Gallie - 1939 - An Interpretation of Causal Laws

Mind Association
An Interpretation of Causal Laws

Author(s): W. B. Gallie
Source: Mind, Vol. 48, No. 192 (Oct., 1939), pp. 409-426
Published by: Oxford University Press on behalf of the Mind Association
Stable URL: http://www.jstor.org/stable/2250703
Accessed: 04-02-2016 09:09 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
Mind Association and Oxford University Press are collaborating with JSTOR to digitize, preserve and extend access to Mind.
http://www.jstor.org
This content downloaded from 128.59.222.12 on Thu, 04 Feb 2016 09:09:36 UTC
All use subject to JSTOR Terms and Conditions
VOL. XLVIII. No. I92.] [October,I939.
M IND
A QUARTERLY REVIEW
OF
PSYCHOLOGY AND PHILOSOPHY
I.-AN INTERPRETATION OF CAUSAL LAWS.

BY W. B. GALLIE.
I.
IN his posthumouslypublished paper, " General Propositions

and Causality", F. P. Ramsey dealt with causal laws as a sub-
class of what he calls 'variable hypotheticals'. The pecul-
iarityof Ramsey's view of causal laws resultsmainlyfromhis
view of variable hypotheticals. This is sufficiently suggested
by the followingquotations: (1) "Variable hypotheticalsare
not judgmentsbut rulesforjudging' If I meeta 0 I shallregard
it as a zf'. This cannotbe negated,but it can be disagreedwith
by one who does not adopt it." (2) The variable hypothetical
"is there just to be deduced from". (3) " Variable hypo-
theticals have formal analogies to other propositionswhich
make us take them sometimesas facts about universals,some-
times as infiniteconjunctions. The analogies are misleading,
difficultthoughtheyare to escape, and emotionallysatisfactory
as they prove to different types of mind." (4) " As opposed
to a purelydescriptivetheoryof science,mine may be called
a forecasting theory."1
Ramsey's statementof his positionis cursoryand incomplete.
For example,he gives us no indicationas to how his theory
should be specifiedto cover our beliefand practicesin different
1 The Foundations of Mathematics, pp. 237-255.
27
410 W. B. GALLIE:
sciencesor at different stages in the developmentof scientific

thought. More particularly,a numberof his statementsmust
be revisedif we are to take accountof statisticallaws. Never-
theless there are good reasons for taking Ramsey's paper as
a standardexpressionof one very importantview of empirical
laws. Its terminology, thoughsometimesloose, is fairlyclear,
and is not chosenin deferenceto, or in supportof,some special,
and highlyproblematic,theoryof language-or of science or
of philosophy. Of course Ramsey had his own view of the
functionof philosophy-thatit should make clear the different
ways in which we use propositionsof different formaltypes.
(That thiswas his view I judge morefromhis practicethan fromi
his dicta on the subject.) Now such an attitudeto philosophy
may be too limited,but at least it does notlimitthe philosophical
discussionof causal laws in one particularlyannoyingway,
viz. by employingsuch a terminologythat views opposed to
one's ownsimplycannotbe stated. In thesecondplace,Ramsey's
paper seemsto me to restatewhat is of lastingvalue in Hume's
treatmentof causation,and to supplementHume's view where
it is most deficient,i.e. in connexionwith the question: How
can we account for our beliefthat there are 'unknown causal
laws' ? For thesereasonsI selectRamsey'spaperas a standard
expositionof one view of causal laws; a view whichI radically
oppose. To urge a contraryview in the formof a criticism
of Ramsey's seems to me usefulat a time when philosophers,
in discussingcausation, speak almostlyliterallydifferent lan-
guages.
II.
Ramsey beginshis paper by comparingand contrastingvari-

able hypotheticals(e.g. 'Arsenic is poisonous', 'All men are
mortal') with such universal propositionsas 'Everyone in
Cambridgevoted', i.e. with enumerativeuniversalpropositions.
These latterare, of course,conjunctionsof primarypropositions
which have certainterms in common. Now the notion of a
variable hypotheticalresultsfromacceptinga certainview of
all non-enumerativeuniversalpropositions, namely,thattheseare
to be interpretedas hypotheticals. And this view, commonly
associatedwiththe ' Russellianlogic', appliesto non-enumerative
universalpropositionsin everyfieldof discourse. But Ramsey
(in the paperto be considered)deals onlywitha certainsub-class
of variablehypotheticals,viz. those which,if true,are not true
a priori. In other words,his concernis with empiricallaws,
AN INTERPRETATION OF CAUSAL LAWS. 411
and in particularwith one species of these, causal laws. His

concern with enumerativeuniversal propositionsis similarly
restricted:viz.to thosewhichreferto matteroffact. I propose,
therefore,to introducecertain clearly needed alterationsinto
Ramsey's terminology.Where he writes of 'variable hypo-
theticals' I shall write of 'L-propositions'-e.g. 'Arsenic is
poisonous'. Where he writes of conjunctions(meaning one
sub-class of conjunctions),I shall write of 'U-conjunctions'-
e.g. 'Everyone in Cambridgevoted'.
Accordingto Ramseythe resemblancebetweenL-propositions
and U-conjunctionsis two-fold. In the firstplace, whoever
assertsan L-propositionis therebycommittedto accept as true
any possible conjunctionof propositionswhich are 'instances
of' this L-proposition. Secondly, when we ask what would
make an L-proposition(x)+x true " we inevitablyanswerthat
it is true if and only if everyx has b". And the same holds
of any U-conjunction. Now this formal analogy certainly
suggeststhat L-propositionsand U-conjunctionsare species of
one genus-the genus of universal propositionsfrom which
deductionscan be made as to the truthof simplerpropositions.
But of course this suggestiondoes not involve the assertion
that the analogy between L-propositionsand U-conjunctions
is complete: and we may admit, with Ramsey, that it is in
certaincases very misleading. But Ramsey goes much further
than this, since he is at pains to show that so-called L-pro-
positionsare in fact not propositionsat all, and hence cannot
belongto the same genus as U-conjunctions.
This bringsus to the differences.Ramsey mentionsthree'
of these in the openingpages of his paper; whilsta prolonged
discussionof a fourth,by far the most fundamentaldifference,
is introducedlater. The firstthree differences hold clearly
between L-propositionsand finiteU-conjunctions;the fourth
holds between L-propositionsand any U-conjunctions,finite
or infinite-ifindeed it be significantto speak of the latter.
(1) An L-propositioncannot in fact be writtenout as a U-
conjunction. Now thismightseem onlyto indicatea deficiency
in the symbolswe happen to use. But this suggestionis quite
inadequate. For (2) evenifone wereto thinkofa U-conjunction,
whose constituent propositions together referredto every
possible state of affairsto which we could apply a certainL-
proposition,still this U-conjunctionwould not exhaust the
meaning,or range of application,of the L-proposition. The
referentsof our U-conjunctionwould constituteonly a sub-
set of the class to whichthe L-propositioncould apply. From
412 W. B. GALLIE:
this we can see (3) that the L-propositioncannotexpressa belief

of the primarysort,a beliefas to matterof fact; and, further,
that it cannot be taken to express conjointlya class of such
beliefsin the same way as can a U-conjunction. From here
Ramsey passes straighton to the statementof his main thesis:
that L-propositionsdo not referto facts at all, but are simply
rules for making inferencesas to particularfacts; and more
positively,that their status or modus operandi is that of
habits. (ComparePeirce'sphrase 'habits of inference '.) Now
habits can be altered and dropped; but cannot themselvesbe
describedas true or false.
It seemsclearto me that in thisargumentRamseyis assuming
that it is senselessto speak of infiniteconjunctions. I have no
special competencefor discussingthis difficult point, and will
only say this: that it does seem to me that the notionof con-
joiningor conjointlyassertinga set of propositionspresupposes
that that set is eitherexhaustivelyenumeratedor is thought
of as admittingexhaustiveenumeration. This at least is the
sense of 'conjoin' which I am familiarwith and understand.
There may, of course,be some othersense of this word which
I don't understand; but I doubt if this can be relevantto the
presentissue. For it is surelya hard thoughtthat no one can
understandany inductivelyestablishedlaw who has not found
a clear meaningfor the term 'infiniteconjunction'. Closely
connectedwith this is a second point; not only does Ramsey
set aside the notion of infiniteconjunctions,he writes(in the
paper we are examining)as if the notionof finiteconjunctions
containedno difficulties.Finite conjunctionsare such as either
are or could be exhaustivelyenumerated,and therefore,he
seems to argue, no finiteconjunctioncan be equivalentto an
L-proposition. But if the matterwere as simple as this, why
should people ever have confusedL-propositionswith conjunc-
tions? Let us considerfiniteU-conjunctions moreclosely.
Clearly we sometimesassert a finiteU-conjunctionon in-
completeevidence. Completeevidenceforit could be obtained,
but we do not trouble to collect it. Take, for example, the
assertionthat all the books in a given librarybear the library
stamp. In such a case we can use incompleteevidenceto infer
the truthof a finiteconjunctionin just the way in whichwe
use evidence to establish an L-proposition. Further,having
assertedsuch a finiteconjunction,we need not proceedto verify
it completelyor to use it in its full range of application; we
could confineour use of the assertioninstancedabove to the
case of those books whichhappenedto interestus. Here again
notice the resemblance,in respectof use, betweensuch U-con-

junctionsand L-propositions. Next, let us ask: Since certain
U-conjunctionsare assertedand used in the ways just noticed,
by what criteriondo we decide,in difficult or border-linecases,
whethera given universalpropositionshall count as a U-con-
junctionor not? Is the criterionthis: that I (the assertorof
the conjunction) could completethe enumerationwhich is
required to show that it is a conjunction? Take a wilder
example than our previousone: " None of the pebbles on the
sea-shores of America contain gardenia-seeds." How can I
knowwhetheror not I could completethe enumerationrequired
in the case of this assertion? The answer is, of course,that
I don't know whetherI could do this, but have good grounds
forbelievingthat I couldn't. On the otherhand, I may have
good groundsfor believingthat a finiteset of observerscould
completethe required enumeration. But what would be my
groundsfor this belief? In part, surely,a numberof L-pro-
positions-i.e. inductivelyestablishedlaws as to men's powers
of observation,endurance,co-operation,etc. Thus it turns
out that we need L-propositionsto decide what are finitecon-
junctions: we need the results of problematicinduction in
order to state the evidence for what seemed to be (in prin-
ciple) non-problematicassertions,viz. assertionsof certainfinite
U-conjunctions.
From the above discussionwe can derive three important
conclusions. First, that the notion of a finiteU-conjunction
is, in border-linecases, very far fromclear-cut. Second, the
suggestionthat a finiteU-conjunctionmeansa set ofpropositions
which I (the assertoror entertainerof the conjunction)could
exhaustivelyenumerate,is certainlyfalse-unless we give a
fancyphilosophicinterpretation to the phrase'I could enumer-
ate'. Thirdly, any very large U-conjunction resemblesan
L-proposition(a) in respectof the way in whichit is evidenced,
(b) in respect of the way in which it may be used. In view
of these facts I think it is worthtryingto make out a case
for the view that L-propositionsare U-conjunctions,though
of a very peculiar kind. I do not believe that this view can
be maintained,for reasons, urged by Ramsey, which I will
consider later. But the task is worth undertaking,because
I thinkthat this view, when properlyelaborated,is not at all
absurd; and that it fails forquite different reasonsfromthose
which would usually be urged against it.
Let us concentrateon those U-conjunctionswhich are cer-
tainly such that I could not enumerateall their constituent
414 W. B. GALLIE:
propositions:for if L-propositionsshould turn out to be U-

conjunctionsin disguisethey would clearlyhave to be of this
special type. Now what account are we to give of this kind
ofU-conjunction ? Can we statethe conditions-overand above
purely formal conditions of consistency,etc.-of the signi-
ficanceof large U-conjunctions ? It seems to me that we can
do so only on the followinglines,or somethingverylike them.
We are all familiarwiththe notionof an 'apperceptiveset'
as a name or summarydescriptionof certainconditions,chiefly
mnemic,which are necessaryfor at least our more complex
perceptualexperiences. Now clearlythere are certainbroadly
analogous conditionswhich must hold if symbolicutterances
of mine (in particularmy statementsas to what I have per-
ceived or could perceive)are to be understood by otherpersons,
whethercontemporariesof mine or people living when I am
dead. I shall not attempthere to distinguishthe main kinds
of conditionwhich are necessaryto this result: it is sufficient
that thereare such conditions-thoughwe seldomgive heed to
them until they fail. Now so long as such conditionshold in
the case of my 'reports' these may be said to be includedin
what I shall call a 'socially apperceptiveset', alongwithreports
of other experientsearlier and later. The analogy between
an ' apperceptiveset' in the usual sense and a 'socially ap-
perceptive set' is as follows. Suppose I am examining a
fairlylarge physical object: and suppose that it has twelve
sides. If, when I have examinedit, I am to judge that this is
a twelve-sidedobject, then this judgmentmust, in some way,
contain referencesto each of the twelve sides in its numerical
identity,or as numericallydifferent from each of the other
sides. Suppose now we have twelvesuccessiveobserversmaking
reports: thenif all thesereportsare to be includedin a socially
apperceptiveset,it is necessarythatthetwelfthobserver'sreport
contain,in some way, references to each of the elevenpreceding
reports,consideredas reports-i.e. in respectof what each of the
elevenreported. (The same kind of relationwhichthe twelfth
reporthas to its predecessors, will of coursehold betweenany
one of the elevenprecedingreportsand its predecessors.) Now,
if a socially apperceptiveset is very large, then naturallyin
later reportsthe contentsof earlierreportswill not be actually
enumerated;theywill be referred to by short-handdescriptions.
It is thereforenecessary(in view of the use I wish to make of
the notionof sociallyapperceptivesets) that everysuch short-
hand descriptionshall be understoodto be of eitheractually
enumeratedcontentsof previousreportsor contentssuch that
that they could (easily or manageably)have been enumerated

in previousreports.
The notionof a sociallyapperceptiveset may now be applied
to our question: What are the special conditionsof the sig-
nificanceof large but finiteU-conjunctions? I should say:
Such a U-conjunctionis significant only so long as he who be-
lieves or entertainsit, believes, or entertainsthe possibility,
that every constituentpropositionwhich it contains could be
asserted by somemember(or members)of a group of thinkers,
whosereportedexperiencesconstitutea sociallyapperceptiveset.
To elucidate this account, we may considertwo alternative
possibilities. (1) I think it will be admitted by all that the
conditionswhich must obtain if a series of experiencesis to
constitutea socially apperceptiveset, are in themselvescon-
tingentconditions. Now supposetheysome timecease to hold.
Then, ipso facto,the U-conjunctionwhose significancedepends
on these conditionsturns out to have been finite. But (2) so
long as the requiredconditionsdo hold,it will also be the case
that the U-conjunctionin question is finite. For what the
above prescribedconditionsof a sociallyapperceptiveset ensure,
is that everyreportin sucha set refers(interalia) to the contents,
the actually enumeratedor manageably numerable contents,
of all earlierreportsin the set. And, as we have noted, the
presenceof short-handdescriptionsin later reportsconstitutes
no argumentagainst the genuinelyenumerativecharacterof
everyreportwhichthe set contains.
I am now goingto take it that the notionof sociallyapper-
ceptive sets does provide the special conditionsrequired for
the significanceof those U-conjunctionswhich are such that
I could not enumeratetheir constituentpropositions. Such
conjunctionsI shall,fromnow on, referto as A.S. conjunctions.
Now it seemsthat we could quite plausiblytake the view that
L-propositionsare simplyA.S. conjunctionsin disguise. And
we could hold that the use of an L-propositionsimplyindicates
that we are using a certainU-conjunction(whetherit be an
A.S. conjunctionor no) as evidenceof the truthof a widerU-
conjunctionwhichis an A.S. conjunction. On this view causal
laws and other L-propositionswould be perfectlygood pro-
positions: for an A.S. conjunctioncan be falsifiedby finding
a singlenegativeinstance,whilstits probabilityis raised as the
numberofits positiveinstances(i.e. the numberofits constituent
propositionswhichare foundto be true) is increased. As such,
the view has certainobvious attractionsand is ideally adapted
to a 'purelydescriptive theoryofscience',suchas Mill's. Never-
theless,I think that it is untenable.
416 W. B. GALLIE:
In the firstplace, it neglectsthe factthat such of our thinking

as is -based on generalisations(and it seems to me that all our
thinkingis so based) proceedsin termsof 'would-bes' and
' could-bes'; not in stringsof 'will-bes' and 'weres '. For
example,we forecastthe futurein termsof what wouldhappen,
if somethingelse,whichcould happen,did happen. But against
this it might be argued: These formsof speech could(!) be
replaced by straightforward categoricalforms. They are, in
part, grammaticalconveniences,and in part ways of expressing
our tentativenessas to certainconclusions. But this argument
will not do, forthe followingreason. As Ramsey shows quite
clearly,people oftendisagree-and this disagreementmay lead
to importantdifferences in practice-in cases where everyone
is agreed as to the facts,i.e. as to what has been observedand
recorded. And the reason for this situationis that the same
factsmay be takento supportradicallydifferent generaltheories
or explanations. (See his examplesof the cake-and strawberry-
abstainers.) Closelyconnectedwith this objectionto our sug-
gested theoryis another,viz. that there are many important
scientificlaws-e.g. the firstlaw of motion-which if taken as
'disguised' expressionsof A.S. conjunctions,must be admitted
to rest on no instantial premisseswhatever. These two ob-
jections are, I think,conclusiveagainst our suggestedtheory.
(That many other objections could be raised against it I am
well aware; but since a numberof these apply equally to my
own positiveview, I will postponeconsiderationof them.)
In thissituationare we forcedto fallback on Ramsey'stheory,
and admit that variable hypotheticalsare not propositionsat
all? Since Ramsey's argumentproceedsby the methodof ex-
clusion,we shouldfirstassure ourselvesthat all plausible alter-
nativeshave been excluded. On the otherhand, Ramseywould
have maintainedthat his 'rules' for judging "If I meet a ?
I shall regardit as a t " have all the propertieswhichwe require
of inductivelyestablishedlaws. But is his argumentacceptable
on either side ? To take the latter point first: it certainly
comes as a shock to hear that one cannot properlysay that
some theory,e.g. the phlogistontheory,is false. In the second
place, it is hard to believe that we can gather evidencefor or
against somethingwhich cannot be said to be true or false.
Again, to take the special case of causal laws, it is very hard
to believethat none ofthesetellus anything(howeverindirectly)
about the world around us. Two points are bound up in this
objection: (a) Our use of a causal law is certainlynot confined
to the predicting' a new instance' by means of it; newly es-
tablishedcausal laws carrywith them a new conceptionof the

world(e.g.the contemporary educatedman's 'idea of the world'
as opposed to that of an educated Greekor a savage). (b) Ou-r
idea of theworld,in this non-philosophicsense,is mostnaturally
givenin termsofa systemofsubstancesand processesofdifferent
sorts. And thingsare sortedor classified,in the main,by their
behaviourproperties,i.e. by referenceto certain causal laws.
In this way causal laws do tell us somethingabout the world
around us.1
Possiblynone of these objectionsare conclusiveagainst Ram-
sey's theory,but I think they provide some reason for look-
ing foran alternativeview. There may be a numberof such
alternatives,but I shall presentonly one.
In my view L-propositionsshould be definedas a species of
explanatorypropositionsor explanations,i.e. universal pro-
positions from which, whetheralone or in conjunctionwith
certain instantial premisses,deductions can be made. It is
clear that they differfrom certain other explanatorypropo-
sitions,e.g. those of formallogic, (a) in the ways in which
they are evidenced,(b) in the ways in which they are used.
(L-propositionsare neitherself-evident nor deduciblefromself-
evident premisses; and from L-propositionsalone we cannot
make the kind of deductionswhich we all habituallydo make
from L-propositionstaken together with certain instantial
premisses.) Again, it is clear that L-propositionsdiffer,as
explanations, from U-conjunctions which simply 'resume'
(to use Karl Pearson's phrase) sets of primarypropositions.
L-propositionsclearly do more than 'resume' the cases they
apply to-otherwise how could we have alternative,competing
explanationsof the same facts? But thus far our definition
of L-propositionsis negative. On the positive side we must
notice the peculiar relationwhich holds betweenthe evidence
forthem and our use of them; i.e. betweenthose primarypro-
positionswhichwe count as evidencefor an L-propositionand
those primarypropositionswhich are its instances,or which
' Consider,forexample,the statement,'Magnets attractmetals'. In
certaincontexts,and for certainreaders,this expressesan informative
proposition,an empirical(causal) law. In othercontexts,and forother
readers,it may be taken to expresspart of the definition of the word
' magnet'. On thislatterinterpretation the statementis notinformative,
does not tell us about the world. But if we accept thislatter interpre-
tation,thenwe mustadmitthat the existentialpropositionexpressedby
"Magnets exist" or " There are magnets"-which certainlytells us
somethingabout the world-containsa causal reference;this reference
beingboundup in the meaningof theword' magnet'.
418 W. B. GALLIE:
constituteits range. This, I hold, is as follows: If p, q and r

(conjoined) are taken as evidence of an L-proposition,then
p, q and r must,in the natureofthe case, be consideredmembers
of a largerconjunction,an A.S. conjunction,whose truth (i.e.
the truth of each of its constituentprimarypropositions)is
deducible fromour L-propositiontaken togetherwith certain
appropriateinstantial premisses. To give an example. Two
or threeexperiments withfallingbodies wouldprovide evidence
forthe truthof the formulad/t2is K. But the assertionofthis
formulaas an L-propositionrequiresthat thesetwo or threein-
stancesbe thoughtofas membersof a verylarge class ofsimilar
experiments, actual or possible,i.e. the class whicha certainA.S.
conjunction would describe. (This A.S. conjunctionwill be
of the form: Experiment1 was of such an observablenature;
Experiment2 was of a similarobservablenature; and so on.)
Now to assert the truthof our L-propositionis not the same
as to assert the truth of the corresponding A.S. conjunction
which verifiesit: because (1) it is logicallypossible to assert
the A.S. conjunctionwithoutregardto any L-propositionwhich
explains it, and (2) there may, conceivably,be a numberof
L-propositionseach of whichwill explain the A.S. conjunction.
Nevertheless,the assertion of any L-propositiondoes imply
the truthof a corresponding A.S. conjunction.
The purpose and meritsof this theorycan be broughtout
by contrastingit (a) with our previouslysuggestedtheory,
viz. that Law-propositionsare A.S. conjunctionsin disguise,
and (b) with Ramsey's theory.
(a) My theoryis, I think,proofagainstthose criticisms which
seemed conclusive against the previously suggested theory.
For on the presentaccount it is easy to see that it is our ac-
ceptance of certain L-propositionswhich in fact determines
what primarypropositionswe are, at any given time,ready to
assert or curiousto verify. In other words, what facts we
thinkit worthwhileto examineand recorddependsupon what
L-propositionswe believe. The restriction thusimposedon what
we shall observeis indeednot logicallynecessary; thatwe make
observationsas to particularfact only in the light of certain
very generallaws is a contingentfact about the workingsof
the humanmind. It is, nevertheless, a factwhichany theoryof
L-propositions musttake into account; and this our previously
suggestedtheoryfailedto do.
(b) On my theory,when we say that certain things would
happen if other things,which could happen, did happen, we
are emphasisingthe fact that certain propositionsare con-
sequencesof-or are at leastcompatiblewith-some L-proposition

which we believe. But these consequences are not thought
of only as consequencesof the L-propositions. They are also
thoughtof as constituentsof that A.S. conjunctionto which
the L-propositionis ' pegged', and save in relationto which
has no propersignificance
the L-proposition and henceno truth-
value. Again, on my theory,the L-propositionis not 'there
simplyto be deduced from'. It is there to explain a certain
describedset of primarypropositions-theconstituents
definitely
of the corresponding A.S. conjunction. Because of this, it is
a genuineproposition,true or false of the world.
III.
I will now state and replyto certaincriticismsthat mightbe

broughtagainst my theory.
(1) It mightbe said: This theoryis eithercircularor plati-
tudinous. For it presupposesthat we all understandhow an
L-propositionexplains certain primarypropositions,which is
the verymatterwhichwas to be explained. Now thisobjection
is trivial,if taken in one way. For it mightbe taken to be
a requestthat we shouldexplainthe generalnotionof explana-
tion. And thisrequestI decline. For surelythisnotioncannot
be explainedat all. The best we can do is to distinguishthe
main speciesof explanation. Now in this paper I have argued,
against Ramsey, that L-propositionsare a genuine species of
explanation. And I have assumed that, as such, they are to
be distinguished fromexplanationsin mathematicsand formal
logic, and fromall explanationswhich rest on the acceptance
of certaindefinitions.And I have tried to show where they
difierfromexplanationswhich rest on completed,i.e. perfect
inductions. To this defenceI will only add that those phil-
osopherswho, like Hume, try to explain away the notion of
explanation,are surelyvictimsof a kind of intellectualnear-
sightedness. Hume, for instance, offersus his psychological
(or meta-psychological) account of why we trustthe inductive
method,and why we believe in a world of persistentthings.
He forgetsto notice that his psychologicalpremissesare used
as explanations-of the workingsof the human mind. The
generalcategoryof explanationis instancedwhereverthinking
occurs.
But, takenin a different way,the above objectionis important.
It could be elaboratedas follows: In manyL-propositions there
420 W. B. GALLIE:
are terms,let us call them explanatoryfactors,to which no

specificexistentscorrespondin the observationswhich verify
these L-propositions. The use of certainconstantsin physics
is an example; or the use of the notionsof ' normalfunctioning
of the organismas a whole' in physiology. Now it mightbe
asked: How do explanatoryfactorsof this kind help to explain
the factswhichwe observe? And it mightbe urgedthat any
adequate theoryof L-propositionsmust answer this question.
I must admit the forceof this argument,and can only answer
that my theoryis deficientin the way indicated. And this
deficiencycould only be made good by a carefuland complete
surveyof the different kinds of explanationthat occur at dif-
ferentlevels of science and common-sense thought. As a first
step towardsthis,I have sketchedout, in the last part of this
paper, my view of the groundsof the distinctionwe all draw
between causal and non-causalLaw-propositions.
(2) On my theoryL-propositionsare true or false 'of the
world'. And I countthisa strongpointin favourofmy theory.
Their truth or falsityis indeed derivative: they are true or
false as explanationsof A.S. conjunctions. But via these they
apply to the world. Yet I have admittedthe possibilitythat
an A.S. conjunctionmight be explained by more than one
L-proposition. Hence I seem to be committedto holdingthe
uncongenialviewthattwo or moreL-propositions may be equally
true of a given set of facts. But this difficultyis not peculiar
to my theory; it arises on any theory that maintains that
L-propositionsare propositions. And, fortunately, abler pens
thanminehave alreadysuggestedthelinesalongwhichitssolution
is to be found. In any case wherewe are presentedwithwhat
seem to be alternativeexplanationsof (apparently)the same
facts,we must firstask: Is the evidence for the different ex-
planationsthe same ? If it is not,theneitherone ofthe theories
can in principlebe shownto be more probablethan any of its
rivals, or else it will turn out that the different theoriesare
explainingdifferent, perhaps overlapping,fieldsof fact. And
if the evidence for the different theoriesshould be the same,
then we must analyse out those parts of each whichare equi-
valent. When this has been done the residualdifference in our
theoriesmust,if it is significant,be a case of the difficulty
dis-
cussed above, viz. that of explanatoryfactors(in this case dif-
ferentexplanatoryfactors)to which no specificitems in our
observationscorrespond.
(3) We come to a far more radical criticismof my theory,
and one that would apply to Ramsey's and many othertheories
as much as to mine. I will call it the realist criticism. It

informsus that L-propositionsare made true or false by the
facts,and in the case of most L-propositions,by facts about
the 'nature of matter'. As such, their significanceis by no
means limitedto what I or some hypotheticalset of observers
may suppose or assertas to fact. They are true or false,once
forall, of the kindsof fact,e.g. factsabout atoms,to whichthey
refer.
Now a fullanswerto this criticismwouldrequirea vindication
of the thesis (in part phenomenalist,in part pragmatic)that
L-propositionsare significantonly in relationto the primary
propositions we can derivefromthem. This,of course,I cannot
attempt. I will only recommendto the realist critic these
two quite simple considerations,which may help him out of
his difficulty.(1) Even if not actually meaningless,L-pro-
positionsare plainly uselesssave in so far as they assist pre-
diction-or retro-diction-as to particular matters of fact.
And save in so faras theyare thus used, it is impossibleto give
any determinatetruth-value,or any degree of probability,to
our Law-propositions.(The realist begins by wanting 'real
laws'. He ends witha set of valuelessformulae on his hands.)
(2) (And here,for brevity,I make use of what can be a very
misleadingterminology)-Truth, consideredas a relation,neces-
sarily requirestwo terms, each of which is ' internal' to it:
viz. factand significantstatementor reference. Now the second
of these clearly depends on certain contingentconditions,in-
cludinghuman memoryand purpose. Should these conditions
fail, then significancefails (as with the Etruscan inscriptions)
and the questionof truthdoes not arise. The questionof the
truth or falsityof any propositionsis significantonly on the
assumptionthat certain conditionshold; and the question of
the truthof L-propositions is significant
only on the assumption
of certain surprisingly complicatedand narrowconditions; as
I have tried to show.
IV.
Let us now turnto the question: What distinguishescausal

laws fromothernon-causalL-propositions? Ramsey'saccountof
the matteris as follows(p. 249). Suppose we assert," If p had
happened,q wouldhavehappened". Then,Ramseysays " p q "
must here " followfroma hypothetical(x)+x 3 Ox and factsr,
pr: q being an instance of Ox n Ox and q describingevents
422 W. B. GALLIE:
not earlierthan any of those describedin pr." Ramsey then

elaboratesa theoryof his own as to the meaningsof past and
future. His theoryseemsto assumethat we have an immediate
knowledgeof the diflerencebetweenbeliefsand volitions: and,
on the basis of this ass-umption,he definesthe past by reference
to the class of our beliefsto the determination of whose truth
or falsityour volitionsare irrelevant. I can see hardlyanything
to be said forthis view, but I shall not discussit here. What
I wish to discuss is the peculiarway in which (as it seems to
me) we take pr : q to be an instance of Oxn Ox, when we
assert pr : q to be an instance of a causal law.
Wheneverwe use a causal law to make a prediction, we predict
trulyor falselythe dates of two or more events,relativelyone
to another. But veryoftenin determining the dates of events
we use non-causal Law-propositions.Suppose, for example,
two Englishmenare disagreeingas to the exact wordsused by
a Frenchman. One mightsay to the other," My view must
be right,for on your view, the orderof his wordswouldn'tbe
grammatical,"and this would be a fair argument;we can use
the laws of grammarto predictthe order,the relative dates,
of certainevents-i.e. actually spokenwords. And yet clearly
the laws of grammararen't causal. They are non-causal in
the same wide sense in whichthe laws of perspective,the laws
of empiricalgeometry,and indeed all the laws of the " back-
ward" or still " purelyclassificatory " sciencesare non-causal.
But what is this sense? We all know it vaguely,for we can
all see that in any instanceof a non-causallaw the condition
doesn't referto a state of affairssuch that we can properlycall
it thecause,orevenpartofthecause,ofthestateofaflairsreferred
to by the consequent. But we must improveon this vague
knowledge. And usually,as a firststep,we are told that causal
laws have a special referenceto change,to temporalpassage:
by causal laws we explainthe orderof events. But, as we have
just seen, non-causallaws can be used to explain the order
of events. Admittedlyin this kind of explanation (see the
above example) certain importantpremissesare suppressed,
and these would of course be causal premisses. But can we
see in what peculiar way these suppressedpremissesreferto
the temporalorderofthefactstheyexplain? Now thisquestion
seems to me to be just ignoredin most 'non-activist' theories
of causation; whilst'activist' theoriesherepresentus withun-
helpfulremarksabout production,prolongationof events,etc.
In this situationI proposeto deal withthe matterin what may
at firstseem rathera round-aboutway.
The instantial judgmentswhich verifyL-propositionsmay

be classifiedby referenceto the threetenses,past, presentand
future. But mostof our uses ofthe presenttenseare inevitably
vague or loose: they containa certainreference to the immed-
iately rememberedpast. They are similarlyloose with regard
to the immediatefuture. This is particularlyobvious when
we are speakingof things-as indeed we always do in common-
sense discourse. The thing we are concernedwith means to
us somethingwe are ready or preparedfor. And this implies
that we thinkofthe thingas a continuant. (Or, in otherwords,
commonsense, like physical science,is committedto a belief
in the 'conservationof something '.) Now I am not here con-
cernedwith the psychologicalanalysis of this characteristicof
many of our beliefs,nor with its justification-ifindeed any
can be given. That almost all our judgmentsas to what is
presenthave this characteristic,which I shall call 'the pre-
sumptionof persistence',is a truism,though for my present
argument an importantone. And it should be noted that
almost all our judgmentsas to what is past or futurecontain
an analogous presumption. (E.g. had a presentjudgmentas
to a past event been made at the time of that past event,it
would have manifestedthe presumptionof persistence:and
similarlywith judgmentsas to the future.) Now it seems to
me that our use of certain predicates emphasisesthis pre-
sunmption, whereas the use of certain other predicatesallows
or even encouragesus to ignoreit. The formermightbe called
elementarycausal predicates,the latter elementarynon-causal
predicates. (As examples of the firstwe have-strained, ap-
proaching,alive. As examples of the second-coloured, cylin-
drical,distant.) We now come to the difference betweencausal
and non-causallaws. The former, I shouldsay,are alwaysused,
and can only be used, to specifythe precisemannerin which
a state of affairs,whichit is presumedwill persistin some way
or other,will develop.
For the purposesof exposition,let us assume that the per-
sistence of any thing can be stated in terms of a succesion
of states. Then, in making a judgmentas to a presentedor
inferredstate of affairs,we usually presumethat this will be
succeededby a run of states,differing in some degreefromthe
firststate, and one fromanother. Causal laws, and even to
some extent elementarycausal predicates,help us to specify
this presumptionboth in respectof the qualities of any of the
succeedingstates and in respect of the order in which these
succeedingstates will be manifested. This, I hold, is the sole
424 W. B. GALLIE:
functionof causal laws.' What then is the functionof non-

causal Law-propositions ?
This cannot be answeredso simply. As a firstsuggestion
we mightsay that non-causallaws can be applied only when
we make that kind of abstractionfromtemporalpassage which
Mead describesas 'thinking of the universe at an instant'.
For example,conceiveof an instantaneously co-ordinatedstate-
mentof all the appearancesof a physicalobject at a givenmo-
ment: thenyou will have an instanceof the laws ofperspective.
But no such co-ordinatedstatementwould give you an instance
of any mechanicallaw: not even a law in statics. Now this
account does bringout an importantdiflerencebetweencausal
and non-causallaws of certainkinds,e.g. the laws of perspective
and of empiricalgeometry. But as it standsit cannotbe applied
to all non-causallaws. Consider,forexample,the purelyclassi-
ficatorylaws characteristicof zoology or psychology,at least
in their earlier phases. The types which these laws classify
couldn't exist at an instant: they requirea certaintime-span
'to be whattheyare'. And yet theseelementaryclassifications
don't give us causal laws. For instance,if we found a skull
and judged it to be a man's, we shouldn'texplainor justifythis
judgmentby recallingthat 'being a man' causes the existence
laws, based on the obser-
of this kind of skull. Classificatory
vation of ' familylikenesses' are thus neithercausal laws (in
the sense properto this term) nor yet are they non-causalin
the same way as are the laws of perspectiveand empiricalgeo-
metry. What then is their status?
This question can only be answeredby noticinghow they
are used. Such an examinationshows us that purely classi-
ficatorylaws are at once make-shiftsand hybrids. Sometimes
1 Certain empiricallaws, which would commonlybe called causal,
may appear to be concernedexclusivelywith simultaneousstates of
substancesor processes; e.g. Boyle's law. But any gas is definedby
referenceto its causal properties,so that any massofgas requiresa period
of timeto be what it is. Boyle's law thenis not concernedwithsimul-
taneousstatesconsidered in completeabstractionfromotherstateswhich
precededthem or will succeed them. On the contrary,if we consider
howit is used,we see thatBoyle'slaw accordsperfectly withourdefinition
of causal laws. pv = RT can and does help us to specifyhow a given
mass of gas will develop. Of course,it does not do this directlyforit
describesa certainproperty(a higher-order property)which the gas
willmanifest in any ofits futurestates-providedthesefallwithincertain
well-known limits. But at least it rules out a vast numberof lines of
development whichit is logicallypossiblethat the gas shouldtake. And
in this way-and here Boyle's law is on all fourswith any theoremin
mechanics-thetask of 'specification'is greatlyadvanced.
we use them(lazilyand even dangerously)in the mannerproper

to causal laws: sometimeswe use them as if they classified
certainfragments ofnatureconceivedas existing' at an instant'.
That this latter use greatly simplifiesour understandingof
nature,by givingus ideal specimensor normsround whichto
classifyactual changingexistents,is clear. On the otherhand,
this use can be dangerous: for,havingorderedand in one sense
explained certain existents by means of classificatorylaws,
we are temptedto thinkof these laws as givingus causal ex-
planationsof theirinstances. We should,of course,admit that
these are veryincomplete-orvague-causal explanations. But
even this is utterlywrong: for classificatoryexplanations-
based on observed resemblances-aren'tin any sense causal.
(As examples, witness the 'vicious' doctrine of substantial
forms,and, equally, many or most of our psychologicaljudg-
ments. E.g. " It was his typical selfishnessthat caused the
trouble".) Recognitionof thisdangerousambivalenceof classi-
ficatory laws notonlybringshometo us the amountofabstraction
involvedin our thoughtsabout the 'furnitureof the universe';
it also re-emphasisesthe peculiarfunctionof causal explanations.
The searchforcausal laws cannot be equated withthe attempt
to explain events by bringingthem under wider and wider
generalisations.In the case of causal laws increasedgenerality
is only prized when by means of it we can describethe per-
sistence and interactionof existents in greater and greater
detail. Neglect of this fact has, I think,vitiated almost all
positivist accounts of scientificexplanation. And on more
particularissuesneglectof it has caused muchconfusion:e.g. in
connexionwiththe question,Can we findand apply causal laws
in the study of human history?
I turn,lastly,to a differentspeciesof non-causallaws, namely
those whichdependon some convention,or habituallyobserved
and enforcedagreement. As an example,I shall considerthe
laws of grammar; but I thinkthat a greatmany of what pass
as moralgeneralisations fallunderthishead. (E.g. in " Manners
makythman ", "makyth" is not, one hopes, used as a causal
predicate.) Here again our question is, by what criteriondo
we distinguishthese non-causal laws, restingon conventional
uses and attitudes,from causal laws ? The answer, at least
for the case of grammaticallaws, is fairlysimple. These are
generalisationsabout certain bits or products of human be-
haviour: writtenor spoken words and sentences. Now we
mightbe temptedto argue that instancesof.grammaticallaws
are " causal facts", on the groundthat the habits and con-
28
426 W. B. GALLIE: AN TNTERPRETATION OF CAUSAL LAWS.
ventionson whichthese laws are said to dependare themselves

largelysustainedby the regularitymanifestedin the instances
of these laws. But we feel this is wrong. Why? Largely,
I think,because sentencesand other signs, in so far as they
are used and understood, don't perceptiblypersist-or at least
don't persistlongenough. Wlhatdo persistare the conditions,
physiologicaland psychological,which determinethe habit
of producinggrammaticalsentences. These, we feel, could be
tracedout in detail,and thissatisfiesourdesireforan explanation
of languagewhichwill be expressedin laws of a totallydifferent
kind fromthe laws of grammar. The requiredlaws would be
causal, and would constituteone of the centraldoctrinesof a
developed psychology. But at the moment,with a very few
notable exceptions,social psychologists have not passed beyond
the 'purely classificatorystage'.
The view of causal laws here sketchedcarrieswith it certain
important implications. The 'presumption of persistence'
(which, I hold, explains the peculiar status and functionof
causal laws) is surely,in part, an ontologicatpresumption. As
suchit is,thepartialexpressionofa certainattitudeto the world:
that attitude which emphasisesthe characterof the world as
process. Philosopherscannot be segregatedinto sheep and
goats by referenceto their acceptance or rejection-or sheer
ignoringof the importance-of that attitude. Not just ac-
ceptanceof it, I would say, but the eflortto showits relevance
in particularfields,is the importantthing. And it is equally
importantto showwhereneglector rejectionof it by competent
thinkershas renderedvague and incompletetheir treatment
of those topics whichthey claim to have made clear.

Gallie - 1939 - An Interpretation of Causal Laws

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gallie - 1939 - An Interpretation of Causal Laws

Uploaded by

Copyright:

Available Formats

Mind Association

An Interpretation of Causal Laws

PSYCHOLOGY AND PHILOSOPHY

I.-AN INTERPRETATION OF CAUSAL LAWS.

IN his posthumouslypublished paper, " General Propositions

sciencesor at different stages in the developmentof scientific

Ramsey beginshis paper by comparingand contrastingvari-

and in particularwith one species of these, causal laws. His

this we can see (3) that the L-propositioncannotexpressa belief

notice the resemblance,in respectof use, betweensuch U-con-

propositions:for if L-propositionsshould turn out to be U-

that they could (easily or manageably)have been enumerated

In the firstplace, it neglectsthe factthat such of our thinking

tablishedcausal laws carrywith them a new conceptionof the

constituteits range. This, I hold, is as follows: If p, q and r

sequencesof-or are at leastcompatiblewith-some L-proposition

I will now state and replyto certaincriticismsthat mightbe

are terms,let us call them explanatoryfactors,to which no

as much as to mine. I will call it the realist criticism. It

Let us now turnto the question: What distinguishescausal

not earlierthan any of those describedin pr." Ramsey then

The instantial judgmentswhich verifyL-propositionsmay

functionof causal laws.' What then is the functionof non-

we use them(lazilyand even dangerously)in the mannerproper

ventionson whichthese laws are said to dependare themselves

You might also like