You are on page 1of 22

Journal of Philosophy, Inc.

Causation: A Matter of Life and Death


Author(s): John Earman
Source: The Journal of Philosophy, Vol. 73, No. 1 (Jan. 15, 1976), pp. 5-25
Published by: Journal of Philosophy, Inc.
Stable URL: http://www.jstor.org/stable/2025447
Accessed: 21-11-2015 05:44 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.

Journal of Philosophy, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Philosophy.

http://www.jstor.org

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
THE JOURNAL OF PHILOSOPHY
VOLUME LXXIII, NO. I, JANUARY I5, 1976

CAUSATION: A MATTER OF LIFE AND DEATH *

I. INTRODUCTION: THE LIVING DEAD


D ESPITE several predictionsof imminent demise and even
a few obituary notices, causation as a topic of philosophi-
cal discussion refuses to die. Each year, books and articles
on causation continue to pour forth. Of course, all this activity
may simply be a symptom of the necrophilia that infects so much of
philosophy. This was Bertrand Russell's diagnosis when he wrote
that "the word 'cause' is so inextricably bound up with misleading
associations as to make its complete extrusion from the philo-
sophical vocabulary desirable.... The law of causality, like so much
that passes muster among philosophers, is a relic of a bygone age,
surviving like the monarchy, only because it is erroneously supposed
to do no harm." 1
Part of Russell's polemic against causation simply doesn't apply
today. Writing in 1912, Russell charged that the words 'cause' and
'effect' do not appear in the advanced sciences. But current physics
journals make reference to a number of causal concepts, including
determinism, causal signals, and microcausality. Indeed, causation
as it is used in the sciences is such a variegated concept that one
would be surprised if it would fall captive to the attempts of
philosophers to trap it in the formula 'X is the cause of Y iff. . ..'
This raises a companion and neo-Russellian point. In assuming that
it is sensible to talk about causation in terms of the formula 'X is
the cause of Y', philosophers presuppose, in effect, that there is
something to the concept of causation over and above the particu-
* I want to thank Mr. Martin Bunzl for a number of helpful conversationson
these topics.
1 "On the Notion of Cause, with Applications to the Free Will Problem,"
reprinted in H. Feigl and M. Brodbeck, eds., Readings in the Philosophy of
Science (New York: Appleton-Century-Crofts,1953).

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
6 THE JOURNAL OF PHILOSOPHY

lar causal concepts used in the sciences and that this something has
wide-ranging applications both in human affairs and in the inter-
action of inanimate objects. Could we but speak with him, Russell's
ghost would undoubtedly tell us that this presupposition is false
and pernicious.
Although I am not entirely cowed by this apparition, I believe
the basic thrust of its message to be correct. In this paper I will
not attempt a general defense of my belief. Rather, I will concen-
trate on attacking one of the most crucial elements of almost all of
the philosophical discussion of causation; namely, the tenet that
causation has a directionality that is grounded in objective physical
features of the world. Part of the attack will be directed at the
widely held notion that temporal order is key to causal order.
The attack will be launched by examining "backward causation."
But backward causation serves a much more important function:
contrasting alleged cases of backward causation with the more nor-
mal sorts of cases helps to pin-point the reasons why we want to say
that causation is normally future-directed. These reasons, I will
argue, collapse under detailed scrutiny.
Causation lives. But, if I am right, the causation of which philoso-
phers so fondly speak, lies mouldering. The least we can do at this
late date is to give it an honest burial.
II. THE HERITAGE OF MILL
Mill's analysis of causation has proved to be the most persuasive in
terms of the number of adherents. Modern versions of Mill's analy-
sis are to be found in the writing of Feigl, Carnap, Hempel, Pap,
and many others.2 On the Millian analysis, causation reduces to
determination. Mill's "law of causation" holds if and only if the
universe is deterministic; indeed, on the cosmological level, Mill's
vision coincides with that of Laplace, whlo said that "we ought to
regard the present state of the universe as the effect of its antecedent
state and as the cause of the state that is to follow." Of course, Mill
thought that complex causal connections can often be resolved into
simpler ones-his "methods of experimental inquiry" were de-
2 See Feigl, "Notes on Causation," in FeigI and Brodbeck, op. cit.; R.
Carnap, Philosophical Foundations of Physics (New York: Basic Books, 1966),
ch. 19; C. G. Hempel, Aspects of Scientific Explanation (New York: Free Press,
1965), pp. 347-354; and A. Pap, An Introduction to the Philosophy of Science
(New York: Free Press, 1962), ch. 14.
3 Actually, Mill's law of causation assumes not only determinism in the sense
defined below, but also time-translation invariance and, hence, periodicity for
the laws of nature; see bk. III, ch. Ori1, sec. 1 of A Systenz of Logic.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH 7

signed in part to effect just such a resolution-but each of the


simpler connections is again reducible to a deterministic uniformity.
The rub comes when one recognizes that there are two types of
Laplacian determinism: futuristic and historical. For simplicity, let
us begin by considering theories that postulate a Newtonian space-
time frameworks By a theory T I will mean the closure under logi-
cal implication of a set of lawlike sentences, and I will assume that
a model for a theory T is a map that assigns to every instant t of
Newtonian time an instantaneous state s(t) of the universe. The
set of dynamically possible models MDT of T consists of all those
models which satisfy T. (The use of the phrase 'dynamically pos-
sible' is motivated by thinking of the law sentences of T as con-
sisting of "laws of motion"; the "dynamically possible models" then
correspond to trajectories satisfying the laws of motion.) T is futur-
istically (respectively, historically) deterministic iff, for every s, S2 e
MDT and every t and t', if t < t' (t' < t) and sl(t) = s2(t), then sl(t')

s2(t').
Following Mill, let us say that, if T is a true, futuristically de-
terministic theory, if the states s(t) and s(t') occur, and if t < t',
then s(t) is the future-directed cause of s(t').5 By analogy, we have
a temporally reversed sense of causation: if T is a true, historically
deterministic theory, if the states s(t) and s(t') occur, and if t < t',
then s(t') is the past-directed cause of s(t). To date, all the basic laws
of physics have proved to be both futuristically and historically
deterministic, to the extent that they are deterministic at all. Thus,
to the extent that physics justifies a belief in future-directed Millian
causation, it should also justify a belief in past-directed Millian
causation.
The common-sense response is apt to be that this talk about past-
directed causation is only a play with words. But to sustain this re-
sponse, it is not sufficient to point to the fact that we don't ordi-
narily talk of past-directed causation. The advocate of common
sense must show that this linguistic fact is not simply the result
of historical accident or a quirk of human psychology. No doubt
the advocate of common sense would further respond to this chal-
lenge with the claim that only future-directed Millian causation de-
4 For a discussion of determinism within relativistic space-time structures,
see my paper, "Laplacian Determinism in Classical and Relativistic Physics,"
forthcoming in the University of Pittsburgh series in Philosophy of Science.
5 Since nothing important in what I have to say is sensitive to the details of
the ontology of causation, I will speak variously of states, events, happenings,
etc., as being causes.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
8 THE JOURNAL OF PHILOSOPHY

serves the name 'causation' since causal influences can "go" only
in the earlier-to-later direction and that it is simply a failure of the
Millian analysis that it does not explain what it is that "'goes"and
why it can "go" only one direction in time and not the other. One
of the main purposes of this paper is to investigate the viability
of the common-sense response.
There is yet another aspect of causation that the Mfillian tradition
does not adequately handle. If there is any paradigm case of causa-
tion, it is force as a cause of acceleration. But the usual type of
force law, both Newtonian and relativistic, relates instantaneous
force to instantaneous acceleration. This poses a problem for
Xfillians, since they distinguish between laws of co-existence and
laws of succession and since they relegate causal laws to the latter
category., But the force laws mentioned seem to belong to the
former category rather than the latter.
Cases of instantaneous causation provide a serious challenge to
the view that causal order is grounded in temporal order. But I
will not be much concerned here with instantaneous causation. The
immediate effect of instantaneous force is instantaneous accelera-
tion; but, intuitively speaking, these immediate effects "spread out"
in the future direction of time. The instantaneous force laws can
be analyzed in terms of determinism, and the fact that they usually
imply a development that is both futuristically and historically de-
terministic seems, from the common-sense viewpoint, to imply that
causation must include something more than just determinism.
The common-sense view should be stated somewhat more cau-
tiously. It should contain an escape clause to allow for the possibility
of some exceptional cases in which there is backward causation.
And here springs the hope that backward causation can help with
forward causation: seeing why one is tempted in the exceptional
cases to say that causal influences can "go" in the later-to-earlier
direction may help us to see why in the usual cases we want to say
that causal influences can "go" only in the earlier-to-later direction.
Or, alternatively, it may help us to see why we should abandon the
common-sense view.
Before turning to backward causation, I will present the second-
most-popular approach to causation.
III. THE SINE-QUA-NON ANALYSIS
Mill was also the first to discuss in any detail what has come to be
another major tradition in causation; namely, the sine qua non

6 See, for example, Hempel, op. cit.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH 9

analysis according to which the cause is that but for which the
effect would not have occurred.7 On this tradition no deterministic
uniformity is needed for causation. Although Mill rejected the sine
qua non analysis,8 it has continued to be advocated, the most recent
proponents being David Lewis 9 and J. L. Mackie.10For purposes of
discussion, I will start with a version of the sine qua non thesis
which neither Lewis nor Mackie would want to hold without
qualification:
Sine qua non thesis: If c and e are distinct events that occur, then c is
the causeof e iff if c hadn't occurred,then e wouldn't have occurred.

Such a thesis faces some grave difficulties. For one thing, it will not
provide a necessary condition for causation unless events are indi-
viduated sufficiently narrowly. On the other hand, if events are
individuated too narrowly, it will not provide a sufficient condi-
tion. And this tension may be enough to tear the thesis apart.1'
However, I am not concerned here with the demerits of the sine
qua non thesis; rather, I am concerned with what light it may
shed on the direction of causation. Whatever difficulties the thesis
faces, one feels that there is a strong connection between 'a caused /3'
and 'if a hadn't occurred, then p wouldn't have occurred'. And it
is this feeling which holds out hope of help for our problem. For the
counterfactual 'if a hadn't occurred, then p wouldn't have occurred'
tends to sound odd when a refers to occurrences whose date is later
than the date of the occurrences referred to in 8. But alas, this
oddity may be a symptom of our problem rather than a cure; it
may simply be a reflection of our intuition that causation is nor-
mally future-directed.
According to Lewis's truth conditions, our counterfactual is true
(in the actual world) just in case some possible world in which
neither a nor /3occurs is closer to the actual world than any possible

7 Some passages in Hume's Inquiry also suggest the sine qua non thesis; but
these passages do not fit with either of Hume's two explicit definitions of
causation.
8 Mill may have wanted to retain a sine qua non clause as a necessary con-
dition for causation; see bk. III, ch. V, sec. 3 of A System of Logic.
9 "Causation," this JOURNAL, LXX, 17 (Oct. 11, 1973): 556-567; parenthetical page
references to Lewis are to this article.
10 The Cement of the Universe (Oxford: Clarendon Press, 1974); parenthetical
page references to Mackie are to this book.
11 More details are given in my review (in preparation for Philosophical Re-
view) of Mackie's book. See also B. Berofsky, "The Counterfactual Analysis of
Causation," this JOURNAL, LXX, 17 (Oct. 11, 1973): 568-569; and J. Kim,
"Causes and Counterfactuals," ibid.: 570-572.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
IO THE JOURNAL OF PHILOSOPHY

world in which p but not a occurs. On my admittedly slippery


grasp of the comparative similarity of possible worlds, Lewis's
truth conditions lead to a symmetry between the truth or falsity of
'if the state s(t) hadn't occurred, then the state s(t') wouldn't have
occurred' in the two cases when t < t' and when t' < t, under the
assumption that s(t) refers to the instantaneous state a la T., where T
is a true theory that is both futuristically and historically determin-
istic.
It is at this point that Mackie and Lewis part company. Because
Mackie's sense of similarity of possible worlds is similar to mine,
he wants to modify the above raw sine qua non analysis by adding
a condition to the effect that c is "causally prior" to e. Mackie's
proposal for explicating causal priority in terms of "fixity" of events
will receive some attention below.
Lewis believes that no such further condition needs to be joined
to the sine qua non thesis. He argues that laws of nature should not
be held sacred in making comparisons of similarity of possible
worlds; for to suppose the contrary in conjunction with (futuristic)
determinism "would mean that if the present were ever so slightly
different, then all of the past would have been different-which is
absurd" (567). 'Where Lewis sees absurdity, I see truth. But, ab-
surdity or no absurdity, to take Lewis's proposal at its face value as
a solution to our problem is to choose theft over honest toil. Theft
is sometimes preferable to toil, but not here. It may be true that,
when t < t', it is less of a departure from actuality to get rid of s(t')
by holding s(t) fixed and giving up some law of nature rather than
by holding the laws fixed and going back and abolishing s(t). But,
even supposing this is true, our problem is to explain why the anal-
ogous assertions with t and t' interchanged is not true. And if the
explanation cannot be grounded on objective physical asymmetries,
causation won't bear the weight philosophers want to put on it.
IV. SOME OBJECTIONS TO BACKWARD CAUSATION
At this juncture I suggest we turn to some alleged cases of back-
ward causation for help. I will not attempt a point by point rebuttal
of the many arguments designed to show that backward causation is
not possible; 12 rather I will try to by-pass them by giving examples
where backward causation obtains. Still, it will be useful to review
briefly some of the more prevalent of these arguments.
The boldest and the least interesting denial of the possibility of
backward causation rests on the assertion that it is just part of the

12 For a nice review of a number of these arguments together with numerous

references to the literature, see ch. 7 of Mackie's Cement of the Universe.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH II

meaning of 'cause' and 'effect' that a cause is never later than its
effect. Even if it is true, this assertion does not have much force.
To take an analogous example, it may be true that the special
theory of relativity changes the meaning of 'simultaneity' because
it may have been part of the pre-Einsteinian meaning of simul-
taneity that simultaneity is absolute, the same for all observers. But
it seems clear to many that Einsteinian simultaneity is recognizably
a concept of simultaneity. Whether or not alleged cases of back-
ward causation are recognizably cases of causal connection is not a
matter to be settled by appeal to ordinary use of 'cause' and
'effect'.
Closely related to the first objection is a second objection which
denies the possibility of backward causation on the grounds that
if the ordinary concept of causation were changed so as to allow
for such a possibility, then absurdities would break out in other
concepts that have logical liaisons with the concept of causation.
This objection can be defused by means of the same examples as the
first. For, to the extent that the claim is true with respect to causa-
tion, it is also true that changing the ordinary concept of simul-
taneity so as to allow for the relativity of simultaneity would
cause absurdities to break out in concepts that have logical liaisons
with the concept of simultaneity.
A third objection dismisses backward causation on the grounds
that, if backward causation were possible, it would also be possible
to change the past; but the latter is no possibility at all since the
past is dead and gone, and no one can change it. But it is no less
true, and no less trivial, that what will be will be than that what
has been has been. Whatever will be is the future, and no one can
change that. So if the present objection is valid, it would be paral-
leled with a valid objection to future-directed causation.
And yet, hasn't something been left out? A fourth objection at-
tempts to provide grounds for a positive answer. According
to this objection, past events are "fixed" in a way that future events
are not; there is then an important asymmetry between past and
future, which is glossed over by the trivialities that what has been
has been and what will be will be. Moreover, the objection con-
tinues, unfixed events cannot cause fixed events. Hence, backward
causation is impossible. I agree that we do have a feeling that the
future is "open" and "unfixed" in a way in which the past is not.
But insofar as I can give a coherent account of this feeling, the
"openness" of the future and the "fixity" of the past are not re-
spectively preconditions for future-directed causation and a block
against backward causation; rather, both are a result of our belief

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
12 THE JOURNAL OF PHILOSOPHY

that our present actions have causal influences which are always
(or almost always) future-directed.
A fifth and more modest objection does not dismiss backward
causation tout court; it is claimed only that backward causation
is incompatible with futuristic determinism. The idea is that, if
events at t are already determined (and, hence, "fixed") by events
at some earlier time to, then events at t cannot be causally influ-
enced by events occurring at a time t, later than t. Futuristic de-
terminism leaves, so to speak, no room for backward causation to
operate. The force of this objection seems to me to rest on the
same mistaken assumptions that underlie the two preceding objec-
tions. Moreover, the present objection proves too much if it proves
anything at all, for it would imply that forward-directed causation
is also incompatible with futuristic causation. Parroting the above
line of reasoning, we can argue that, if events at t are already de-
termined (and, hence, "fixed") by events occurring at the earlier
time to, then events at t1, where to < t1 < t, cannot causally in-
fluence events at t. Given that time does not have a beginning, it
follows that futuristic determination leaves no room for future-di-
rected causation to operate. Mackie bravely swallows this conse-
quence. Though I admire the state of his digestion, I also believe
that this consequence is so inherently implausible as to prompt
skepticism about the whole fixity idea.
In attempting to motivate and defend his asymmetrical treat-
ment of past and future in his doctrine of fixity,13 Mackie is driven
to falling back on the claim that "what happens next flows from
what is there already. The immediate future is, so to speak, ex-
truded by the present and the immediate past" (225). The only
way this 'so to speak' is cashed in is by means of the further claim
that the ideas of flow and extrusion help to explain the (alleged)
fact that most long space-time "worms" are temporal rather than
spatial. I see no need for such a metaphysical explanation over and
above what physics tells us. And, in any case, the crucial asymmetry
Mackie needs is otiose for purposes of the explanation; for the
explanation would work just as well if the immediate past were
extruded from the present and the immediate future.14
13 For Mackie, an event e is fixed at time t if e or a sufficient cause of e
occurs at or before t. c is a sufficient cause of e iff c and e occur and if e hadn't
occurred, then c wouldn't have occurred. I will ignore problems caused by the
relativity of simultaneity.
14 For additional remarks on fixity, see the first reference of fn 11; I argue
there that Mackie's fixity notion fails to capture causal directionality for
"sufficient causes."

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH 13

Contrary to Mackie, I do not see determinism as any block to


causation, forward or backward; indeed, it seems to me that a com-
bination of backward and forward causation requires determinism
-futuristic, or historical, or perhaps both-as a necessary condi-
tion for enough "mesh" between past and future that inconsisten-
cies are avoided. And here arises a sixth, and, to my mind, the most
serious challenge to backward causation: determinism does not
always seem sufficient to guarantee the relevant "mesh," and so
backward causation is haunted by the specter of inconsistencies. I
propose to discuss this challenge in terms of a concrete example
given in the following section.
V. A PRIMA FACIE CASE OF BACKWARD CAUSATION
In order to construct an example in which there is a strong prima
facie case for backward causation, it is not necessary to resort
either to mythical-land stories or to extrasensory-perception phe-
nomena. In fact, one need look no further than physics text books.
The Lorentz-Dirac equation of motion for classical relativistic
charged particles is a third-order (in time) differential equation.
When certain asymptotic conditions are imposed, the equation is
converted into a second-order integro-differential equation accord-
ing to which the acceleration of a charged particle at proper time
To is equal to the integral over the "effective force" acting for all
times T > ro. As a result, the integro-differential equation predicts
"pre-acceleration" effects; e.g., if a sharp impulsive force acts on
the particle at time a, then the particle will begin to accelerate
before r.15 For the moment, I will assume that this result consti-
tutes prima facie evidence that if the equation in question were a
true law of nature, then force would have backward causal effects.
The ground for this assumption will be probed in detail in later
sections.
The first question we must face then is not whether the integro-
differential Lorentz-Dirac equation provides a true or even a plausi-
ble account of classical relativistic electrodynamics. Rather, the
question in the first instance is whether the account is consistent and
coherent, and, if so, what light it casts on causation, forward as well
as backward.

15 For some specific examples, see G. N. Plass, "ClassicalElectrodynamicEqua-


tions of Motion with Radiation Reaction," Review of Modern Physics, xxxiii
(1961): 37-78. No uniqueness proof is known for the initial-value problem for
the integro-differential Lorentz-Dirac equation. If uniqueness, and therefore
determinism, should fail, the example can be shifted so as to satisfy the re-
quirementsof determinismwhile retainingthe relevantnonlocal time dependence.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
I4 THE JOURNAL OF PHILOSOPHY

VI. PARADOXES OF BACKWARD CAUSATION


Imagine a system governed by the integro-differential Lorentz-Dirac
equation. If I can control the force acting on a given particle in
the system in such a way that I can turn the force off or on at any
time r 2 ro depending upon whether or not the particle pre-accel-
erates at ro, then a contradiction results. Does the particle pre-
accelerate at ro? If it does, then I switch off the force after ro; so,
according to the equation of motion, it doesn't pre-accelerate after
all. On the other hand, if it doesn't pre-accelerate at ro, I switch
on the force after ro; so, according to the equation of motion, the
particle does pre-accelerate after all. But either it pre-accelerates or
it doesn't. Contradiction.
The paradox is most easily and vividly illustrated in terms of
agency, but it can be restated so as to avoid all mention and use of
human agency, volition, etc. All we need is two mechanical devices,
one of which is programmed to deliver a sharp blow at ri (>To) iff
its switch is in the "on" position, and a second device which "senses"
the presence or absence of pre-acceleration at To and which is
programmed to turn the switch on the first device to the "off" or
''on" position accordingly.
Should such paradoxes be regarded as a proof of the impossibility
of backward causation? No. At best, the paradoxes demonstrate the
impossibility of a certain combination of backward and forward
causation. Leave out our ability to affect the future, and we cannot
reset the switch so as to generate an inconsistency. To emphasize
this point, imagine a world in which all causal influences are in the
future-to-past direction. No paradoxes arise. But it would probably
be a safe bet that in such a world some philosophers would argue
that forward causation is impossible on the grounds that forward
causation would lead to paradoxes.
Secondly, the paradoxes don't even show that a combination of
forward and backward causation is impossible. What is shown is
that the possibility of such a combination requires the satisfaction of
a set of consistency conditions. This is a familiar conclusion to those
who have followed the literature on faster-than-light particles in
special relativity theory and closed timelike curves in general rela-
tivity theory.16If, for example, there could be approximately closed
world lines in space-time, then the following results: an observer
could travel into his own near past and shoot himself-at-an-earlier-
16 For further details see the reference of fn. 3 and my "Implications of
Causal Propagation outside the Null Cone," Australasian' Journal of Philosophy,
L, 3 (December1972):222-237.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH 15

time, thus preventing himself from living long enough to shoot him-
self. (Again, nothing important turns on human agency.) Such para-
doxes do not demonstrate the logical or physical impossibility of
closed causal curves in space-time. They are simply a dramatic way
of showing that closed causal loops impose consistency conditions
on equations of motion, and this conclusion should have been ap-
parent from the beginning. And so it is with backward causation.
Of course, this is not the end of the story. Two further questions
present themselves. Is there any reason to believe that such con-
sistency conditions are realized in possible worlds accessible from
the real world? And is there any way of explaining the satisfaction of
such conditions, short of invoking a benevolent God who does not
want the world to lapse into inconsistency and is therefore careful
to arrange things so as to avoid contradictions?
In considering similar paradoxes which threaten their theory of
classical relativistic electrodynamics, J. A. Wheeler and R. P. Feyn-
man claim that the paradoxes depend on "the postulate that dis-
continuous forces exist in nature." Their resolution is that "we are
led to make just the contrary assumption, that the influence of the
future on the past depends in a continuous way on the future con-
figuration." 17 But, in their original paper, Wheeler and Feynman
provided no justification for their assumption other than the ques-
tion-begging desire to ward off inconsistencies.
However, one step can be taken toward a justification: it can be
shown that the Wheeler-Feynman continuity assumption is con-
sistent with their theory. This follows from the fact that, for a sys-
tem of charged particles moving solely under their mutual electro-
magnetic interactions, the Wheeler-Feynman equation of motion
implies that the force acting on any particle is an infinitely differen-
tiable function of proper time.'8
But, of course, there are other types of force in addition to elec-
tromagnetic ones, so the truth of the Wheeler-Feynman assumption
remains to be settled.
Moreover, the Wheeler-Feynman assumption does not exhaust the
consistency conditions for their theory. The paradoxes that threaten
their theory, as well as the paradox outlined at the beginning of this
section, need not rely on the use of discontinuous forces. They do,
however, assume that forces can be "switched on" in sufficiently
17 "Classical Electrodynamics in Terms of Direct Interparticle Action," Re-

views of Modern Physics, xxi (1949): 427.


18 See J. Cohn, "Considerations on the Lorentz-Dirac Equation," Journal of
Mathematical Physics, xi (1970): 294/5.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
i6 THE JOURNAL OF PHILOSOPHY

small intervals of time; from the macro-human perspective, the in-


tervals may be so small that the forces appear discontinuous. In the
case in point, the integro-differential Lorentz-Dirac equation implies
that the acceleration at proper time r depends effectively only on
forces acting during the interval from r to r + A (A > 0). For an
electron, A is of the order of 10-23 seconds; for typical macroscopic
objects, it is even smaller. Thus, in order to implement the para-
doxical gedanken experiment it is necessary to employ forces that get
"'switchedon" in less than 10-23 second. But, according to one author-
ity, "No classical interactions . . are known that have space or
-

time variations corresponding to such short intervals. No classical


particle can enter from a free state into interaction over a time inter-
val of order [A]." 19 If this is correct, we have an explanation of
how the required consistency conditions can be fulfilled in the
actual world. Unfortunately, the explanation also rules out the pos-
sibility of constructing clear-cut cases of pre-acceleration effects.
Still, the theories under discussion do imply that earlier events de-
pend on later events in a fashion that gives prima facie support to
an attribution of backward causation.
In contrast, there seems to be no non-question-begging reason for
thinking that the consistency conditions required by tachyons and
closed timelike curves will be satisfied in a possible world that is
not very remote from the actual world; but this is another story.
VII. INTERVENTION
What justifies an ascription of forward or backward causal influ-
ence? We know by now that comparisons of similarity and differ-
ence among the elements of MDT will not by themselves supply
the answer. We want to know how differences at a given time in the
elements of MDT make for differences at other times. The rub comes
in trying to pin down the relevant sense of 'make for'. If T is de-
terministic, then, for any s1,s2 e MDT and any t', if sl(t') / s2(t'), s1(t)
#4 s2(t) for any t < t'. But if we want to deny that T implies back-
ward causal effects, as we surely will in most common cases, then
we must deny that the differences in the states at t' make for differ-
ences at the earlier time t in the relevant causal sense of 'make for'.
But unless some yet to be revealed consideration can be brought
forward, symmetry of reasoning blocks us from concluding that dif-
ferences at earlier times make for differences at later times in the
relevant causal sense of 'make for'.
We seem then to be stuck as long as we hold the laws of T in-
19 F. Rohrlich, Classical Charged Particles (Reading, Mass.: Addison-Wesley,
1965), p. 151.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH 17

violate. Should we then contemplate violations of laws? A positive


motivation for such a course suggests itself: a seemingly promising
way to test the directionality of causal influences of forces is to
imagine what would happen if God were to intervene by creating
a miracle that changes the laws only by imposing some additional
force.20If it is then not consistent with such a miracle that the his-
tory of the world be the same before (alternatively, after) the
miracle, then backward (forward) causation reigns.
The machinery already introduced can be extended so as to make
such imaginings tractable. For this purpose, consider a theory T
formulated within the framework of Newtonian space-time, and
suppose that T is both futuristically and historically deterministic
and that the laws of T are invariant under time translation, so that
the time origin can be shifted at will without affecting the truth
of the theory. Let T* be a theory that is like T except that (a) the
state variables of T* include an additional variable Fext, called the
external force function, and (b) the laws of T* differ from those of
T including Fextin the expression for the total force. Typically, T*
will be conditionally deterministic on Fext; i.e., for any s *, s2l' e
MDT* and any time t, if sl*(t) = s2*(t) and sl* and s2* agree at all
times on the value of Fext, then sly = s2*. I will assume that T* is
conditionally deterministic in this sense.
From T* we can generate an infinity of subtheories by particular-
izing Fext.For instance, let T1* denote the theory in which Fext satis-
fies: Fext(t) 0 for t < 0 and fit) for t > 0, where f(t) may be non-
zero for t > 0. It is assumed that, for a wide class of physically
reasonable f's, MDT, #4,i. Each Tf * if, of course, deterministic.
We now seem to be in a position to capture the intuitive idea
that a necessary condition for T not to entail backward causal
effects is that each s e MDT can be held fixed for all t < 0 while the
external force is varied in any manner we like after t = 0. For,
evidently, this condition will be fulfilled just in case
(Cl) For everyf such that MDT, 7#h 0, there exists a map 1t MDT > MDA,.
such that (i) -I is bijective and (ii) for every s e MDT and every t < 0,
s(t) -4,(s) (t), where 'I' signifies agreement on all the state variables
of T.

20 This is not to concede Lewis's point about counterfactuals and the violation
of laws of nature. I still hold that if the world is fully deterministic, it is true
that if the present state of the world were different than it actually is, then the
entire past and future would also be different. It does not follow that causation
cannot be analyzed in terms of counterfactuals. But it does follow that the coun-
terfactuals have to be more complicated than those Lewis considers; e.g., they
may have to make explicit reference to intervention events.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
I8 THE JOURNAL OF PHILOSOPHY

Note that, if bDexists, it is unique, as follows from the fact that T*


is conditionally deterministic on Fext.
If T posits an equation of motion with features like those of the
integro-differential Lorentz-Dirac equation, then in general (Cl)
will be violated; and, as expected, the analysis yields the desired
result that backward causation is entailed by such a theory.
By parallel reasoning, we arrive at a necessary condition for the
absence of forward causation. Let T.* denote the theory in which
Fext(t) = 0 for t > 0, g(t) for t < 0, where g(t) may be nonzero for
t < 0. If, for T., force does not have forward causal effects, it must
be possible to hold fixed each s E MDT for all t > 0 while the ex-
ternal force is varied in any manner we like before t = 0. In terms
of our apparatus, the consequent condition comes down to
(C2)For every g such that MDT: = 0, there exists a map ma:MDT - MDTM
such that (i) *g is bijective and (ii) for every s e MADT and every t > 0,
s(t) , *(S) (t).

As before, if *, exists, it is unique.


If T posits a temporally nonlocal equation of motion, but now
with a temporal asymmetry opposite to the integro-differential
Lorentz-Dirac equation, so that now the acceleration depends on
the past rather than on the future behavior of the system, then in
general, (C2) will be violated; and again we obtain the expected
result that such a theory entails forward-directed causal effects.
If the above analysis is correct, it has three important conse-
quences. First, it shows that there are logically and physically co-
herent cases in which we can give good reasons for assigning a
direction to causation. Secondly, it shows that this direction can
either be future- or past-oriented. Hence, as Mackie maintains, tem-
poral order is not the key to causal order. Thirdly, and contrary to
Mackie, causal directionality is not incompatible with determinism;
indeed, determinism is central to the above construction.
Despite its virtues, the analysis might seem too weak; for it as-
signs a direction to causation only for cases of temporally nonlocal
equations of motion. Whether or not this incompleteness is a defect
in the analysis will be investigated in the following sections. But
before turning to this matter, another doubt about the approach
must be mentioned.
The approach might be accused of smuggling in a causal asym-
metry. The construction underlying (Cl) and (C2) rests on the
notion that God intervenes directly by changing the force; the ac-

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH I9

celeration is changed indirectly through the modified force laws of


Try and T. But this distinction between direct and indirect inter-
vention seems to presuppose that force is a cause of acceleration
but not vice versa.
Admittedly, the form in which the above construction was pre-
sented was influenced by the common-sense intuitions about causa-
tion, but these intuitions are not an essential part of the analysis.
Describing God's intervention in terms of an "external force func-
tion" rather than in terms of an "external acceleration function"
was also motivated in part by the fact that, for temporally nonlocal
equations of the type considered above, the external force uniquely
determines the external acceleration, but not conversely; thus, the
intervention may not be completely described by giving only the
external acceleration. Of course, this very non-uniqueness might
possibly provide a way to escape attributing backward causal effects
to force; for it might be that, for each admissible external accelera-
tion function, there exists an external force function that is com-
patible with both the acceleration and the holding fixed of the past
history in accord with (Cl). Then, assuming that a Principle of
Redemption were in operation, whenever God intervened by di-
rectly changing the acceleration, the external force would adjust
itself so that the past was not violated. Fortunately, there is no need
to launch into the theology of Redemption, for we can supply cases
(the integro-differential Lorentz-Dirac equation being one such)
in which not every external acceleration can be matched by an
external force function that redeems the past.
In sum, the causal directionalities that the above analysis cap-
tures do not appear to depend in any essential way on prior judg-
ments of causal asymmetries. And, in any case, what is important
for the main thesis of this paper is that the backward and forward
directionalities revealed by (Cl) and (C2) stand or fall together. If
they both fall, so much the better for my skeptical conclusion.
VIII. THE LIMITATIONS OF INTERVENTION
In the preceding section, (Cl) and (C2) were touted as necessary
conditions, respectively, for the absence of backward causation and
the absence of forward causation .To accept them as sufficient con-
ditions as well would be disastrous for common sense: since New-
ton's theory satisfies (C2), it would follow that that force does not
have forward causal effects in his theory.
There is an independent reason not to accept (C1) and (C2) as suffi-
cient conditions. Consider the theory Th* where Fext is required to

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
20 THE JOURNAL OF PHILOSOPHY

satisfy:
Fext(t) = 0 for t > t, > O or t <-tl
Fext(t) h(t) for - t1 < t < t1

where h(t) may be nonzero. If, for T. forces have neither backward
nor forward causal effects, then it should be the case that
(C3)For every h such that MDTh # 0. there exists a map XI,: MDT > MDT*
such that (i) xh is bijective and (ii) for every s e MDT and every t such
that either t ? t1 or t < -t1, s(t) (t).
,X1ds)
The simultaneous satisfaction of (Cl) (with the time origin shifted
to -t1) and (C2) (with the time origin shifted to +tl) does not
guarantee the satisfaction of (C3) as we would expect if (Cl) and
(C2) were truly sufficient conditions.
If common sense escapes refutation, it also remains unsupported.
In most cases where we ordinarily want to assign a direction to
causation, no nontemporally nonlocal laws are at work, and both
(Cl) and (C2) will hold for the operant theories.
In cases where (Cl) does hold, we can consistently ask, "What
would have happened if the history as described by s had been the
same up to t = 0 but after t 0 the force had, by some miracle,
been made different?" And the answer seems unequivocal. (Df is
unique; so if the difference in forces is given by f, what would have
happened is described by 1D(s). Suppose now that attention is re-
stricted to those f's which are null for all t > t, > 0. In typical
cases, we would expect that bf
"[(f(S) 2(S)] for t > t, when f,
Z f2. In such cases, we would seem justified in attributing the dif-
ferences after t1 to the differences in the forces acting between t = 0
and t = ti-there is nothing else to attribute them to-and hence,
in attributing forward-directed causal effects to force.
Unfortunately, this story does not provide the sought-after basis
of causal directionality for the temporally local laws. The story rests
on the assumption that the history as described by s is held fixed up
to t = 0. But since, by hypothesis, (C2) also holds, we can equally
well imagine that the history as described by s is held fixed after
t 0. We can thus parrot the above story, substituting 'backward'
for 'forward' and reversing the time order. Thus, when (C2) is
satisfied, we can consistently ask, "What would have happened if
the history as described by s had been the same after t = 0 but the
force, by some miracle, had been made different before t = O?" And
the answer seems unequivocal. If the difference in forces is de-
scribed by g, then what would have happened is described by *I(s),
since "la is unique. Suppose now, says the parrot, that attention is

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH 2I

restricted to cases where the value of g is null for all t ? t1 < 0. In


typical cases, we would expect that '-4[*g1(S) -I+92(s)] for t ?-tt
when gl' g2. In such cases we would seem justified in attributing
the differences before t, to the forces acting
- between t = 0 and
t = -tl-there is nothing else to attribute them to-and hence, in
attributing backward causal effects to force.
We certainly want to reject the parrot's conclusion. But, unless
some relevant asymmetry can be supplied, we must also reject
the line of reasoning on which the parrot operates. The following
section explores some possible sources of relevant asymmetries.
IX. THE ROOTS OF A PREJUDICE
When we ask what would have happened in a possible world w if
someone or something had intervened, we usually proceed by
imagining a scenario in which events "run on" as they actually did
in w up to the time of the intervention. (Call this a type 1 scenario.)
But we could also proceed by imagining that events are the same
following the intervention as they actually were in w. (Call this a
type 2 scenario.) Assuming that the laws of w are temporally local, the
apparatus introduced in section VII shows how both types of scenario
can be set up for the case where the intervention involves the im-
position of additional forces. In both types of scenario, the "size"
of the miracle can be the same; i.e., the external forces can have
the same values.21 From what, then, does our favoritism toward type
1 scenarios derive? My suggestion is that this favoritism is a form of
prejudice which reflects certain linguistic, psychological, and epis-
temological asymmetries and that none of these asymmetries sup-
plies an objective physical basis for causal directionality.
The first source of the prejudice is given away by the very
words used to describe type 1 scenarios. We do tend to think of
events as "running on" until their flow is diverted by the interven-
tion. But it hardly needs to be argued here that this metaphor of
the "flow" of events from the past into the future is without any
hard philosophical cash value.
A second source derives from a combination of the first source
and a tendency to imagine the effort needed to construct possible
worlds. To build a world w' that is like w up to time t but different
thereafter, I can simply imagine myself letting w "run on" up to
time t (no sweat so far), creating a small localized miracle at t, and
21 Lewis, in Counterfactuals (Cambridge: Harvard, 1973), argues that the
"size" of a second miracle needed to assume reconvergence of a type 1 scenario
with w will typically have to be much larger than the "size" of the miracle that
started the divergence from w. To the extent that this is true, a temporally
reversed analogue also holds for the type 2 scenario.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
22 THE JOURNAL OF PHILOSOPHY

then resting from my brief labors while Nature takes Her course.
But to build a world w" that is like w after t but which diverges
from w for time before t, I have to imagine myself changing records,
traces, memories, etc., in w, and this is a sweaty job. Needless to
say, however, this imagined labor is not a relevant commodity. In so
far as it makes sense to talk about possible worlds, it makes sense
simply to take them as existing, and not to think of them as arti-
facts that have to be fashioned by human or divine labor. If our
comparisons of similarities and differences of possible worlds do
depend on imaginings of such labors, then so much the worse for
counterfactuals, causation, and anything else that depends on these
comparisons.
A third, though less important, source of prejudice derives from
the fact that we know much more about the past than about the
future and we also know that our knowledge claims about the past
tend to be much more reliable than our knowledge claims about
the future. In constructing possible scenarios we naturally tend to
defer to this asymmetry by favoring those scenarios which, ceteris
paribus, do as little violence as possible to our knowledge. Obvi-
ously, this epistemological asymmetry can at best provide only a
very partial explanation of prejudice in favor of type 1 scenarios;
for this favoritism is shown even in cases where the time in ques-
tion is so far in the past that we know little of what happened be-
fore t and much about what happened after t. But it is significant
that, if we know literally nothing about what happened before t,
our normal unwillingness to entertain type 2 scenarios is somewhat
diminished. Also it may be that our intuitions about counterfactual
situations are formed largely for cases where the epistemological
asymmetry does apply and that the resulting prejudice against type
2 scenarios is then carried over to cases where the asymmetry does
not apply.
Yet another impetus for the prejudice is our tendency to conceive
of the intervention situation in terms of human agency. For full-
bodied human actions are intentional, and intentionality is future-
directed.
No doubt our prejudice has other roots' as well, but the four
mentioned above illustrate the range of cases. The first two involve
the importation of confused or irrelevant notions. The third and
fourth involve genuine asymmetries in our attitudes toward the
past and future. But insofar as causation is something which exists
in the objects themselves and which acts independently of our atti-
tudes toward them, these asymmetries cannot 'be the source of

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
CAUSATION: A MATTER OF LIFE AND DEATH 23

causal directionality. Of course, it might be that a full account of


human knowledge and intentions will reveal that the asymmetries
in our mental attitudes rest on physical asymmetries which are
relevant to causal order; but until such a revelation occurs, we
must remain skeptical.
X. FAREWELL TO COMMON SENSE
Must we then bid farewell to common-sense intuitions about the
directionality of causation? Since parting is such sweet sorrow, it
will be prolonged by several considerations.
Much of the above discussion focuses on contexts where the rele-
vant laws are fully deterministic. Changing the focus to other cases
where the laws are neither futuristically nor historically determinis-
tic is not apt to produce any helpful asymmetries. Changing the
focus to cases where, say, futuristic but not historical determinism
reigns might produce some helpful asymmetries. But, to repeat,
there is no strong evidence of any basic physical laws that are
deterministic in one sense but not the other.
Another defense of common sense might start from the propo-
sition that commonsensical causal laws are often temporally non-
local; e.g., the opening of a sticking drawer (the effect) may re-
quire a prolonged pulling and tugging (the cause). But to defend
common sense in this way is to convict it of the sin it commits.
The appearance of a temporally nonlocal phenomenon results from
a crude course-graining of states: while the tugging is going on,
we may perceive no change of state of the drawer (it remains shut).
But even the most untutored common sense can be brought to ap-
preciate a finer-grained distinction between states under which the
final opening of the drawer is the culmination of a sequence of
states and under which the appearance of temporal nonlocality
disappears.
It seems then that if our only purchase on causal priority is
through something like the intervention model, we must, inescap-
ably, make our adieu to common sense. But one other way out
remains to be explored. It has, I am thankful to note, nothing to
do with ever more complicated comparisons of possible worlds.
Jerrold Aronson22 has suggested that a paradigm causal sequence
involves two bodies, that the cause body possesses some physical
quantity (e.g., momentum, kinetic energy, etc.) that is transferred
to the effect body, and that it is this transference of the relevant
quantity which constitutes the nexus of the cause-effect relation.
22 "The Legacy of Hume's Analysis of Causation," Studies in the History and

Philosophy of Science, ii (1971): 135-156.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
24 THE JOURNAL OF PHILOSOPHY

Moreover, the direction of transfer of the relevant quantity deter-


mines the direction of causal influence.
Note that, if the laws of motion are invariant with respect to a
nontrivial class of reference frames, Aronson's story will not always
identify the causal agent in an observer-independent fashion, for
different observers may see a different direction of transfer of the
relevant quantity. But let us leave aside such complications and
concentrate on a single observer.
Imagine that our observer inhabits a world governed by laws of
impact rather different from Newton's. In particular, let us suppose
that if our observer sees a body A which is initially at rest and
which has mass mA and a body B which has mass mB > mA and
which approaches A with a velocity v, then after impact he will see
both A and B move off with velocity v. Now it might be possible
for our hypothetical observer to preserve the transference-of-quanti-
ties idea by postulating that, say, in addition to its mechanical
momentum and kinetic energy, body B has stored up within itself
some other form of momentum or energy which is transferred to A
on contact. But suppose no such view can be worked through con-
sistently with the other principles of motion our observer discovers
in his world. Even so, I would claim, our hypothetical ob-
server, if he is anything like us, would still want to call the sequence
described above a causal sequence and would want to say that B's
impact on A caused A to move.23
It is not a deductive consequence of my parable that in the actual
world our intuitions about causation are not founded on and
justified by Aronson's transfer-of-quantities story. But I do think
the parable creates a strong presumption in favor of this conclu-
sion. And, in any case, I believe that an adequate account of causa-
tion must be broad enough to cover the possibility that Newton
might have been wrong about the laws of impact.
The transfer-of-quantities idea promised to provide a simple and
satisfying explanation of the common-sense belief in the coinci-
dence of causal order and temporal order. But to derive an explana-
tion of this coincidence from the coincidence of temporal order
and the transfer-of-quantities order, one needs an additional premise
to the effect that the transfer of quantities constitutes causation.
Our hypothetical world undermines this premise.
If pressed to justify their ascriptions of causation, both the New-
tonian observer, who does see a transfer of some quantity from B
23 Experimental evidence in support of this claim is to be found in A.
Michotte, The Perception of Causality (New York: Basic Books, 1963).

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions
BOOK REVIEWS 25

to A, and our hypothetical observer, who sees no such transfer, would


be driven to fall back on assertions about how A would have be-
haved differently if B's initial motion had been different. And here
the Newtonian and our hypothetical observer are on equal footing;
their counterfactuals both stand or both fall. The footing, as we
have seen, is seriously unfirm.
XI. CONCLUSION: THE DEAD DEAD
The search for the basis for causal directionality led us to investi-
gate "backward causation." The investigation revealed that tem-
poral order cannot be the key to causal order, since there can be
cases in which the effect precedes its cause. But, more importantly,
a study of alleged cases of backward causation helped to pin-point
the reasons for assigning a forward or a backward direction to
causation. In certain physically coherent cases involving temporally
nonlocal equations of motion, the reasons can be grounded in
genuine physical asymmetries. There is, however, no good evidence
that such temporally nonlocal laws are at work in any of the cases
where we commonly talk of causal directionality, and the reasons
for assigning a forward direction to causation in these cases were
found to derive from a felt determination of the mind to discrimi-
nate in favor of the future direction in time. This felt determina-
tion could not be linked to any relevant physical asymmetry. Of
course, no proof of the nonexistence of such a link has been given,
but the weight of the evidence is in favor of this conclusion.
The assumption of a directionality to causation is the last vestige
of the doctrine that causes have a "power" or "efficacy." It is high
time that we paid more than lip service to Hume's rejection of this
doctrine.
JOHN EARMAN
University of Minnesota

BOOK REVIEWS
Thomas Reid's Inquiry. NORMAN DANIELS. New York: Burt Frank-
lin, 1975. $12.95.
There are many occasions in the history of science in which an
important scientific discovery, normally attributed to one scientist,
can (with some justice) be attributed to an earlier figure. The
views of the earlier figure had little influence, however, and the
discovery in question really entered the scientific literature as a
result of the work of the later scientist.

This content downloaded from 155.69.24.171 on Sat, 21 Nov 2015 05:44:58 UTC
All use subject to JSTOR Terms and Conditions

You might also like