatory privilege. He compels his imaginative subject to think along the following lines: “I know I’m imagining p. But I could equally well be imagining ¬p. I have to take both of these possibilities into account, so – aw, shucks! – that just leaves the data of sense perception.”
Imagination is here understood as a willful aberration that knows in advance it lacks solid justification. This is a fundamentally mistaken way of construing ‘imaginative’ belief. Everything we know about faith (or cognitive dissonance) shows that coherent, reality-opposed beliefs do not weigh themselves against their opposites. Coherence wins out and the world goes to hell. Subjects simply are not in a state of transparent awareness regarding what they have imagined (and thus could imagine otherwise) and what they see before them. Yet I can only make sense of Bradley’s response by reading him as assuming the opposite: that I know when I imagine, and hence know I could imagine otherwise.
Maybe I haven’t understood Bradley’s fanciful interchange, but it seems to me that he misses the point of his own objection. The thrust of the objection is that if coherentism does not anchor beliefs in sense judgements, then it is possible to form a coherent system completely independent from such judgements – indeed, in direct contradiction to them. What matters, after all, is the coherence of that system. And this objection is very serious. It goes right to the heart of the logical gap we saw emerge earlier between our beliefs about the world and our sense impressions of it. For a strict (Davidsonian) coherentist who holds that only a belief may justify another belief, it is important to account for the justificatory role sense impressions take on when they cause beliefs about the world.
Whilst I may be misunderstanding Bradley, my feeling is that, because he did not conceive of these issues in terms of Sellarsian logical spaces, he failed to formulate this particular objection as clearly as he could have. In fact, he failed to see its importance, and certainly did not answer it convincingly. Yet, by pointing to the privileged role that sense data undeniably have in our justificatory practices, Bradley anticipated Sellars again, who, as I pointed out, admits as much in §38 of his famous essay – he just does not construe sense impressions as a foundation sensu stricto.
Sellars might thus be seen as advocating a non-traditional form of empiricism. Indeed, this is John McDowell’s interpretation. In this essay he criticizes Brandom’s cleaned-up account of concept acquisition in Sellars, which I more or less followed earlier. I certainly find Brandom helpful in clarifying many aspects of Sellars, particularly his argument that we may come to have knowledge of things to which we were only previously responding (the sentience/sapience distinction). Yet, McDowell is right that Brandom reads a lot of his own philosophy into passages of Sellars that do not aim to give an explicit account of concept acquisition, despite implying one. Furthermore, this version of Sellars’ coherentism reproduces the ‘logical gap’ in an implied distinction between reception and articulation. It is clear that concepts are at work in articulated knowledge, but what about sense reception itself? Can I, as a linguistically-adept adult, continue to divide my experience into sentience (things I see but do not bother to articulate propositionally) and sapience (specific propositional thoughts I draw from my experience)? Brandom seems to imply that I could, but this seems very remote from my phenomenology of perception.
The endpoint of the dialogue between Bradley, Sellars et al. is, I think, McDowell, who takes up this quasi-phenomenological perspective and attempts to close the aforementioned gap by arguing, daringly, that spontaneity is already at work in reception. This profound thesis about the interpenetration of the normative and natural, the subject of Mind and World (1994), takes us far beyond what Bradley offers in this essay. Alas, it also takes me beyond the possibilities of a single post, which I believe I have already stretched (perhaps along with the reader’s patience) beyond breaking point. Such are the dangers of holism.
Besides his own, the only English versions of Nabokov’s Russian poems are those of his son Dmitri, included in the Penguin Collected Poems. However, whilst Dmitri Nabokov’s skill at imitating paternal practice contributed greatly to the authority of his prose translations, in the case of the poetry, English readers are left with a very narrow window onto Nabokov’s early, lyric works. Nabokov’s own views on translation were highly idiosyncratic. Apotheosized in his edition of Evgeny Onegin, he strove for “rigid fidelity” to the original, seeking to capture the semantic nuances of words whilst abandoning the formal and phonic features of a text. His Pushkin thus reads like an elegant crib.
My translation attempts to replicate Nabokov’s verse form as faithfully as possible, preserving both rhyme scheme and metre, with as little divergence from the actual semantic content as I could manage. The highly-synthetic nature of Russian means that it often requires more syllables to express an idea than English. Thus, maintaining tetrameter sometimes calls for some (strictly speaking, pleonastic) additions. Of course, this can sometimes aid in the search for rhymes. The greatest challenge was the third stanza, which breaks with the ABCB rhyme scheme for something much tighter. I had no choice but to add significant embellishments, which I hope do not dilute the translation too much.
Amongst these was Nelson Goodman’s Languages of Art (1968), a classic of analytic aesthetics. (Some of the others would be Philosophy and The Mirror of Nature, The Claim of Reason, and Wahrheit und Methode.) The book’s aura was enhanced by the emotional circumstances in which I acquired it – at the Librairie philosophique J. Vrin on the Place de la Sorbonne, whilst taking the ENS entrance exams. I shan’t go into detail about the concours, but it was an immensely stressful week, and somewhat eye-opening, given the variety of people who presented themselves. Trying to get on with all these people, against whom I was effectively in competition, added to the intensity of the experience, but I had some wonderful discussions in doing so, particularly with a philosophy student who recommended the Goodman book to me.
We should now consider the two modes of exemplification. At first blush, this seems rather more complex than denotation: in exemplification, a symbol corresponds to a predicate that could itself be used to denote the symbol. Just as in denotation, a symbol refers to something, but in this case, the referent is a predicate that belongs to the symbol itself. To put it differently, the symbol possesses the qualities to which it refers. Goodman thus offers as an example of literal exemplification a tailor’s swatch: “Exemplification is possession plus reference […]. The swatch exemplifies only those properties that it both has and refers to” (53). The specific qualities to which the symbol refers are determined by the symbol system in which it is used (conventionally, the swatch acts as an example of the texture of a particular material, and not of the property of rectangularity). Thus, the difference between exemplification and denotation is that the former deals only with predicates. As Goodman writes, “anything may be denoted, only labels may be exemplified” (57).
In metaphorical denotation, an arbitrary symbol is used to describe an object. Metaphorical exemplification, on the other hand, means that a predicate is attached to a symbol that possesses this predicate itself. This allows us to talk of a “sad picture”, for example. The picture “metaphorically exemplifies sadness if some label coextensive with ‘sad’ is referred to by and metaphorically denotes the picture” (85). Metaphorical denotation is thus also a part of metaphorical exemplification, in that this second moment relates the predicate ‘sad’ to the picture.
Now we do, of course, talk in our everyday language about the emotional character of artworks. Emotional responses to artworks are undeniable and a central part of art’s cognitive value. Yet, the complexity of these phenomena is totally lost by Goodman’s theory, which reposes upon (implicit, unexamined) “conventions” that would allow us to correlate, say, a certain way of painting boulders with agitation, or a minor chord on the subdominant with an exquisite if saccharine melancholy. It just does not suffice, philosophically, to say that a certain symbol possesses a certain predicate. How? Not only is Goodman’s theory unsatisfying philosophically, as it leaves this question hanging, but it also does a woeful disservice to its object: artworks themselves. Goodman both ignores the constitutive role of individual subjectivity in response to artworks and, by skimming over the centrality of “conventions” in the way he does, represses the dimension of intersubjectivity, which is in turn constitutive of art itself.
This idea of gauging and grasping in relation to emotion is extended to art’s cognitive value as a whole, which Goodman sees as fundamental to determining the worth of artworks. All symbolism “is to be judged fundamentally by how well it serves the cognitive purpose: by the delicacy of its discriminations and the aptness of its allusions; […] by how it participates in the making, manipulation, retention, and transformation of knowledge” (258). These criteria are also valid for natural science. As Goodman says (262-3), simple truth is not necessarily of value: multiplication tables can produce an endless stream of truths without meaningfully transforming our knowledge. Rather, truth should always be placed in relation to knowledge that we already possess, in order to transform and develop it. Goodman names this criterion “appropriateness”. Nonetheless, in doing so, he shies from assigning truth to artworks in the same way that he assigns it to scientific theories: “Truth and its aesthetic counterpart amount to appropriateness under different names.” (264)
This rings, to my ears, like a fairly convincing account of the self-reflective nature of modernist (and abstract) art – the kind of account more thoroughly elaborated by Cavell, on Caro, in ‘A Matter of Meaning It’. Yet, Goodman does not really draw out the implications of self-reflexivity in the way Cavell does (viz., the condition of modernity implies that all art is meta-art, all philosophy meta-philosophy). Instead, within the framework of his theory, this account of abstract art threatens to strip it of its expressive power. If a movement simply exemplifies itself, then how is this anything but literal exemplification? Do we not potentially reduce abstract forms to the level of a tailor’s swatch?
