THE OMNIPOTENCE PARADOX AS A

PROBLEM OF INFINITE REGRESS

TZACHI ZAMIR
Department o[ Philosophy
Tel-Aviv Unwersity,
Tel-Aviv, Israel

PREAMBLE

In this paper, I argue that attempts to explain away the

Omnipotence Paradox by referring to the way in which its
standard presentation is temporally confused fail. I offer an
alternative formulation, which presents the paradox as a
problem of infinite regress. Unlike epistemic concepts, the
concept of ability becomes self-contradictory when maximal-
ly extended. Finally, I argue that this peculiarity plagues
another modal concept: that of possibility.
Can an omnipotent Being create a stone that It cannot lift?
A now popular reply to this medieval paradox is that, with
appropriate limitations, it can. I will argue that it cannot. My
main concern will be to show that the traditional presentation
of the paradox is wrong, and that it requires reformulation in
terms of a problem of infinite regress (I). I shall then address
possible objections to my suggestion (II). In the third and final
section, I shall argue for some general conceptual conclusions
that may be drawn from the omnipotence paradox.

Call a being 'omnipotent' if it can succeed in accomplishing

any logically consistent task. Since performing a task in a par-
ticular way can be itself formulated as a task, it seems fair to

Sophia Vo138 No 1 1999, March-April. 1

 require of omnipotent beings to be able to accomplish any
logically consistent task in any logically consistent way. By
'logically inconsistent' tasks, I mean tasks like drawing a
round square or naming the largest prime number. Most of
the philosophers who have discussed the paradox regard
these demands as wrong in one way or another and argue
that, therefore, no being - omnipotent or not - should be
required to meet them. Creating unliftable stones is a logical-
ly consistent task (it can, for example, be easily met by non-
omnipotent beings like us). A being that fails this test cannot
be omnipotent.
Under its standard formulation, the Omnipotence para-
dox supposedly yields two possibilities which lead to the same
result: either the Being cannot create the stone (thereby proving
itself not omnipotent), or the Being can create the stone (again
proving itself not omnipotent since a stone exists which it can-
not lift). Therefore, omnipotent beings do not exist.
A possible refutation of this is to claim that an omnipo-
tent Being can create such a stone and thereby cease to be
omnipotent. This is not paradoxical or strange because there
is no contradiction between a Being that is omnipotent on t
and not so on t'. 1 An omnipotent Being has the ability to vol-
untarily lose Its omnipotence, much as a pianist has the abil-
ity to voluntarily lose her playing ability by cutting off her
hands. Granted, an omnipotent Being cannot create such a
stone and remain omnipotent, but this does not affect the
Being's omnipotence any more than Its inability to comply
with any contradictory demand. Since the demand to create
such a stone was ipso facto a demand to give up omnipo-
tence, the Being cannot meaningfully be challenged both to
retain and give up omnipotence. The answer to the original
question would be positive: the Being can non-paradoxically
create such a stone and the paradox is explained away. 2
This answer is, on the face of it, plausible, and has been
used to show that according to some notions of omnipotence,

2
the alleged paradox is not a genuine paradox at aU.3 However,
I will attempt to show that it entails an infinite regress, which
proves that the Being cannot create the stone. My proposed
reformulation would, thereby, retain the idea that omnipo-
tence involves paradox. My argument is as follows:
Granted that it is not contradictory to think that an
omnipotent Being can lose one of Its abilities, thereby becom-
ing non-omnipotent (through, for example, creating the stone
or losing Its ability to lift existing stones), what would stop It
from regaining Its lost ability? In opposition to the pianist
mentioned earlier, who unfortunately cannot bring back her
lost playing ability, an omnipotent Being has, by definition,
all abilities. These include no doubt, the ability to cancel lim-
itations that It had previously set upon Itself. It should be
obvious that if the Being retains an ability to restore Its abil-
ity to lift the stone, then It can in fact lift the stone. This
means that there will exist no point in time in which it could
be truly asserted of the Being that it cannot lift the stone. The
Being, therefore, fails to comply with the original challenge to
create a stone that It cannot lift. +
So the refuter of the paradox has to show that the Being
can create the stone, while not being able to regain the abili-
ties It necessarily loses by performing that action, s This seems
to be simple: If the Being really intends to stop being omnipo-
tent, It also has to eliminate Its ability to restore Its lost abil-
ities. However, this would mean that when the Being decides
to stop being omnipotent, It has to lose not one ability, but
two. This involves two distinct actions. 6 The Being has to:
(1) Stop being omnipotent (by, for example, creating the stone).
And:
(2) Eliminate Its ability to restore the abilities It loses by per-
forming)
But surely, an omnipotent Being which has lost only two
of Its abilities, namely, the ability to lift up a certain stone and

3
the ability to regain that ability, still retains all the rest of Its
abilities. Amongst these, no doubt, is the ability to restore the
second ability It has lost. However, one might say, if the Being
has really eliminated Its ability to restore the ability It has lost
in (1), then It has also eliminated Its ability to restore that
ability as well. From this it would, of course, follow that we
were wrong to assume that the Being only performed two
actions when It created the stone. It necessarily had to per-
form a third action. The Being also had to:
Eliminate Its ability to restore the second ability It had
lost by performing (2).
One can easily see how this process goes on, yielding the
unfortunate result that the Being has to perform an endless
number of actions so that It can eventually become non-
omnipotent. But if they are indeed endless, then the Being
cannot finish doing them and get on with being non-omnipo-
tent. This means that It cannot create the stone and that It
was therefore not omnipotent from the start. The paradox is
thereby reinstated.
However, what if the Being could eliminate any ability to
restore eliminated abilities? Alternatively, eliminate Its ability
to revise any of Its actions? This ability itself could not be
later regained since any restoring of lost abilities was elimi-
nated by it. This way, the Being need not perform an endless
number of actions, but only two. Namely:
(a) Stop being omnipotent (through losing an ability to lift a
certain stone).
(b) Eliminate Its ability to restore any ability It loses.
This would bypass the infinite regress argument but
would unfortunately fall short of answering the paradox,
which would be reformulated thus: 'Can an omnipotent
Being create a stone It cannot lift, and not eliminate Its abili-
ty to restore any ability It loses?' There is no reason to regard
this demand as self-contradictory or empty (non-omnipotent

4
beings can easily comply with it). It does not challenge the
Being in the same illegitimate ways that the vacuous demands
mentioned above do. 7
A negative answer to this question would expose the
Being to be non-omnipotent from the start (since a task exists
that It cannot perform). A positive answer, on the other hand,
would simply succeed in relocating the paradox. The reason
for this is that a positive reply means that the Being can some-
how eliminate the set of all abilities connected with the lifting
of the stone or with restoring such abilities if they are elimi-
nated (call this set S). It also maintains some of Its other abil-
ities to restore eliminated abilities. Therefore, according to
this hypothesis, in order to create an unliftable stone, the
Being supposedly performs two actions:
(I) Creates a stone that It cannot lift.
(n) Eliminates S.
However, if S is composed solely of the infinite (1), (2),
(3) type abilities, then eliminating S is not enough for the
elimination of the stone-lifting ability. The reason is that in
such a case the Being retains Its ability to restore sets of abil-
ities that It eliminates. Therefore, eliminating the stone-lifting
ability via the elimination of S necessitates that S should
include as members of it the elimination of an infinite num-
ber of abilities involved by restoring the S set itself once it is
eliminated. This involves eliminating the ability to restore S,
eliminating the ability to restore the eliminated ability to
restore S, and so on. S becomes incredibly big. This makes it
possible to simply reformulate the paradoxical question. The
demand of the omnipotent Being would be to be able (like
non-omnipotent beings) to create the stone and not lose the
ridiculously large number of eliminated abilities included in S.
A negative answer now appears mandatory. The Being is
exposed as non-omnipotent from the start thereby reinstating
the paradox.

5
II
I have argued that attempting to explain away the paradox by
assuming that an omnipotent Being can enter a state in which
it is not omnipotent cannot work since it entails an infinite
regress that stems from the omnipotent Being's unique nature.
The following general objections could be raised against my
argument:
(I) One could argue that God is not only omnipotent but
also necessarily successful, therefore He can create such a
stone.
(II) One could argue that the Being does not in fact have to
perform an infinite number of tasks.
(RI) One could argue that performing an infinite number of
tasks is possible.
(IV) One could accept the validity of the paradox but deny
that its conclusion is necessarily that the Being does not
exist.
Variations of these objections will be addressed below.
One could object to the above formulation of the para-
dox through claiming that an omnipotent God is also always
successful. Given such a necessary efficacy of God, then if He
wishes at t to lose His ability to lift the stone at t', then - con-
tra the infinite regress argument - He necessarily succeeds.
Such an objection, however, begs the question since the para-
dox investigates whether or not the Being can in fact succeed
in performing certain tasks. The paradox challenges the suf-
ficiency of omnipotence to necessary success, hence necessary
success cannot simply be assumed.
A more subtle variation on such an objection is to claim
that God is both omnipotent and necessarily successful (with-
out assuming that the former implies the latter). Presented
with such a concept of God, the infinite regress argument will
fail, since (again) if the divine will wishes to eliminate any of
God's abilities, It necessarily succeeds. Such an argument
would succeed in saving God's omnipotence at the price of
introducing a notion that is no less vulnerable to paradoxical
constructions. For consider the following: 'Can a necessarily
successful Being (a Being that necessarily succeeds in per-
forming any logically consistent task) succeed in devising
tasks that It can never successfully perform?' A negative
answer would entail that the Being is not necessarily success-
ful. A positive reply would entail that such a necessarily suc-
cessful Being - that is, a being which is successful at all pos-
sible worlds - is necessarily unsuccessful at some possible
worlds (hence not necessarily successful from the start).
A second possible objection can be based on the claim
that the infinite regress argument involves an inadequate
description of the Being's actions when It becomes non-
omnipotent. Eliminating or limiting an ability does not nec-
essarily involve performing an endless number of actions. As
the case of the pianist above shows, non-omnipotent beings
can in fact eliminate or limit some of their abilities without
being involved in an infinite regress. Why should there be a
specific problem for omnipotent Beings?
This objection rests on falsely carrying over the meaning
of 'eliminating (or limiting) an ability' from non-omnipotent
contexts to omnipotent ones. Strictly speaking, in order to
ascribe to any being - omnipotent or not - a loss of an abili-
ty, there must be no existing defeaters to that claim. For a
non-omnipotent being, this need not be problematic given
knowledge of possible defeaters, and the fact that there are
not, usually, an infinite number of them. Losing the ability to
walk because of some neck injuries, for example, has - in the
present state of medical knowledge - no existing defeaters. 8
However, things are different for a Being that has all abilities.
Since the ability to restore an eliminated ability is not only
itself an ability which the Being must, by definition, possess,
but also an actually existing defeater, the omnipotent Being
has to somehow eliminate it. The infinite regress springs from
the fact that this elimination, instead of settling the matter, is
exactly what constitutes the coming into being of a new
defeater.
Another objection that argues for the same conclusion is
to accept the adequacy of the description as a possible one,
but to deny that it is actually true. When one looks at a
coloured surface, for example, this action can be described as
involving looking at an infinite number of colored points.
However, from the fact that it is possible to describe the
action as such, it in no way follows that the action in fact
requires actually performing an infinite number of actions.
Therefore, although the Being's action can be described as
involving an infinite number of separate actions, it does not
necessarily follow that It has to actually perform them.
In answering this, one may claim that there is a false
analogy between the case of looking at a colored surface, and
creating the stone. There is no prima facie reason to think
that it is impossible to look at an infinite number of points (if
they are in fact infiniteg), although it is indeed impossible to
look at each one individually. However, there is no reason to
think that one needs to look at all of its individual points in
order to look at a coloured surface. In the case of the infinite
regress above, on the other hand, the Being starts by per-
forming one action - creating the stone - then another - elim-
inating Its ability to restore the lost ability to lift the stone -
and so on ad infinitum. The problem is not a temporal one.
It is, rather, that each successive action creates a new defeater,
making a 'final' action impossible. This is also the reason why
allowing the Being to start with other actions than the cre-
ation of the stone does not circumvent the difficulty.
A third possible critique is to argue that there is no rea-
son to suppose that performing an infinite number of tasks is
impossible. J.E Thomson gave strong arguments against such
a claim, l~ However, even if one rejected Thomson's reasons,
an ability to perform an infinite number of tasks would not
save the Being's omnipotence. The paradoxical question
would then be reformulated in terms we have already met
earlier: a demand to create the stone without performing an
infinite number of tasks. Non-omnipotent beings can easily
meet such a demand, omnipotent Beings - as it turned out -
could not.
Finally, one could claim that all that the argument shows
is that an omnipotent Being cannot lose Its omnipotence. This
would mean that if a Being is omnipotent, It is necessarily
omnipotent. If one regards 'omnipotence' as meaning 'neces-
sary omnipotence' from the start, then the infinite regress
argument is irrelevant. An omnipotent Being cannot lose Its
omnipotence, and therefore It cannot create the stone in ques-
tion. This, though, is not a limitation on Its abilities but a
consequence of Its special nature. All that the so called 'para-
dox' proves is that if a necessarily omnipotent Being exists,
then the existence of such a stone is logically impossible. 11
Such a move seems attractive to some, since it ultimate-
ly accepts the validity of the paradox, but instead of drawing
conclusions concerning the abilities of the Being; it makes
ontological limitations on the world in which the Being exists.
This hardly makes it a persuasive move, however, since the
inability to break free of these ontological limitations is pre-
cisely what revealed the Being to be non-omnipotent from the
start, and thereby exposed the concept of omnipotence as
self-contradictory. It is indeed true that if such a Being exists,
then the existence of such a stone is logically impossible. But
it is the exactly the inability to create such a stone (an inabil-
ity that - to the omnipotent Being's disgrace - non-omnipo-
tent beings do not share) that makes the existence of such a
Being impossible.
A close variation of this objection is to Claim that from
the beginning the paradox is not genuine since there can exist
no object X that can satisfy the description 'a stone that can-
not be lifted by an omnipotent Being'. This boils down to the
same conclusion as the one reached by the previous 'necessary
omnipotence' objection (though this time without the condi-
tional clause): the existence of the sort of stone the paradox
postulates is logically impossible. The 'paradox' is no chal-
lenge to the Being since it does not meet the initial requirement
of demanding the Being to perform a logically consistent task.
For someone who wishes to defend the traditional for-
mulation of the stone paradox this last objection might prove
insurmountable. But since my purpose here is not to defend
the traditional formulation - a formulation that could be eas-
ily dismissed by the sort of rejection of the paradox with
which we begun - but to suggest a close reformulation of it
such an objection can be circumvented. My suggested formu-
lation of the paradox in terms of a problem of infinite regress
enables seeing that the omnipotence paradox does not, in
fact, depend on the idea that a previously non-existent stone
has to be created. The paradox can be formulated as a
demand to eliminate an ability to lift an already existent stone
(if the Being cannot eliminate such ability, It is not omnipo-
tent. To mistakenly suppose that It can, is to over-look the
infinite-regress which, as we have seen, such elimination
involves). It would hardly make much sense to say that the
existence of such stones is 'logically impossible'.
If the Being cannot lose any of Its abilities (that is, it is
'necessarily omnipotent') then - as was traditionally conclud-
ed from the paradox - it would turn out that such a being
lacks many abilities that non-omnipotent beings possess. In
fact, if being able to call aloud the name of a number is admit-
ted as an ability that can be eliminated, non-omnipotent
beings are able to eliminate an infinite number of abilities by
losing the power of speech. That is, they possess an infinite
number of abilities that a being, which is supposed to have all
abilities, does not. Such an objection would therefore entail
defending a rather strange concept of omnipotence that is
hardly worthy of its name.

10
Ill
What general conceptual conclusions can be drawn from
this? If the Omnipotence Puzzle is to be regarded as more
than an intellectual exercise, it should surely reveal something
about the concept of ability. The infinite regress argument, if
it is correct, proves that, unlike non-omnipotent beings, an
omnipotent Being cannot lose any of Its abilities. 12 The inter-
esting point that emerges concerns a peculiarity of the con-
cept of ability when maximally extended to include all abili-
ties. This peculiarity can be shown through contrasting abili-
ties with knowledge. If A knows some tunes and B knows all
tunes, than B knows all the tunes known to A and of course
many more. This asymmetry remains if the notion of knowl-
edge is maximally extended to include everything knowable.
Limited cases of ability are also asymmetrical: If A can lift
some stones and B can lift all stones, than B can lift any stone
A can. However, unlike knowledge, in the case of ability the
asymmetry is broken when maximally extended: if A has
s o m e abilities and B has all abilities, it paradoxically turns
out that B cannot have all of A's abilities. The omnipotence
puzzle shows that, unlike the epistemic concepts of knowl-
edge or belief, the modal concept of ability collapses when
maximally extended.
Because some uses of the notion of possibility are syn-
onymous with ability, it is hardly surprising that similar infi-
nite regress problems can be constructed pertaining also to
this modal concept: can a being which can realize any possi-
bility realize a state of affairs (S) in which it can permanently
realize fewer possibilities? An argument analogous to the one
above would lead to a negative answer here too. If realizing S
requires permanently losing the possibility to realize A (call
this possibility P), this can only be truly achieved through los-
ing the possibility (P') of regaining P as a realizable possibili-
ty after it has been lost, and so on. Here too, the infinite num-
ber of actions that the Being would be required to perform

11
would render the move to S impossible. As in the case of abil-
ities, a Being for which anything is possible paradoxically
cannot realize possibilities that are easily accessible to a less
fortunate being (e.g. our former pianist which had easily
enough entered such an S state).
Some notions of t3ossibility, however, do not require any-
one to do anything. An obligation to perform an infinite num-
ber of actions would, therefore, not seem to be problematic for
all cases of possibility. Unfortunately, this does not save the
concept of possibility from contradicting itself when maxi-
mally extended. This can be shown through the following:
Suppose that anything is possible, that is, all conceivable
(logically consistent) possible worlds exist. Call possible
worlds that have access only to some - but not to all - other
possible worlds 'third' worlds. For third possible worlds, not
everything is possible. Call possible worlds to which all other
possible worlds are accessible 'rich' worlds. If all possible
worlds exist, then both third and rich possible worlds exist.
So far, none of this is problematic. But if all logically consis-
tent possible worlds exist, then there is no way through which
to exclude the existence of a possible world - call it a 'bene-
factor' world - which in some strange, affirmative action,
way allows itself to be accessible only to third worlds. Given
the previous assumptions, the existence of benefactor worlds
and rich worlds is mutually exclusive. Therefore, if all con-
ceivable possible worlds exist, it cannot be the case that all
possible worlds exist.
Unlike other concepts, something strange happens to
modal concepts when linked to the notion of 'all'. 13

12
ENDNOTES
1 This observation was made by Igal Kvart: 1982, 'The
Omnipotence Puzzle', Logic and Analysis, and David E
Shrader: 1979, 'A Solution to the Stone Paradox', Syntbese.
2 Some theorists demand that the Being be 'necessarily omnipo-
tent', meaning that It cannot lose Its omnipotence. I will dis-
cuss such a possibility below. A different objection which I
have sometimes heard theorists use (though never noted in the
literature) is that the Being can create the stone and remain
omnipotent through Its ability to destroy the unlikable stone.
The elimination of the stone is tantamount to restoring
omnipotence and so, as long as the Being retains an ability to
destroy the stone It never loses Its omnipotence (that is, when-
ever the Being's omnipotence would be questioned, there
would exist no stone that It could not lift). There are several
problems with this suggestion, the most important of which is
that, at best, it would simply require changing the paradoxical
demand from creating an unlikable stone to creating an inde-
structible one.
3 Kvart (Ibid.) used this to reject the notion of what he calls 'a
merely omnipotent' being. Such a being can non-paradoxical-
ly lose Its omnipotence.
4 One could obviously claim that the Being still succeeds in
entering a state in which It temporarily cannot lift the stone
('temporarily' referring here to the point in time between cre-
ating the stone and restoring the ability to lift it).
Circumventing this objection simply requires clearing a possi-
ble ambivalence in the formulation of the paradoxical task.
The Being is required to create a stone that It can never lift.
5 I am assuming here that creating the stone involves losing an
ability. This assumption is not crucial to the argument and, if
challenged, would merely change the original question from
which the paradox begins. For example: Can an Omnipotent
being lose Its ability to lift stones? A negative answer would
show It to be non-omnipotent; a positive one immediately has
to face the infinite regress argument that follows above.
6 I shall assume that such actions can be individuated and
counted.
7 The reader might wonder why a similar reformulation, relat-

13
 ing abilities (1) and (2) into the paradoxical question, was not
suggested instead of the infinite regress argument which was
introduced at that stage. The answer is that in opposition to
the question above, such a formulation could have been reject-
ed as a meaningless demand, since in that case it does seem
contradictory to challenge the Being to create the unliftable
stone while retaining an ability to lift it.
8 In fact, one should draw a distinction here between actual and
possible existing defeaters. A possible medical procedure that
would bring back the ability to walk can be envisaged, and
therefore is a possible defeater. This distinction is, however,
irrelevant to the above argument.
9 Strictly speaking, they are not infinite. The coloured surface
actually contains a finite number of points of any specific size.
An unlimited ability to divide spatial points further does not
entail that there are an infinite number of them. This fallacy
constitutes one of Zeno's paradoxes of motion.
10 J.E Thomson: 1954, ' Tasks and Super-Tasks', Analysis, (15)
pp. 1-13.
11 Kvart (Ibid.) and others make this point.
12 Though It can, non-problematically lose all of them through
suicide.
13 This paper owes much to helpful and insightful comments by
Ruth Manor, Anat Matar, Ruth Weintraub, and an anony-
mous reader.

I4