ANALYSIS

OCTOBER

34.1

1973

HUME AND JULIUS CAESAR

By G. E. M. ANSCOMBE
I
IV of Part III of Book I of the Treatiseis a doubly unusual
SECTIONof
piece
philosophical writing for Hume. Read very casually, all
seems uncommonly smooth and acceptable. A little attention, and it
collapses. Revision is incontrovertibly needed to secure coherence. The

neededrevisionthenrevealsthe positionas incredible.

The topic is our belief in matters falling outside our own experience
and memory:
When we infer effects from causes, we must establish the existence of
these causes.., .either by an immediate perception of our memory or
senses, or by an inference from other causes; which causes we must
ascertain in the same manner either by a present impression, or by an
inference from their causes and so on, until we arrive at some object
which we see or remember. 'Tis impossible for us to carry on our inferencesin infinitum,
and the only thing that can stop them, is an impression
of the memory or senses, beyond which there is no room for doubt or
enquiry. (Selby-Bigge's edition, pp. 82-3.)
Now this is a credible account of a kind of prognosis from what is
seen or remembered. That once noted, what must be our astonishment
on observing that in illustration Hume invites us
To chuse any point of history, and consider for what reason we either
believe or rejectit. Thus we believe that CAESAR
was kill'd in the senatehouse' on the ides of March;and that because this fact is established on
the unanimous testimony of historians . . . Here are certain characters
and letters ... the signs of certain ideas; and these ideas were either in
the minds of such as were immediately present at that action; or they
were deriv'd from.., .testimony... and that again from another
testimony... 'till we arrive at ... eye witnesses and spectators of the
event. 'Tis obvious all this chain of argument or connexion of causes
and effects is at first founded on those charactersor letters, which are
seen or remember'd.
This is not to infer effects from causes, but rather causes from effects.
We must, then, amend: 'When we infer effects from causes or causes
from effects', etc. For historical belief:
1
Sticklingfor accuracy,I believethis is false,if by 'senate-house'Humemeantto indicate
a building.The Senatewas not meetingin the senate-house.
I
This content downloaded from 130.49.198.5 on Fri, 13 Mar 2015 22:54:15 UTC
All use subject to JSTOR Terms and Conditions

ANALYSIS

When we infer causes from effects, we must establish the existence of

those effects, either by perception or by inference from other effects;
which effects we must ascertain in the same manner by a present impression or by an inference from their effects and so on, until we arrive
at an object which we see or remember.
For Hume, the relation of cause and effect is the one bridge by which
to reach belief in matters beyond our present impressions or memories.
(That is why the Section 'On the idea, or belief' is in the middle of the
Part which we would think of as the Part on cause.) But also, cause and
effect are inferentially symmetrical.
The historical example is an inference of the original cause, the
killing of Caesar, from its remote effect, the present perception of certain
characters or letters. The inference goes through a chain of effects of
causes which are effects of causes, etc. What is its starting point? It is
natural to say the starting-point is the present perception.
But that cannot be a sufficient exegesis! For what on this account has
become of the argument that we cannot go on in infinitum?The end of
the chain is now the death of Caesar or the perception of its eyewitnesses,
not our perception. But it has to be our perception. What is in question
isn't a chain nailed at both ends, but a cantilever.
The impossibility of running up with our inferences in infinitumwas
not occasioned by our incapacity or exhaustion. The chain of inference
has to stop or else 'there wou'd be no belief nor evidence. And this
actually is the case with all hypotheticalarguments; there being in them
neither any present impressions, nor belief of a real existence'. (Ibid.)
"Tis impossible for us to carry on our inference in infinitum'means:
thejustificationof thegroundsof our inferencescannotgo on in infinitum.Where
we have chains of belief on grounds believed on grounds ... we must
come to belief which we do not base on grounds. The argument here is
that there must be a starting point of the inference to the original cause,
not that inference must terminate. Indeed, one reason why this passage
of Hume's seems fairly ordinary and acceptable at first sight is, that he
strikes one as just making this point, together with the one that the
starting point must be perception.
Does our original amendment 'When we infer effects from causes,
or causes from effects . . .' still stand? Yes, it must. But Hume is arguing
not merely that we must have a starting point, but that we must reacha
starting point in the justification of these inferences. He would have been
clearer if he had said, not 'we cannot carry our inferences on in infinitum'
but 'we cannot trace them back in infinitum'.But as we have said, cause
and effect are taken by him to be inferentially symmetrical. So for him
the tracing back is inference too. But note that it must be purely hypotheticalinference.
Let us see what this looks like in the case in hand.

This content downloaded from 130.49.198.5 on Fri, 13 Mar 2015 22:54:15 UTC
All use subject to JSTOR Terms and Conditions

HUME AND JULIUS CAESAR

Let
p = Caesarwas killed
q = There were [at least ostensible] eyewitnessesof Caesar's
killing
r = There was testimony from the eyewitnesses
s = There were recordsmade, deriving from the testimony
t = There are charactersand letters to be seen which say that
Caesarwas killed.
We must suppose that we start (how?-but let that not delay us)
with the mereidea of Caesar'sdeath. Perhapswe reallydo infer an effect
from it as cause: 'There will have been chaos and panic in the Senate
when Caesarwas killed'. But 'we must establish the existence of this
cause'. As we have seen, this will not be as (at the beginning) Hume
suggests, by deriving it as an effect from a cause; we shall rather,have
to derive it as a cause from an effect. So we reason-and here our
reasoningmust be 'purelysuppositiousand hypothetical'-: ifp, then q;
if q, then r; if r, then s; if s, then t. Not all thesehypotheticalpropositions
are equally convincing, but only this is a chain of inferences through
causes and effects such as Hume envisages. It terminatesin something
that we perceive. That is the last consequent. We can assert this consequent. Now we go in the other direction: since t, s; since s, r; and so
on back to p.
So Hume's thesis falls into four parts: First, a chain of reasonsfor a
belief must terminate in something that is believed without being
founded on anythingelse. Second, the ultimatebelief must be of a quite
differentcharacterfrom the derivedbeliefs: it must be perceptualbelief,
belief in something perceived, or presently remembered.Third, the
immediate justificationfor a belief p, if the belief is not a perception,
will be anotherbelief q, which follows from, just as much as it implies,p.
Fourth, we believe by inferencethrough the links in a chain of record.
There is an implicit corollary:When we believe in historicalinformation belonging to the remote past, we believe that therehas been a chain
of record.
Hume must believe all this: otherwise he could not, however confusedly, cite the chainof recordbackto the eyewitnessesas an illustration
of the chain of inferencesvia cause and effect,with which we cannot run
up in infinitum.

But it is not like that. If the written records that we now see are
grounds of our belief, they are first and foremost grounds for belief in
Caesar'skilling, belief that the assassinationis a solid bit of history.
Then our belief in that original event is a ground for belief in much of
the intermediate transmission.
For let us ask: why do we believe that there were eyewitnesses of

This content downloaded from 130.49.198.5 on Fri, 13 Mar 2015 22:54:15 UTC
All use subject to JSTOR Terms and Conditions

ANALYSIS

thatkilling? Certainlyfor no otherreasonthanthatwe believeit happened.

We infer q fromp, not p from q. I have heardthat the Rabbis held that
the 6oo,ooo witnesses to the crossing of the Red Sea must be credited.'
6oo,ooo witnesses! That's a lot. But now: why does anyone believe
there were 6oo,ooo witnesses?-Because he believes that 6oo,ooo passed
through. And let us make no mistake: it is not otherwise for belief in
there having been eyewitnessesto Caesar'sassassination.
Compareone's belief in the spatio-temporalcontinuity of the existence of a man whom one recognizes and identifies as a man seen last
week. We don't believe in the identity becausewe believe in the spatiotemporal continuity of a human pattern from now here to then there.
It is the other way about.-On the other hand a proof of a breakin the
continuity-a proof that this man was in New York in between, while
that man was not-would destroy our belief in the identity. Mutatis
mutandisthe same holds for the chain of transmission of historical
information.

It is so alsowith propernames.In usingpropernamesthatwe take

to be the namesof peoplewe don'tknow,or peoplein the remotepast,
we implicitlydependon an 'apostolicalsuccession'of users of these
of them-going backto originalusers,
names-or linguistictransforms
who knewthe people.We do not, andusuallycouldnot, tracethe chain
of use of the name.Buta discoverythata namebelongedoriginallyto a
periodlaterthanthe life-timeof the supposedbearerof the nameat any
ratereducesthe statusof the name:it becomesequivalentto some set
of definitedescriptions.
Beliefin recordedhistoryis on the wholea beliefthattherehasbeena
chainof traditionof reportsand recordsgoing backto contemporary
knowledge;it is not a beliefin the historicalfactsby an inferencethat
passesthroughthe linksof sucha chain.At most,thatcanvery seldom
be the case.
II
All thisis not justcatchingHumeout in a mistake.Thatwouldnot
be veryinterestingor important.The mistake-whichI thinkit is now

to referto as such-has
not a bit of patronizingsuperioritymorehodierno
while
demonstrated
of
the rare character being easily
yet it touches the
nerve of a problem of some depth. It is a lot more difficultto see what
to say, than it is to point clearlyto errorin Hume.
One of the rare pieces of stupidity in the writings of Wittgenstein
concerns this matter:
Thatit is thinkablethat we may yet find Caesar'sbody hangsdirectly
togetherwith the senseof a propositionaboutCaesar.But so too does
1I owe this informationto Dr. StephenKatz.

This content downloaded from 130.49.198.5 on Fri, 13 Mar 2015 22:54:15 UTC
All use subject to JSTOR Terms and Conditions

HUME AND JULIUS CAESAR

the posssibilityof finding somethingwritten, from which it emerges

that no such man ever lived, and that his existencewas made up for
IV 56.)
Bemerkungen,
particularends. (Philosophische

Whatdocumentor inscriptioncouldbe evidencethatJuliusCaesar

neverexisted?Whatwouldwe thinkfor exampleof aninscriptionsaying
'I, Augustus Caesar, invented the story of the divine Julius so that

Caesarsshouldbe worshipped;buthe neverexisted'?To aska question

Wittgensteinaskedmuchlater:Whatwouldget judgedby whathere?1
Take somethinga bit less extreme:a documentrecountinga conversationaboutsiege-enginesbetweenCaesarandArchimedes.We will
supposethatthe documentitselfgets acknowledgedby expertsin such
mattersas a genuineold MS. Disputeexists,perhaps,whetherit was
madein the tenthor eleventhcenturyandit comesundermuchcritical
scrutiny.(It is no Piltdownskull.)The Hellenicityor Latinityis authenticallyancient;it seemsreasonableto placethe writingof it in the first
centuryB.C. The styleis suchas mightfit, if possible,with its beinga
is an exampleof
piece of historicalwriting.(Xenophon'sCyropaedeia2
suchwriting.)The contentprovesit to be fictitious.
It mightwell be that the discoveryof such a piecewould compel
someadjustment
in our pictureof whatwas"on"in theliterature
of the
time.It couldnot forcean adjustmentin our idea of the relativedates
of ArchimedesandCaesar.
Of courseWittgensteindoesn'ttell us fromwhatcharacterof documentthatcould"emerge".I do not believehe could.
If you go to an experton JuliusCaesar,you will findhe is an expert
on whetherCaesarconductedsuchandsuchnegotiationswith Pompey
or when he wrote his books, for example.Not on whetherCaesar
existed.Contrastan experton King Arthur.
I wastaught,I think,thatwhenLucretiuswasfirstpublishedduring
the Renaissance,
the De RerumNaturawas suspectedof beinga forgery;
but its Latinityand the absenceof "giveaways"won its acceptance.
Thatmeansthat therewere standardsby whichto judge.The ancient
Latinityof Horace,Ovid,Virgil,CiceroandCaesarwassucha standard,
itselfknownby traditionandneversubjectto question.The attemptto
constructa seriousdoubt whetherwe have writingsof Cicero-how
couldit finda groundfromwhichto proceed?
We knowaboutCaesarfromthe testimonyof ancienthistorians,we
evenhavehis ownwritings!Andhow do youknowthattheseareancient
historians,theseworksof Caesar?You weretold it. And how did your
teachersknow?Theyweretold it.
We knowit frombeingtaught;not justfromexplicitteaching,but
1 In On Certainty.When he wrote On CertaintyWittgenstein would not have made such a
I am a good deal indebted to On Certaintyin this article.
suggestion.
2
Thought to be history in a time of very sketchy impressions of ancient Persian history.

This content downloaded from 130.49.198.5 on Fri, 13 Mar 2015 22:54:15 UTC
All use subject to JSTOR Terms and Conditions

ANALYSIS

by its being implicit in a lot else that we are taught explicitly. But it is
very difficultto characterizethe peculiarsolidity involved, or its limits.
It wasn't an accident that Hume took the killing of Caesaras his example; he was taking something which existed in his culture, and exists
in ours, with a particularlogical status of one kindof certainty.And yet
he got a detail wrong! And yet again, that detail's being right would
not be an important aspect of what he knew. I mean, if he had been
careful, he could have called that in question; he could even perhaps
have calledthe datein question (mightit not have been a false accretion?)
-but that that man, Caesar, existed and that his life terminated in
assassination: this he could call in question only by indulging in
Cartesiandoubt.
I cannot check that there was such a person as Julius Caesar.No
one can, except by finding out the status of the informationabout him.
I mean: suppose there were a schoolchild who first ran into Caesar
through Shakespeare'splay. Somehow he doesn't learn at once that
this is not a purely fictitious story. He refers to Caesaras a fictitious
character, and then someone tells him Caesar was a real historical
character.He can check this; he can look into history books and find
out that that's what Caesaris. But I cannot. I alreadyknow-I can at
best remind myself of-the status of 'Julius Caesar'as a name of an
immensely famous man "in history". (To be sure, these things can
change.)
Or again: suppose a Chinese man, of a time when there was little
contact, who hears of Caesarfrom a traveller. He is accustomed to
chronicles and traditionalinformation.(But we should not forget that
it is by traditional, oral, information that one knows that these are
chronicles, or are editions of ancient books.) He learns that Caesaris
supposed to be such a characterin our history. He cancheck on it. He
can learn our languages,come to our countries,find out that the corpus
of solid historical information belonging to our culture does indeed
include this. But I cannot. The most I can do is: frame the hypothesis
that Caesarnever existed, or was not assassinated,and see that it is
incapableof statuseven as a wild hypothesis. So I do not mean that it is
vastly improbable.I mean that either I should start to say: 'How could
one explain all these references and implications, then? ..

but, but,

butif I doubt the existenceof Caesar,if I say I may reasonablycall it in

question, then with equal reason I must doubt the status of the things
I've just pointed to'-or I should realizestraightaway that the "doubt"
put me in a vacuum in which I could not produce reasons why such
and such "historicalfacts" are more or less doubtful.
I once asked an expert on Galen how he knew that his subject
existed. His reactionwas to consider the hypothesisthat Galen did not
exist. 'It wouldn't do, you know', he said; 'we know too much about

This content downloaded from 130.49.198.5 on Fri, 13 Mar 2015 22:54:15 UTC
All use subject to JSTOR Terms and Conditions

HUME AND JULIUS CAESAR

him-' and went on to mention Galen's connection with Marcus

Aurelius as an example. The response was surely a correct one. What
does the hypothesisamountto in face of our informationabout the time?
But if all that is irrelevant-as we could have no reason for doubting
the existence of Caesar,say, but continuing to believe in Cicero and
Pompey-then the effectof the hypothesisis to makea vacuumin which
there is nothing by which to judge anything else.
The hypothesis about Galen is merely one that "won't do"! That
is: one can relate him to better known historicalmatters.But in face of
such an hypothesis about Caesarone would have to ask: 'What am I
allowed to count as evidence, then?'
People "in history", as we say, are not in any case hypotheseswhich
we have arrivedat to explain certain phenomena.No more than is the
fact of my birth or the existence of my great-grandmothers....
Though I have never given the question any thought before this, I
know I had more than one. Do I know I had four? I would have said so.
But not in the sense that the hypothesis that one of my grandfathers
was a half brother, say, of the other is such that the supposition of its
truth involves destroying bases and standards for discovering any
historicalfacts at all.-And so also about people "in history" there are
gradations;and thereis the possibilityof discovering that some obscure
supposed historicalfigure is probablymythical,or is a conflationor the
like. Things get corrected or amended because of inconsistencies. But
not everythingcan be put up for checking. Von Neurath'simage of the
ship which we repair-and, I suppose, build on to-while it is afloat:if
this suggests that we can go round tapping every plank for rottenness,
and so we might end up with a wholly differentship, the analogyis not
good. For there are things that are on a level. A generalepistemological
reasonfor doubting one will be a reasonfor doubting all, and then none
of them would have anything to test it by.

of Cambridge
University

This content downloaded from 130.49.198.5 on Fri, 13 Mar 2015 22:54:15 UTC
All use subject to JSTOR Terms and Conditions