John McDowell Knowledge and The Internal PDF
John McDowell Knowledge and The Internal PDF
John McDowell Knowledge and The Internal PDF
1.1 first delivered this essay in the 1989190 lecture series of the Center for the Philoso-
phy of Science at the University of Pittsburgh. I presented an ancesto r to a conference on
Belief and Knowledge in Albi, France, in July 1981. Cllarleli Trav is commented helpfully
on the ancestor, and Simon BJackbum on a revised version of it. A less distant descendan t
of some of that material for ms part of Essay 17 above. More recently, I have benefited
from conversations with jonath an Dancy; from romments on a more recent drah by
Roben Brandom; and from the responses of audiences at Ohio State University, Ha verford
CoIkgc, and the Chapel Hill Co lloq uium, especially Robert Kraut, L Arych Kosman, and
Jay Rosenberg. Ro bert Brandom ' s respon se at the Chapel Hill Co lloquium has been pub-
Iished as " Knowledge and the Social Articulati on of the Space of Reasons".
2. " In characterizing an episode or a state as that [bet ter: one) of know ing, we are not
giving an empirica l descripti on of tha t episode or state; we arc placing it in the logical
spac e of reasons, of justifying and being ab le to justify what one says" and
the Philosop hy of Mind", p. 298-9). I put in the parenthcrical qualification so as to allow
that a co ncept of know ledge might be applied to non -rational animal s too; but nothing in
this essay will depend on that.
396 IS S UES I N E P IS T E MO LOG Y
3. I am deliberately leaving the idea of blameworthiness., in one 's moves in the space of
reasons., unspecific. If sornt<lI'e arrives a t a bIsc belief h om which she would have been
dereered by some investigation she choK not to enpge in because of its high CC5t an d low
probability o f overrurning lhe olher evidence, is she bla meworthy? Different answers are
poaible. BUI no reawn.able interprnation of th e idea of doxasr:ic obIigarions could mak e
falsehood in an em pirical belief show, all by itself, that an obligation has nOl been met.
That is the cmtral insight (a genuine insighl , even thou gh it is typically mishandled ) of the
familiar genre of philosophy acco rding 10 wh ich empi rica l knowledge is pr oblematic. I
want to foclI$ on this ga p, nea rly un iversally ackno wledged, between doxasric blameless-
neSll in a sense that connects w ith doxaslic obligarion, o n the ant: hand, and em pirical
knowledge, on the other, without being distracted by det ails a bout how dcxasoc blame-
lessness should be understood. (Th e epistemological ou tloo k I shall recom mend should
make such questions seem less ur sern.)
4. In the case of testi mony, one is literall y done a favour by an infonnaJll. But the re will
have 10 be,l1gurativdy, a f.avoor from the world at so me point in lhe epiSlemic ancestry of
a piece of knowledge by resrimon y. (At leaSl o utside th e a rea of, for instance, being IOIdof
something tha t ha s been pr oved; sec the text below.]
398 ISSUES IN EPISTEMOLOGY
interiorized and still recognizably the space of reasons a run for its
money.
So we are to try to reconstruct the epistemic satisfactoriness im-
plicit in the idea of seeing that things are thus and so, using the fol-
lowing materials: first, the fact that it looks to a subject as if things
are that way; second, whatever further circumstances are relevant
(this depends on the third item); third, the fact that the policy or
habit of accepting appearances in such circumstances is endorsed by
reason, in its critical function, as reliable. And now the trouble is
this: unless reason can come up with policies or habits that will
never lead us astray, there is not enough here to add up to what we
were trying to reconstruct. Seeing that things are thus and so is a po-
sition that one cannot be in if things are not thus and so. Given that
one is in that position, it follows that things are thus and so. And if
reason cannot find policies or habits that are utterly risk-free, the re-
constructing materials cannot duplicate that. However careful one is
in basing belief on appearances, if one's method falls short of total
freedom from risk of error, the appearance plus the appropriate cir-
cumstances for activating the method cannot ensure that things are
as one takes them to be.
There are various possible responses to this point. The one I re-
commend is that we should jettison the whole approach to knowl-
edge that structures epistemology around the Argument from Illu-
sion. I shall mention three others.
Obviously one response is scepticism. In my Sellarsian framework,
I can put the sceptical response like this. An epistemically satisfac-
tory position would have to be a standing in the space of reasons-
Sellars is right about that. But the argument I have just sketched
shows that we cannot reconstruct a standing in the space of reasons,
suitable to amount to knowledge, with respect to the fact that things
are as they perceptually appear. So it must be a mistake to think we
can achieve knowledge through perception. This thought clearly gen-
eralizes, in a way that matches the generalization of the Argument
from Illusion.
A second response would be to claim that there must be policies or
habits of basing belief on appearance that are utterly risk-free. It is
obvious how this response might be attractive, in the context of the
threat of scepticism; but I do not think it has any plausibility in its
own right. It would express a rather touching a priori faith in the
power of human reason to devise fully effective protections against
400 I S S U E S I N E P ISTE MOLOGY
6 . It is impoetanr nor to assume Iha t, in reiect ing this response, I am making unavail-
able the oornmon- sense th c e ghr rhal we so meti mes kn ow how things ate by seeing how
Ihey a re. Thai wou kl be so on ly if the epistemic S1at us of such knowledge had to consist in
the excdlmo;:e of ;;l policy()( habit of basing belief 00 appearance. focused a s it we re on the
panicular case a t hand. Bur m at assu mpti on is simply a form of what is under att ack. (The
S1atus con sists, rather, in the fact that one sees thai lhingi are so.)
7 . On " blarne!cu", see n. J a bove. H owever prec isely it is s pelled out, the idea of
b1amew on hinc:ss that we need must belong with an XIea o f o bliga tionl as within one's
power to discharge, on pain of losing con tact wilh the point o f intCfiori zing the space of
reason s. So it is nOl to lhe poin t here 10 sugges t thar one can be blamed for ;;I false belief
based on apPearance iUS! beca use of its falsehood, on the an alogy of the idea that on e can
be blamed fOt" unintended comcquences o f one 's inttmional acts .
Knowledge and the Internal 401
can reason have the resources it would need in order to evaluate the
reliability of belief-forming policies or habits? If we press this ques-
tion. the idea that something can be both interiorized in the way I
am considering and recognizably a conception of the space of rea-
sons starts to unravel, as I have already hinted that it would .
I shall return to that point; meanwhile I want to urge another
problem about the hybrid conception of knowledge. In the hybrid
conception. a satisfactory standing in the space of reasons is only
part of what knowledge is; truth is an extra requirement . So two
subjects can be alike in respect of the satisfactoriness of their stand-
ing in the space of reasons. although only one of them is a knower,
because only in her case is what she takes to be so actually so. But if
its being so is external to her operations in the space of reasons, how
can it not be outside the reach of her rational powers? And if it is
outside the reach of her rational powers, how can its being so be the
crucial element in an intelligible conception of her knowing that it is
so-what makes the relevant difference between her and the other
subject? Its being so is conceived as external to the only thing that is
supposed to be epistemologically significant about the knower her-
self, her satisfactory standing in the space of reasons. That standing
is not itself a cognitive purchase on its being so; it cannot be that if
the space of reasons is interiorized. But then how can the uncon-
nected obtaining of the fact have any intelligible bearing on an epi-
stemic position that the person's standing in the space of reasons is
supposed to help constitutes How can it coalesce with that standing
to yield a composite story that somehow adds up to the person's
being a knower?
One way to appreciate what I am driving at here is to consider the
familiar point that true belief need not amount to knowledge. Why
not? A good simple answer is that mere truth in a belief leaves it
open that the believer has hold of the truth by accident, and knowl-
edge excludes that. Now in the hybrid conception of knowledge, it is
admittedly not a complete accident, relative to someone's standing in
the space of reasons, if things are as she takes them to be; the posi-
tion of her belief in the space of reasons makes it likely to be true.
But the reason why the extra stipulation that the belief is true-what
is distinctive of the hybrid approach-is needed is that likelihood of
truth is the best that the space of reasons yields, on the interiorized
conception of it: the closest we can come to factiveness. The extra
404 I S S U E S I N E P ISTE MO LOG Y
11. When dou ble-aspect views of oontmt. involving menml stateS were a novelty, it
used to be routi ne to cite the supposedly obvious composi leness of kll(lwkdgt' as an al-
read y familiar parallel See, e.g., Da niel C Dennen, "Beyond Belief', pp. 11-1 2; and
Co lin McG inn, uThc: Stroo:tun: of Coeeenr", p. 215. In querying 1M credentials of a hybrid
conce ption of knowledge, I mean to do more than remove an expository prop from under
rbose double-aspect views; I believe that direct extensions o f the conside rations in rhis
essay show lhat those views miss the poin t of the oonceptu al appa ratuS they aim 10 ex -
plain-I shall noc be able to elaborate this here, though it will be close to the surface in S6.
Knowledge and the Internal 405
12. M 1 noted in S2, we owe the world th a nks for presenting us with appearances at
a U. But that po int is accommodated by the form ula tion in the text . (If the world withheld
appea rances from us, reason wo uld achieve its goa l by deterring us from unsu ppcrted
guesses as to hew things are.)
13. Of course one can ma ke mista kes; but the idea is that proper exer cise of ru son
would d iminate them.
14. On the paraUd s in the sphere of pract ical reaso n, sec Bernard W iUiams, " Mo ral
Luck-.
406 I S S U E S IN EPI STEM OL O GY
pos itions such as seeing that things are a cert ain way. Whe n some-
one enjoys such a pos itio n, th at involves, if you like, a stroke of good
fortu ne, a kindness fro m the world; even so, the position is, in its
own right. a satisfactory sta nding i.n the space of reasons, not a com-
pos ite in which such a sta nding is combined with a co nditio n exter-
nal to the space of reaso ns. Jj Whether we like it o r not, we have to
rely o n favours from the world: not just that it presents us with ap-
pearances-which, as I remarked, the fantasy view can already ac-
cept as a favour the world does us-but that o n occasion it actually
is the way it appears to be . But that the world does someo ne th e nec-
essary favour, o n a given occasion, of being the wa y it a ppears to be
is not extra to the person's sta nding in the space of reasons. Her
coming to have an ep istemically satisfactory standing in the space of
reasons is not what the interiorized conception would require fo r it
to count as her own unaided achievement, But o nce she has achieved
such a standing. she needs no extra help from the world to co unt as
knowing.16 lf we rescue th e idea of the space of reasons from the dis-
tortions of fantasy, we ca n say that the particular facts th at the
world does us the favour of vouchsafing to us, in the various relevant
modes of cognition, ac tually shape th e space of reasons as we find it ,
I S. Thi s fOl'mulation shoul d make it dear how wildly off-target Blackburn is ("Knowl-
edge, Truth, and Reliability", p. 176) in supposing that my appea l to "gua ranteeing" in-
formational states belongs within the t;enttal framewor k of the attempt " to ensure that
there is no clement of luck, or even con tingmcy, in the tRIC believer' s title to knowledge" .
The trad ition al effecr of this artempe to transcend luck is that area of known fact
shrinks "po tmrially down to an entirely subjecti ve realm" . Blackbu rn tam me to offer a
diffetmt option with in the same:genera l fra mework , in whkh, instead of that shrinkage in
what ca n be know n, the mind (the seat of these supposed leek- free "guaranteeing" sta ttS)
expands to "embrace" all sorts of wor ldly states of affain. Th is idea. wtUch Blackburn
righrly finds bizarre, has nothing to do with what I am proposing here, and was proposi ng
in the work Blackburn is discussing (Essay 17 Blackburn is so locked in to the
ftamework tho ught that epistemology m ust Centre on a luck-free ZOIlt (a role played in his
favoured epistemol ogy by the "indicatift" sta tes to whach we arc pu shed back by the gen -
era lized Argument from lIIusion ) that he cannot: com prehend how I ca n have been qun-
honing the frame work; 50 he has sadd led me with the insane position that is the onl y in-
terpreta tion my wo rds will bear with in the framework .
16. Exorcizing the fantilsy sho uld weaken the inclination to say that such a standing is
nor one's own unaided achievemen t. Co mpa re one of the practical analogues. The concept
of w hat one docs, understood as applying to one's; inttrventions in the objective wo rld,
cannot: mark out a sphere within which one has total con troJ, immune to luck. It is only if
we recoil from this into a fant asy of a sphere with in which one's control is tota l that it ca n
seem to follow that what one gmuillt ly achicvu is.1ess tha n one's interventions in the ob-
jective world . (fhis is one of many plICa at wh ich much more discussion is needed.]
Knowledge and the Internal 407
17. S«ing (o r mo re genera lly perceiving) that things ate a certai n way is just one of th e
(or , in Blackburn's term, "guaranteeing") stat es th at is restored to its proper su-