Vous êtes sur la page 1sur 17

TIME TRAVEL AND CHANGING THE PAST:

(OR HOW TO KILL YOURSELF AND


LIVE TO TELL THE TALE)
G.C. Goddu
Abstract
According to the prevailing sentiment, changing the past is logi-
cally impossible. The prevailing sentiment is wrong. In this paper, I
argue that the claim that changing the past entails a contradiction
ultimately rests upon an empirical assumption, and so the conclu-
sion that changing the past is logically impossible is to be resisted. I
then present and discuss a model of time which drops the empiri-
cal assumption and coherently models changing the past. Finally, I
defend the model, and changing the past, against objections.
Oh, if you could but travel back in time, just think what you could
do. You could provide advance warning and prevent the assassi-
nation of Abraham Lincoln or the bombing of Pearl Harbor. You
could prevent Caligula or Hitler or Pol Pot from ever being born.
You could bring back a cure for the Black Death and prevent the
decimation of Europe. You could play the stock market and
become richer than Bill Gates. You could . . . .
According to almost all philosophers writing about time travel,
however, these speculations about potential time travel exploits
are woefully misguided, for while time travel itself is logically
possible, changing the past is not. For example, in Tips for Time
Travellers, Monte Cook provides the following advice for prospec-
tive time travellers: Dont worry about changing the past. You
cant.
1
John Hospers writes, Not all the kings horses or all the
kings men could make what has happened not have happened,
for this is a logical impossibility.
2
Gilbert Fulmer argues that the
idea of changing the past is logically incoherent, and, therefore,
the act logically impossible.
3
According to Geoffrey Brown, not
1
Monte Cook, Tips for Time Travellers, in Philosophers Look At Science Fiction, ed.
Nicholas Smith, (Chicago: Nelson-Hall, 1982), p. 49.
2
John Hospers, An Introduction to Philosophical Analysis, 2nd ed. (Englewood Cliffs, NJ:
Prentice Hall, 1967), p. 177.
3
Gilbert Fulmer, Cosmological Implications of Time Travel, in The Intersection of
Science Fiction and Philosophy, ed. R.E. Meyers (Westport: Greenwood Press, 1983), p. 33.
Blackwell Publishing Ltd. 2003, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350
Main Street, Malden, MA 02148, USA.
Ratio (new series) XVI 1 March 2003 00340006
even God has the ability to alter the past (which is admittedly
senseless).
4
Numerous other authors also maintain the logical
impossibility of changing the past.
5
Hence, according to the
prevailing sentiment, while traveling to the past is perfectly coher-
ent, changing it is not. You just could not prevent the bombing of
Pearl Harbor or the birth of Hitler, even if you could travel back
in time.
In this paper, I shall argue that the prevailing sentiment is
wrong. Not only is time travel logically possible, but so is chang-
ing the past. In section I, I shall argue that the claim that chang-
ing the past entails a contradiction ultimately rests upon an
empirical assumption, and so the conclusion that changing the
past is logically impossible is to be resisted. In section II, I shall
present my own model of time travel and argue that it coherently
models changing the past. In section III, I shall defend the model
against objections. Finally, I shall conclude, contra the prevailing
view, that changing the past is logically possible.
I. Changing the past
What exactly is involved in changing the past? One might say that
the past is all that has come before now. Hence, tomorrow the
past will be different. All that comes before tomorrow includes
more than all that comes before today. Hence, the past has
changed. What the past was today differs from what the past will
be tomorrow, or even one second from now. This sense of chang-
ing the past is not at issue. At issue, rather is, can the past be
TIME TRAVEL AND CHANGING THE PAST 17
Blackwell Publishing Ltd. 2003
4
Geoffrey Brown, Praying About the Past, Philosophical Quarterly 35 (1985), p. 85.
5
For example, see Bob Brier, Magicians, Alarm Clocks, and Backward Causation,
Southern Journal of Philosophy 11 (1973), p. 361; William Lane Craig, Tachyons, Time Travel
and Divine Omniscience, Journal of Philosophy 85 (1988), p. 147; Larry Dwyer, Time Travel
and Changing the Past, Philosophical Studies 27 (1975), p. 347; John Earman, Implications
of Causal Propagation Outside the Null Cone, Australasian Journal of Philosophy 50 (1972),
p. 232; Antony Flew, Can A Cause Precede Its Effect?, Aristotelian Society Supplementary
Volume 28 (1954), p. 48; Samuel Gorovitz, Leaving the Past Alone, Philosophical Review 73
(1964), p. 367; Richard Hanley, The Metaphysics of Star Trek (New York: Basic Books, 1997),
p. 205; Jonathan Harrison, Dr. Who and the Philosophers or Time Travel for Beginners,
Aristotelian Society Supplementary Volume 45 (1971), p. 7; Paul Horwich, On Some Alleged
Paradoxes of Time Travel, Journal of Philosophy 72 (1975), pp. 43536; Murray MacBeath,
Who Was Dr Whos Father?, Synthese 51 (1982), p. 423; Paul J. Nahin, Time Machines,
(New York: AIP, 1993), p. 181, p. 209; Larry Niven, The Theory and Practice of Time
Travel in All the Myriad Ways (Ballantine Books, 1971), p. 113, p. 121; Hilary Putnam, It
Aint Necessarily So, Journal of Philosophy 59 (1962), p. 669; J.J.C. Smart, Is Time Travel
Possible?, Journal of Philosophy 60 (1963), p. 241; R. Swinburne, Space and Time (Macmillan
& Co., 1968), p. 161.
undone? Can some part or event of the past be made not to be a
part or event of the past? Consider, for example, the following
time travel scenarios:
Scenario 1: Paul, at the age of thirty, departs from February 12,
1998, for the past and arrives on January 28, 1972. On this trip,
in a fit of self-nihilism, he finds and kills his three-year-old self,
after which he returns to 1998.
Scenario 2: Sarah has just completed building her time
machine. She decides to test the machine on herself tomorrow
morning at which time she intends to travel back one day. In
the meantime, she goes home, puts some salve on the burn she
received that day, and goes to bed. In the morning, Sarah, with
coffee in hand, sits down to read the morning paper. She opens
the paper to the following headline: Famous physicist found
dead. On the front page is a picture of her body, salved burn
clearly visible on her arm, inside her pristine time machine.
Underneath is the caption: Nobel-prize winning physicist
found dead yesterday in mysterious device that materialized
near city hall. Extremely shaken, Sarah returns to the lab and
destroys the machine.
Are these scenarios logically coherent? Can the thirty-year-old
Paul, for example, make it such that he does not survive past the
age of three? Can Sarah make it such that she does not appear
dead in her time machine the previous day? The standard answer
to these questions is no? But why?
According to R. Swinburne, To change something is to make
it different from what it was at another temporal instant. I cannot
make a thing at t
1
different from what it is, was, or will be at t
1

for this would imply that it both did and did not have some prop-
erty at t
1
.
6
For example, assume that thirty-year-old Paul was never
killed and resurrected. Paul cannot, therefore, at some time, say
1998, change any instant of 1972 and make it such that thirty-year-
old Paul kills his three-year-old self. To do so would make some
instant of 1972 such that Paul is both killed and not killed during
that instant and make it the case that Paul both survives past the
age of three and does not survive past the age of three. Similarly,
since Sarah dies yesterday in her time machine, she cannot now
prevent herself from getting into her time machine and make it
18 G.C. GODDU
Blackwell Publishing Ltd. 2003
6
Swinburne, p. 161. See also, Dwyer, p. 348; Harrison, pp. 23; Horwich, p. 435;
Hospers, p. 177; and Smart, p. 241.
the case that she does not die yesterday, for then some instant of
yesterday would be such that both Sarah appears in her time
machine during that instant and Sarah does not appear in her
time machine during that instant. According to Swinburne,
therefore, changing the past makes it the case that things simul-
taneously have contradictory properties, which is logically impos-
sible, and so changing the past must be logically impossible.
But consider again Swinburnes claim that to change some-
thing is to make it different from what it was at another temporal
instant. According to Swinburnes claim, Paul changes if Paul is
one way at one temporal instant and another way at a different
temporal instant. For example, Paul, at some temporal instant in
1970 is under three feet tall. At some temporal instant in 1998,
Paul is over five feet tall. Hence, Paul has changed from being
under three feet tall at one temporal instant to being over five
feet tall at another. But what is it for some instant of 1972, call it
i
1
, to change? Presumably, at one temporal instant i
1
was one way,
for example with Paul not killed, but at another temporal instant
i
1
was different, with Paul killed. But is the result of applying
Swinburnes claim to i
1
a contradiction?
To begin to see how to answer, consider David Lewis statement
of why one cannot change the past. According to Lewis, If
change is qualitative difference between temporal parts of some-
thing, then what doesnt have temporal parts cant change. For
instance, numbers cant change; nor can the events of any
moment of time, since they cannot be divided into dissimilar
temporal parts.
7
Why then can one make sense of Paul changing,
but apparently not the events of particular instants of 1972 chang-
ing? Because instants or the events of particular instants are not
the sort of thing that have temporal parts in the same way that
Paul has temporal parts. Paul exists at many times, eg. the first day
of 1978, the first day of 1988, and the first day 1998. On the other
hand, neither i
1
nor the events of i
1
exist at many times. Neither
i
1
nor the events of i
1
exist during 2000 B.C.E or January 1, 1998,
or any other instant. If i
1
can be properly said to exist at a time at
all, then i
1
exists at i
1
.
But is the claim that the events of a temporal instant do not
themselves have temporal parts a logical truth? Isnt it possible,
even physically, that the events of a particular temporal instant
TIME TRAVEL AND CHANGING THE PAST 19
Blackwell Publishing Ltd. 2003
7
David Lewis, The Paradoxes of Time Travel, in Philosophical Papers, vol II. (New York:
Oxford University Press, 1986), p. 69.
might have some kind of temporal part? Might not the best expla-
nation of the nature of our universe ultimately involve two (or
more) temporal dimensions? Lewis himself explicitly restricts his
claim about the impossibility of changing temporal instants to
universes with a single temporal dimension and acknowledges
that time could be two-dimensional and that an event might be
momentary along one time dimension but divisible along the
other.
8
Presumably the exact nature of the universe and time is a
matter for empirical science to discover and so the claim that the
events of temporal instants do not have temporal parts is an
empirical truth, not a logical one. Hence, the claim, At one time,
the temporal instant i
1
was one way, but at another time was differ-
ent is meaningless or contradictory only given the empirical
assumption that temporal instants themselves lack temporal
parts. Since Swinburne and others are claiming that changing the
past is logically impossible, empirical assumptions about the
nature of time will not necessarily support their claim.
Drop then the assumption that temporal instants or events of
temporal instants do not themselves have temporal parts and
suppose for the moment that temporal instants do have temporal
parts. Call the temporal parts of temporal instants hypertimes.
Swinburnes claim should, in the case of events of particular
temporal moments, be rephrased as To change events of partic-
ular temporal moments is to make them different from what they
were at another hypertemporal instant. Suppose, therefore, that i
1
at hypertemporal instant h
1
, is such that Paul is not killed, but
that at some later hypertemporal instant, say h
2000
, Paul is killed
during i
1
. i
1
has therefore changed from having Paul survive to
having Paul not survive. At one hypertime, i
1
was such that Paul
was not killed, but at a later hypertime i
1
is such that Paul is killed.
Thus, even though at a particular instant, viz. i
1
, Paul is both
killed and not killed, Swinburne is wrong to claim that a contra-
diction results. A contradiction results only if one assumes that
temporal instants themselves do not have temporal parts.
Otherwise, i
1
is one way at one hypertime and a different way at
another, just as Paul is under three feet tall at one time and over
three feet tall at another. Neither change entails a contradiction.
To reiterate: Introducing the, admittedly abstract, notion of
hypertimes allows one to make sense of the statement At one
time, the temporal instant i
1
was one way, but at another time was
20 G.C. GODDU
Blackwell Publishing Ltd. 2003
8
Lewis, p. 69.
different. At one hypertemporal instant, the temporal instant i
1
was one way, but at another hypertemporal instant was different.
Given that changing temporal instants does not entail that tempo-
ral instants or the events of temporal instants have contradictory
properties at one and the same hypertemporal instant, changing
temporal instants or the events of temporal instants does not
entail a contradiction. Hence, the standard claim that changing
the past is logically impossible is to be resisted, since the basis for
it is an empirical assumption about the nature of time, viz. that
temporal instants do not themselves have temporal parts.
II. A consistent model of changing the past
Is there a way of coherently modeling a universe in which the past
changes? I think there is. Consider first a standard video cassette
recorder and monitor set up. The VCR has a clock that displays
normal, everyday time call this time clock-time. The monitor,
on the other hand, displays how much time has elapsed on the
video cassette being viewed call this time cassette-time. When
a cassette is played forward at normal speed, clock-time and
cassette-time are synchronous, i.e. for every second of cassette-
time that passes one second of clock-time passes. When a cassette
is rewound (or advanced), however, cassette-time decreases (or
increases) much more rapidly than clock-time. For example,
suppose the clock reads 16:00:00 and the monitor, 0:00:00. Now
let a cassette play the first twenty minutes of the Star Trek episode,
City on the Edge of Forever. At the end of twenty minutes the
clock reads 16:20:00 and the monitor displays 00:20:00. Now let
the cassette be rewound for fifteen (clock-time) seconds until the
monitor display reads 0:05:00. The clock now reads 16:20:15. Now
let the cassette record twenty minutes of the episode, Yesterdays
Enterprise. When the cassette stops after twenty minutes the
clock reads 16:40:15 and the display reads 0:25:00. The interval
between 0:05:00 and 0:20:00 has occurred twice, though at differ-
ent clock-times. The images and sounds on the cassette during
that interval are now different from what they once were. Before
16:20:15 the cassette-time interval between 0:05:00 and 0:20:00
contained parts of The City on the Edge of Forever, but by
16:35:15 that same temporal interval contains parts of Yesterdays
Enterprise. The content of the cassette has been changed.
The VCR and monitor set up provides one conceptual basis for
understanding time travel in a universe in which temporal
TIME TRAVEL AND CHANGING THE PAST 21
Blackwell Publishing Ltd. 2003
moments have hypertemporal parts. Consider then a hypothetical
universe that has two temporal dimensions related much like
clock-time and cassette-time are related. Let the images and
sounds on the cassette over a particular interval be analogues of
the events of the world over a particular interval. Since the inter-
vals on the cassette are measured in cassette-time, cassette-time
will be an analogue of the time we live in; call this normal time.
Clock-time is then an analogue of hypertime. If all proceeds
normally, i.e. no travelling in normal time occurs, then there will
be a one-to-one correspondence between hypertime and normal
time. January 1, 1978 in normal time will correspond to January
1, 1978 in hypertime. There will be no duplication of normal time
points or intervals. Travelling backward in normal time will be an
analogue of rewinding the cassette. (Instantaneous travel would
be like a jump back on the tape.) If time travel into the past
occurs, then the one-to-one correlation between hypertime and
normal time will fail to hold. Some points of normal time will
correlate with two or more distinct points of hypertime. Just as, in
the case above, 0:10:00 of cassette-time corresponds with both
16:10:00 and 16:25:15 of clock-time, so might January 1, 1978 of
normal time correspond with both January 1, 1978 and January
1, 2078 of hypertime. Finally, changing the events of previous
times, i.e. changing the past, will be an analogue of recording new
images and sounds on the cassette.
Consider Paul from Scenario 1. On February 12, 1998 he
departs for the past. Assume his trip takes one instant of hyper-
time. Paul arrives one hypertime instant later at normal time
January 28, 1972. On this trip, he finds and kills his three-year-old
self, after which he returns to 1998. The following rough diagram
can be used to help explain Pauls travels in terms of normal time
and hypertime.
Hypertime: T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18
Normal time: t1 t2 t3 t4 t5 t6 t7 t8 t9 t3 t4 t5 t6 t7 t8 t9 t10 t11
Universe State: e1 e2 e3 e4 e5 e6 e7 e8 e9 e3 e4 e5 e6 e7 e8 e9 e10 e11
Let t1 be February 12, 1968, t3 be the afternoon of January 28,
1972, and t9 be February 12, 1998. Part of e1 includes Pauls
birth. e2e9 includes Pauls life until February 12, 1998; Paul
grows up, acquires his time machine and on February 12, 1998
(T9) he departs. One instant rewind later he arrives at January
28, 1972 (t3 at T10). His mere arrival changes the past. Before,
the state of the universe associated with t3, viz. e3, included the
22 G.C. GODDU
Blackwell Publishing Ltd. 2003
presence of only one Paul. Now that state has changed and
become e3 and includes the presence of two Pauls; a young one
and an older one who mysteriously appeared out of thin air. And
so the re-recording continues. At t4(T11) Paul kills the younger
version of himself and returns to the future. Paul arrives at
t10(T17) and continues on with his life. E4 includes the
murder/suicide of the three-year-old Paul and the subsequent
departure of the thirty-year-old Paul. E5 perhaps includes the
futile search for the murderer and so on up to e10 which includes
the appearance out of thin air of the thirty-year-old Paul.
Paul has changed the past from being one way, his younger self
living, to being another, his younger self not living. The fact that
some of the es are italicized just means that those events have been
changed and are no longer accessible (just as the recorded over
video images are no longer accessible). They are not a part of the
way the normal time world is, though there was a hypertime at
which they were. A normal time historian at t11(T18) would
describe the order of events as, e1 e2 e3 e4 e5 etc, while a hyper-
time historian, if there could be such a being, would describe events
as e1 e2, the way the interval t3t9 used to be, i.e. e3e9, the way t3t9
is now, i.e. e3e9, e10, e11. Hence, a hypertime historian would be
able to explain the origin of time travellers and the origin of the
time travellers memories while a normal time historian would not.
One significant divergence between the time travel case and the
VCR set up requires mention. When one is done recording on a
normal VCR one leaves all images and sound after the period of
re-recording untouched. Hence, a cassette could have five minutes
of one program, five minutes of another and then a sudden cut
back to the first program. Pauls time travels, however, leave noth-
ing after his arrival in the past, the moment re-recording starts,
untouched. The reason for the divergence is that images and
sounds on different parts of the tape are not causally related to
each other, but rather to the source of the recording. Events, on
the other hand, are causally related to each other through time.
For example, the events of e4, the death of Paul, cause the inves-
tigation of e5, etc. Thus, Pauls brief appearance in the past sets in
motion a complete re-recording of everything he knew, e3e9.
When Paul arrives back at t10 in the future he will have to live with
the consequences of the changes he has wrought.
9
TIME TRAVEL AND CHANGING THE PAST 23
Blackwell Publishing Ltd. 2003
9
As a result of a choice concerning ease of diagramming, time travel into the future,
given the current model, is not analogous to fast forwarding the tape in a less significant
What of Sarah and her time machine? Sarah, when she exam-
ines the paper, has not yet travelled in time, so the normal time
and hypertime lines should match perfectly. But if Sarah does not
get into her machine she changes the past even without time trav-
elling, since she fails to appear in the past. How then will the
current model account for Sarah? The problem is easily resolved
once one realizes which part of the normal time-line the Sarah
story actually describes. The Sarah story describes events happen-
ing during a re-recording. Consider again our diagram:
Hypertime: T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18
Normal time: t1 t2 t3 t4 t5 t6 t7 t8 t9 t3 t4 t5 t6 t7 t8 t9 t10 t11
Universe State: e1 e2 e3 e4 e5 e6 e7 e8 e9 e3 e4 e5 e6 e7 e8 e9 e10 e11
At t3(T3) Sarah does not appear dead in her machine in the
middle of town. Thus, the next morning [t6(T6)] when she reads
the paper, there is no front page story detailing such an event
because it did not happen. The same afternoon [t9/T9] Sarah
travels back in normal time to t3, somehow dying in the process.
This event occurs at T10 in hypertime. The events of t3 have now
changed because these events now, T10, include the arrival of
dead Sarah out of thin air, whereas previously, T3, they did not.
Thus, at t6(T13) Sarah reads the paper and discovers to her
horror that if she gets into the machine she will die. Thus, at
t9(T16) Sarah, paradox free, destroys her time machine. No para-
dox results because, hypertemporally at least, Sarah has already
gotten into her machine and the consequences of that action tip
Sarah off not to get into her time machine at the later hypertime.
Are Sarahs and Pauls escapades truly examples of changing
the past? Granted, neither Sarah nor Paul travel backwards in
hypertime and hence never alter the hypertemporal past. But
Sarah and Paul are perceivers of normal time and the only past
they, or anyone else they interact with, care about is the normal-
time past. All our history books are descriptions of the normal-
time past. Therefore, given that before Pauls journey back to kill
24 G.C. GODDU
Blackwell Publishing Ltd. 2003
manner as well. Time travel into the future involves a jump in both hypertime and normal
time. One can avoid jumps in hypertime by one of at least two methods. Firstly, time travel
into the future, unlike the past, could be by standard relativistic methods. That is, in order
to get back to the future one does not just push the button and voila!, but rather the
machine accelerates away from Earth at some significant fraction of light speed and then
returns. Secondly, one could suppose that travel to the future, like travel to the past,
requires at least one instant of hypertime, but during that instant all the intervening
normal time is played out. Hence, T12 might correspond to the entire period t5-t9. Other
methods, and hence other models, are most likely possible.
his younger self the true description of the normal-time past
included Paul surviving past the age of three, but that after his
successful journey the true description of the normal-time past
does not, we must conclude that Paul can and does change the
past. Given that before Sarahs time travel the true description of
the normal- time past did not include her appearing dead, but that
after her travel it does, we must conclude that Sarah can and does
change the past. Hence, changing the past is logically possible.
III. Objections and replies
Objection I: Two-dimensional time or the temporal slices of
objects recurring just seems unimaginable. So how can any model
involving the recurrence of temporal instants or slices of objects
be coherent?
Reply: Imagining two-dimensional time or imagining temporal
slices enduring or recurring is admittedly difficult. So is imagin-
ing a hypercube or imagining two straight lines in the same plane
that are not a constant distance apart and yet never intersect.
Regardless, hypercubes are straightforward four-dimensional
objects which can be described as follows: take a three-dimen-
sional cube with edges of length one; project this cube one unit
along the spatial dimension at right angles to the normal three;
connect the corresponding vertices of the original and projected
cubes; the result is a hypercube. Pre-Lobachevski, two straight
lines in the same plane never intersecting and yet not a constant
distance apart, was thought absurd, unimaginable, impossible,
etc. Yet, even if difficult to truly imagine, such lines are a perfectly
coherent part of hyperbolic plane geometry. Similarly, while
imagining temporal slices of objects enduring or recurring is
difficult, there are perfectly coherent descriptions of the
endurance or recurrence of temporal slices. For example, things
endure or recur by existing at more than one time and since
temporal slices of objects are momentary in normal time, for such
slices to endure or recur involves existing at more than one
moment of some kind of time other than normal time. On my
model, for example, if time travel occurs, temporal slices of
objects recur at later hypertimes.
10
TIME TRAVEL AND CHANGING THE PAST 25
Blackwell Publishing Ltd. 2003
10
For other examples of descriptions of multi-dimensional time see, C.T.K. Chari, A
Note on Multi-Dimensional Time, British Journal for the Philosophy of Science 8, (1957): pp.
155158; and T.E. Wilkerson, Time and Time Again, Philosophy 48, (1973): pp. 173177.
Objection II: Whatever the ts are in the model they are not
times. The only real times are the hypertimes. Since the hyper-
temporal past never changes, if the hypertemporal moments are
the only real temporal moments, then the model does not model
changing the past.
Reply: Why might one think that the hypertimes are the only
real times? Because one might think that e1 causes e2 and e2
causes e3 and so on and when Paul pushes the button on his time
machine at e9 he causes the next state to be e3 and e3 causes e4
and so on. If one holds that the ordering of causation determines
the order of time, then one might think the universe being
described by my model is better represented and described as
follows:
Hypertime: T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18
Universe State: e1 e2 e3 e4 e5 e6 e7 e8 e9 e10 e11 e12 e13 e14 e15 e16 e17 e18
Paul pushes the button on his time machine at T9. The result,
however, is not time travel to 1972, i.e. T3, but rather a drastic,
though instantaneous, change in the universe the universe
becomes a duplicate of the state of the universe in 1972, with the
exception that Paul and his time machine are present. (This
duplicate is e10, for it is not a changed version of e3, but rather a
completely new state of the universe qualitatively very similar to
e3). In other words, Pauls button pressing does not cause him to
travel in time, but rather it causes the universe to become such
that at the very next instant, T10, the universe is almost in the
same state as it was in 1972.
The world just described is a logically possible world.
Regardless, this world is not the world my model is describing, for
both the causal and the temporal relations differ significantly. In
the world of my model temporal slices of objects can become
hypertemporally extended. How so? By the activation of time
machines. When Paul pushes the button on his time machine at
T9 what happens is that (i) the temporal slice at t3 becomes
hypertemporally extended and (ii) Paul and his time machine
appear in the hypertemporal extension of the slice. The events
of t2, viz. e2, are still the cause of (most of) the events of t3, but
because the events of t3 are now(T10) also caused by e9, the
events of t3 at T10 are e10. Finally, because the events of t3 are
hypertemporally extended, the events of each time up to Pauls
original departure become, in turn, hypertemporally extended.
26 G.C. GODDU
Blackwell Publishing Ltd. 2003
The universe-change world is not the world described by my
model.
11
Objection III: Of planar two-dimensional models of time, Lewis
writes:
On closer inspection, however, this account seems not to give
us time travel as we know it from the stories. When the traveller
revisits the days of his childhood, will his playmates be there to
meet him? No; he has not reached the part of the plane of time
where they are. He is no longer separated from them along one
of the two dimensions of time, but he is still separated from
them along the other.
12
Doesnt Lewis objection also apply to the new model? After all,
the temporal part of the three-year-old Paul that leads to the time
traveller is the one in e4 at t4(T4), but the temporal part that is
killed is the one in e4 at t4(T11). Hence, one might argue that
we have yet to reach the relevant temporal part of the younger
Paul.
Reply: I cannot deny that different hypertemporal parts of
three-year-old Paul are involved in (i) Paul living past the age of
three and (ii) dying at the age of three. After all, different hyper-
temporal parts are necessary to avoid contradiction. Regardless, I
am not persuaded that Paul fails to reach the relevant part for
three reasons. Firstly, e4 is a hypertemporal extension of e4, just
as thirty-year-old Paul is a temporal extension of three-year-old
Paul. Put another way, e4 is the way e4 is now (hypertemporally)
just as thirty-year-old Paul is the way three-year-old Paul is now
(temporally). Secondly, e2, the state of the universe in which Paul
turns three, is an immediate causal antecedent of both e3 and e3
and so a causal antecendent of both e4 and e4. For e3, e2 is the
sole immediate causal antecedent. e3, on the other hand has two
immediate causal antecedents e2 and e9. Either way e2 is
TIME TRAVEL AND CHANGING THE PAST 27
Blackwell Publishing Ltd. 2003
11
The universe-change world, and other similar worlds, may raise epistemological prob-
lems for any would-be time traveller. For example, how could Paul know whether he was
in a world in which he actually travelled in time or one in which the rest of the universe
changed? Note that the epistemological problem is not restricted merely to alleged cases
of changing the past. For almost all purported time travel stories, regardless of whether or
not the story advocates changing the past, one could describe a universe in which there is
no time travel at all, but merely a radical change in certain successive states of the universe.
For a discussion of the problem of scepticism as it relates to time travellers, see Roy A.
Sorenson, Time Travel, Parahistory, and Hume, Philosophy 62 (1987): pp. 227236.
12
Lewis, p. 68. For an example of the sort of model Lewis is discussing, see Jack
Meiland, A Two Dimensional Passage Model of Time for Time Travel, Philosophical
Studies 26, (1974): pp. 153173.
common to both and it is partly the fact that the state of the
universe that Paul reaches, e4, is caused by e2, the state of the
universe in which the one and only Paul turns three, which makes
it the relevant part of the universe. Thirdly, Paul has to live with
the consequences of all his actions. When Paul kills his younger
self, the future to which Paul returns will be one in which he
never went to school or paid taxes or built a time machine.
Hence, Paul has reached a relevant part of the past because e4 is
the hyptertemporal extension of e4 and e4 has both the right
causal antecedents and the relevant causal consequences.
Objection IV: Consider a standard branching model, such as:
Time: t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 t12 t13 t14 t15 t16 t17
e3 e4 e5 e6 e7 e8 e9 e10 e11 e12 e13 e14 e15 e16 e17
Universe State: e1 e2
e3
1
e4
1
e5
1
e6
1
e7
1
e8
1
e9
1
e10
1
e11
1
e12
1
e13
1
e14
1
e15
1
e16
1
e17
1
At t3, the afternoon of January 28, 1972, the universe branches.
On the top branch no time traveller Paul appears and Paul
survives past the age of three to build his time machine. At t9 on
the upper branch Paul departs for the past and arrives at e3
1
on
the bottom branch. At t4 Paul kills the younger version of himself
on the bottom branch and returns to the future, again on the
bottom branch. Paul arrives at t10 and continues on with his life.
e3
1
includes the murder/suicide of the three-year-old Paul and
the subsequent departure of the thirty-year-old Paul. e4
1
perhaps
includes the futile search for the murderer and so on up to e10
1
which includes the appearance out of thin air of the thirty-year-
old Paul. On this model the past never changes Paul merely
makes the bottom branch the way it always will be. At no time do
the events of t3 change the events of t3 are e3 and e3
1
and those
events remain unchanged.
So, the objection goes, (i) the scenarios can be accounted for
on standard branching universe models
13
and (ii) your model is
only a complicated branching universe model and the past does
not change in branching universe models.
14
Hence, your model
does not show that changing the past is logically possible.
Reply: First, the fact that the scenarios can be accounted for or
shown consistent on a branching universe model is irrelevant to
the project at hand. The project is to produce one consistent
28 G.C. GODDU
Blackwell Publishing Ltd. 2003
13
See for example, Paul Sukys, Lifting the Scientific Veil, (New York: Rowman &
Littlefield, 1999), p. 238.
14
Lewis, p. 80.
model that both (a) accounts for all the events in the scenarios
consistently and (b) allows the past to change. The fact that there
are models that do (a), but not (b) is irrelevant.
Second, I deny at least one of the conjuncts of (ii). In other
words, either my model is not just a complicated branching model
or changing the past is possible on some (though clearly not all)
branching models.
I take it that my model is complicated because it, unlike the
standard model presented above, (a) appeals to hypertimes and
(b) has branches that terminate. So the objector might claim that
my model is just the following branching model:
Time: t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 t12 t13 t14 t15 t16 t17
e3 e4 e5 e6 e7 e8 e9
Universe State: e1 e2
e3
1
e4
1
e5
1
e6
1
e7
1
e8
1
e9
1
e10
1
e11
1
e12
1
e13
1
e14
1
e15
1
e16
1
e17
1
Hypertemporal references have been suppressed, but events
e1e9 occur at hypertimes T1T9 respectively and events
e3
1
e17
1
occur at hypertimes, T10T24 respectively. On this
model e3 occurs at T3 and e3
1
at T10 even though both occur at
normal time t3. In addition, Pauls departure from e9 terminates
the one branch and his arrival at e3
1
at hypertime T10 starts
another branch.
Is the complicated branching model my model? The answer
hinges on whether e3
1
is a hypertemporal extension of e3. If the
branching on the complicated model is like the branching on the
standard model, then e3
1
is not a hypertemporal extension of e3.
On the standard model e3 and e3
1
are not related as thirty-year-
old Paul and three-year-old Paul are related, but more like two
amoebas which result from the fission of one. On my model e3
and e3 are hypertemporal parts of one event slice, the event slice
of t3, just as thirty-year-old Paul and three-year-old Paul are
temporal parts of one object, Paul. Hence, if e3 and e3 are not
related in a manner similar to the manner in which thirty-year-old
and three-year-old Paul are related, then the complicated branch-
ing model is not my model.
Suppose e3
1
is a hypertemporal extension of e3. In other
words, not only do the hypertimes mark when the universe
branches, but the universes branching just is e3 becoming hyper-
temporally extended as e3
1
. But then, regardless of whether the
model is my model or not, on the complicated branching model
the past changes. At one time, T3, the events of t3 were one way,
viz. e3, and at another, T10, the events of t3 are different, viz. e3
1
.
TIME TRAVEL AND CHANGING THE PAST 29
Blackwell Publishing Ltd. 2003
More generally, terminated branches just are the ways the past
used to be, and the unterminated branch is the way the past is
now.
Either my model is the complicated branching model or it
isnt. If it isnt, then the complicated branching model is irrele-
vant to the question of whether or not my model demonstrates
the logical possibility of changing the past. If it is, then there are
branching models, viz. the complicated one, on which the past
changes.
Objection V: If Pauls surviving past the age of three is part of
the universe, as it surely must be if Paul is to build his time
machine and travel in time, how can one say that the past has
been changed. If Pauls surviving past the age of three is always
part of the universe, isnt it always part of the past?
Reply: To change something, the thing must first be one way.
For Paul to change from being under four feet tall to being over
four feet tall, he must first be under four feet tall. If one consid-
ers the entire universe, i.e. the universe in all its spatial expanse
from the beginning of time to the end of time, part of that
universe will contain Paul being under four feet tall, while a later
part will contain Paul being over four feet tall. Likewise, for the
past to change it must first be one way and then later another. For
example, to change the events of t4 those events must first be one
way, in Pauls case e4 at T4, and then later some other way, e4 at
T11. If one considers the entire hypertemporal history of the
universe, part of the universe, e4, will contain Paul surviving past
the age of three.
Will Pauls surviving past the age of three always be part of the
past? Assuming no travel into the hypertime past occurs, e4 will
always be part of the hypertemporal past when a hypertemporal
historian at T18 looks back at what has come before, e4 will be the
events that occured at T4. But e4 is not always a part of the
normal-time past the past that Pauls or Sarahs or our accurate
history books describe. At hypertimes T4 and T5 e4 is part of the
normal-time past, but by hypertime T11 it is not. So the events of
t4 are at one hypertime the way the events of normal time were,
i.e. part of the universe, and at some later hypertime not the way
the events of normal time are, i.e. not part of the normal-time
past, though still part of the universe. Thus, contrary to the objec-
tion, even if the events of t4 are a part of the universe, as they
must be if they are to be changed, those events are not necessar-
ily always a part of the past.
30 G.C. GODDU
Blackwell Publishing Ltd. 2003
Conclusion
According to Gilbert Fulmer, The idea that the past could be
changed by a time traveler is a confusion which stems from the
mistaken notion that time travel would cause a repetition of the
past.
15
But it is Fulmer who is mistaken. The issue is not whether
time travel would cause a repetition of the past, but rather could it.
If temporal moments have temporal parts, then time travel might
cause temporal moments to hypertemporally occur again and be
changed.
In fact, once one admits that there are coherent models in
which the past can change, many of the traditional problems and
paradoxes associated with time travel disappear. For example,
suppose the prevailing sentiment is correct Paul can travel back
in time, but he cannot kill his three-year-old self. But then,
according to Samuel Gorovitz, Pauls murder weapon is not
behaving as the normal physical object we take it to be or the
notion of voluntary action does not apply in the usual way.
16
To
what degree Gorovitzs concerns are warranted, if changing the
past is logically impossible, is the source of tremendous debate.
17
Regardless, Gorovitzs concerns are completely sidestepped once
one admits that Paul can change the past and kill his three-year-
old self.
Does the hypertemporal model of time travel thereby rescue
from incoherence all time travel stories that either take changing
the past for granted or make changing the past a fundamental
part of the storyline? Not necessarily. The hypertemporal model
has certain consequences that rule out many such stories, espe-
cially stories along the following lines. Paul and Sarah together go
back in time and unknowingly change the past. They jump
forward and find a drastically altered future. They then go back
into the past and fix the problem and finally jump back into the
TIME TRAVEL AND CHANGING THE PAST 31
Blackwell Publishing Ltd. 2003
15
Gilbert Fulmer, Time Travel, Determinism and Fatalism, Philosophical Speculations in
Science Fiction and Fantasy 1, (1981): p. 41.
16
Gorovitz, p. 367.
17
See, for example, Craig, pp. 140-141; Dwyer, p. 349; Paul Fitzgerald, Tachyons,
Backwards Causation, and Freedom, Boston Studies in the Philosophy of Science VIII, (1971):
pp. 425434; Paul Fitzgerald, On Retrocausality, Philosophia 4, (1974): pp. 534547;
Gilbert Fulmer, Understanding Time Travel, Southwestern Journal of Philosophy 11 (1980):
pp. 151153; Fulmer, Time Travel, Determinism, and Fatalism, pp. 4148; Horwich, pp.
433437; Lewis, pp. 7580; Paul Thom, Time-travel and Non-fatal Suicide, Philosophical
Studies 27 (1975): pp. 211216; and Kadri Vihvelin, What Time Travellers Cannot Do,
Philosophical Studies 81 (1996): pp. 315330.
original future. On the current hypertemporal model, however,
there is no jumping back into the original future it is, after all,
in the hypertemporal past. At most Paul and Sarah will jump into
a very similar future. The hypertemporal model allows changing
the past, but with the consequence that when one changes the
past one thereby sets in motion a new future.
Are there models that will allow one to get back to the original
future and still allow for changing the past? I do not know. Could
there be models of time travel different from my own that consis-
tently allow for changing the past? Probably. The argument that
changing the past is logically impossible rests on the empirical
assumption that temporal moments do not have parts. Any model
of time that discards this assumption might produce an adequate
example of the past being changed.
18
Do I think time is the way I
have described it, i.e. that temporal moments in fact have hyper-
temporal parts? I have no evidence that it is. Could temporal
moments have temporal parts? Could time be two-dimensional?
No contradiction results by assuming that it is. If temporal
moments have hypertemporal parts, then time travel might cause
temporal moments to hypertemporally occur again, in which case
the events of temporal moments changing over hypertime would
be no more contradictory than enduring objects changing over
normal time. Changing the past is logically possible.
University of Richmond
Richmond, VA 23173
USA
ggoddu@richmond.edu
32 G.C. GODDU
Blackwell Publishing Ltd. 2003
18
At least some such models fail to model changing the past. For example, see Meiland,
pp. 153173. On Meilands model what he is calling the past is not the states of the universe
causally responsible for the present state of the universe, and since the states causally
responsible for the present state of the universe never change, Meilands model does not
allow changing the past. See also, Objection III. For arguments against changing the past
in multi-dimensional relativistic models, see Alasdair M. Richmond, Plattners Arrow:
Science and Multi-dimensional Time, Ratio XIII (2000): pp. 256274.

Vous aimerez peut-être aussi