. 8
( 10)


4 Damasio (1994).
5 Fredrickson and Kahneman (1993); Kahneman et al. (1993). We will soon see reason to
question the probity of the aspect of these intuitions that involve a differential treatment of
ends and beginnings.

people even coolly and re¬‚ectively discount future suffering in favor of
present gain.
The third point is that normative intuitions, and for that matter features
of our actual practice, that do favor or re¬‚ect prudence are schooled and
hence perhaps corrupted. They do not represent common sense in its
pristine state. At least since Plato, philosophers have been engaged in an
attempt to sell to recalcitrant and ¬‚ighty common sense the thought that
we should act prudently to maximize our happiness over our lives, and
then should commit to moral treatment of others on the basis of that
reformation of our naturally imprudent selves. Attempts to bribe us to
morality by promises of salvation have reinforced this. Intuitions favoring
traditional prudence certainly aren™t natural and inevitable, but are the
result of a long tradition of teaching and training which may of course be
The fourth point is that we should be wary of a tendency to discount
the testimony of common sense in bad circumstances over its testimony in
good circumstances, as we need to be wary of discounting the testimony
of those who are badly off in general. Remember a case from Chapter 4.
As Ahab has his gangrenous leg sawed off by the ship™s surgeon, without
anesthetic, he may cry out, “This is so painful that I™d rather die.” As
surgeons, you and I would likely ignore such cries as best we could,
expecting that Ahab would thank us in the morning. And if he did,
we might think that proved we did the right thing. But would it? Any
judgment that, for instance, his later pleasure in life is or is not worth the
pain of the surgery is essentially comparative, and neither Ahab during the
operation nor Ahab on the morning after concurrently experiences both
of the values that are being compared. It really isn™t clear that Ahab on the
morning after has greater comparative authority. For one thing, we are
happily adept at forgetting the intensity of past pains. Maybe only Ahab
as he is experiencing the pain really knows what it is worth.6 His cries
are, after all, commonsense testimony, and maybe the only such testimony
that is suitably and fully confronted with the single most relevant fact.
All this is not meant to suggest that traditional prudence does not
enjoy some apparent intuitive advantages over an egalitarianism of periods.
But the ¬fth point is that some of these advantages do not lie at a helpful
spot for opponents of that egalitarianism. Individual lives, as opposed to
periods of lives, matter most to our normative intuitions when we consider

6 I owe this case to a public lecture by Thomas Schelling.

goods that are thought to be rewards for desert. And this isn™t the kind
of distribution of well-being or good with which we are immediately
concerned here, the kind of distribution that we dominantly consider
when we make judgments about how well lives go.
Still, there is some distance between real tendencies in our current
normative intuition and an egalitarianism of periods. That is why my
main argumentative strategy will be to put commonsense intuition under
the pressure of confrontation with certain unacknowledged possibilities,
and to argue that the properly re¬‚ective and revised common sense that
results would support my view.
But there are two more pieces of background that are necessary before
we can traverse that argument. The ¬rst can be quickly sketched. Our
central argument will incorporate a foil, the Traditional View. That is one
obvious expression of the traditional conception of prudence. It is the view
that one™s rational individual self-interest is served by the maximization of
one™s good or well-being over one™s life, where that good is the additive
sum of one™s good during each period of one™s life, irrespective of the
temporal order of those periods or the distribution of that good among
Our argument will evaluate the Traditional View relative to more plau-
sible competitors by considering certain cases. Four assumptions must
govern our treatment of these cases. These four assumptions are the last
piece of necessary background.
First assumption: To engage the Traditional View and some alternative
views, we will need sometimes to assume, for argument™s sake, that we can
speak meaningfully of the sum of the well-being or good in a life. This
assumption is intuitively questionable, and indeed violates the ordinality
of basic value. But it will allow proper interaction of our discussion with
relevant extant literatures, and charitably allow for the coherence of some
of the competitor positions that are in play.
Second assumption: To achieve proper understanding of the key issues
involved here, we will need to presume that we can somehow control
for the uncertainty of the future, and for secondary effects occasioned by
the pains and pleasures of anticipation and memory. With the literature,
we will presume that we can talk about certain choices between arbi-
trary patterns of the distribution of well-being or good among relevant

7 Whether one™s well-being can be affected by things that happen after death will not be
directly in play in our argument, but one might ascribe any such goods to the last period of
life for bookkeeping purposes.

periods of a life. If the effects of anticipation and memory matter, they
are somehow subsumed into that well-being. We will also presume with
the literature that these very hypothetical features of our cases will not
disable our intuition.
The third assumption is a related though somewhat subtler point, and
will itself require some background. The issue ultimately in question
for us is the traditional normative issue of how alternatives are to be
properly and rationally assessed, in particular when the outcomes involve
the distribution of well-being within lives. That means that what is in
question here is the rationally proper way to distribute well-being or
good in order to maximize the quality of a life.
But talk about the proper way to distribute well-being and to assess
the weight of that distribution within a life may invite objections. Some-
one may worry that there are all sorts of preferences about the shape of
one™s life, about how well-being is best distributed in that life, that it isn™t
irrational to have. None of these preferences may seem rationally priv-
ileged over the others. For any given person, the distribution of good
across their life is better, it may seem, if it more closely mirrors their
preferences, whatever they are. Or, perhaps it may seem, while certain
preferences for the shape of one™s life seem irrational and hence should
properly be ignored, not many are.8
The correct response to this objection involves two key aspects. The
¬rst aspect is precision about the questions at issue here. Our ultimate goal
is to determine the proper and rational way in which the distribution of
well-being determines the overall value of outcomes or single lives. But
note that “rational” in one sense of the word, the sense mostly invoked by
the objection, is weaker than “rational” in the sense invoked by our goal.
“Rational” in one sense means simply not wildly irrational. If preferences
are not wildly irrational, they may still be mistaken in the sense at issue
here. And of course we are talking not about the distribution of ordinary
goods like money and health but rather about the temporal distribution
of well-being or basic normative good in lives.
The second aspect of the response is that the objection invokes prefer-
ences, but the relationship between preference satisfaction and the well-
being and good that are our immediate concern is complex in relevant
ways. One standard view of well-being identi¬es it with desire satisfaction.
But after Chapter 4, we know that this is problematic. And in any case, it

8 Thanks to Mark van Roojen for this objection.

is widely recognized that only the satisfaction of informed and corrected
desires can plausibly be identi¬ed with well-being. Perhaps the success of
the Traditional View suggests that some actual preferences for the shape
of a life do seem contrary to reason. And there is a special dif¬culty that
would face any attempt to root judgments about the overall well-being of
a life in individual preferences about the temporal shape of a life. Accord-
ing to the most straightforward desire-satisfaction accounts of individual
well-being, a preference for the shape of a life would be a higher-order
preference about the temporal order of the relative satisfaction of other,
lower-order preferences. We are hence faced with the dif¬culty of balanc-
ing off the satisfaction of the higher-order and lower-order preferences.
And even if we grant the higher-order preference complete dominance
in determining well-being, or specify that lower-order well-being is not
de¬ned by reference to preference satisfaction, still it is important to see
that not everyone has such a higher-order preference for the shape of their
life, and that even when they do, it will very likely change over time. It will
very likely do that even if it is corrected. So we cannot properly assess the
overall well-being of many lives, on the way to making an overall norma-
tive assessment of outcomes, by a simple appeal to individual preferences
about the shape of one™s life.
More complex desire-based accounts are, of course, possible. For
instance, we might consider each of the hypothetical fully informed pref-
erences for distribution of goods within a life that an individual would
have at each moment of that life if asked speci¬cally about such distri-
butions, and then maximize the sum of the satisfaction of those higher-
order hypothetical momentary preferences.9 But such preferences are too
abstract and hypothetical, and their summation too unfamiliar, to give
such a controversial normative alternative an obvious intuitive advantage
over alternative views on the simple ground that we need to respect peo-
ple™s preferences. Such preferences are too far removed from actual people™s
actual preferences.
All this is not to suggest that individual preferences for the shape of a
life should not matter at all in the evaluation of outcomes. And here we
come back to the necessary third assumption of our central argument:
Just as we will presume that we can control for the effects of anticipa-
tion and memory by incorporating them into the well-being or good dis-
tributed within the life, so also the satisfaction or frustration of individual

9 For something like this proposal, see Carson (2000: 86“87).

preferences for the shape of a life should play exactly the same role in
determining the well-being or good of a period that any satis¬ed or frus-
trated preference of that period appropriately plays. If a life with a certain
shape is for someone the most important thing at any particular time,
then that will show in their relevant frustration or satisfaction at that time
and hence their good at that time, at least, that is, if the frustration or
satisfaction of preferences is crucial to their well-being or good. And for
those for whom such a shape isn™t that important, frustration or satisfac-
tion of such abstract temporal preferences won™t signi¬cantly affect their
well-being or good. That is proper respect for actual individual differences
in preference.
We are of course ultimately concerned here with distribution within
a life not of normative value tout court, but merely of basic value that is
in fact intrinsic hedonic value. But since frustration or satisfaction often
affects hedonic tone, my fourth and ¬nal presumption will be that this
detail will not derail our argument. In any case, our present concern is
not with hedonism, but merely with a certain abstract structural feature
of HMP, independent of the speci¬c conception of basic good that HMP
also incorporates.


With the necessary preliminaries out of the way, we can now traverse
our argument. It has three steps. The ¬rst step is the recognition that
the Traditional View, concerned to maximize individual well-being irre-
spective of its temporal distribution, is intuitively wrong. Indeed, it is so
obviously wrong that we can wonder if anyone ever really held it except
on paper. Intuitively, the distribution of good within a life matters some
to the overall quality of that life.
The most obvious cases that reveal this involve equalized sums of well-
being in two possible lives of equal duration, and reveal that distribution
matters, at least once the sum of well-being is ¬xed. For instance, it
intuitively matters if two lives with equal sums of well-being involve
respectively compensating moments of agony and bliss or simply an even
keel. It matters if all the joy of a life is in youth or in some limited period
of age, or spread about more evenly through the life.10

10 And even if it is not coherent to talk about equalized sums of well-being in two distinct
lives, the temporal patterns noted remain intuitively salient. And of course if summation
doesn™t even make coherent sense, then a key assumption of the Traditional View is false

We need to exercise due care. There are many related but distinct issues
in play in such examples. For instance, cases of this sort may uncover indi-
vidual preferences for one™s own life, which may differ between people
or even for a single person over time. But we need the cases to reveal
intuitions about normative facts that support some individual preferences
rather than others. Still, many do have the relevant intuitions. The dom-
inant pattern of intuition seems to be that at least some concern for each
reasonably long period of one™s life is rationally required, that there is
at least a ¬‚oor beneath which, other things equal, each period shouldn™t
go, that there is at least that very weak egalitarianism of long periods
present in commonsense intuition and uncovered by these cases. But for
the moment, let™s rest with the weaker and less controversial claim that
distribution intuitively matters some.
More interesting cases involve trade-offs between distribution and
maximization. There are apparently well-founded intuitions that imply
that distribution should matter even at some cost to the total sum of
well-being in a life. John Mackie has uncovered some:

Suppose that as a young adult one could look forward to a fairly long life, and
had some rough idea of the various satisfactions and frustrations likely to be
experienced as a result of each of several alternative choices of a plan of life.
Suppose that one could also allow for probable changes in one™s preferences and
ideals as one grew older, and was able to detach oneself from one™s present youthful
purposes and values and to look fairly at the alternative plans of life from the points
of view of all one™s future selves as well as the present one. Is it obvious that if
a reasonable person were able to do all these things, he would opt for whatever
plan of life promised the greatest aggregate utility? Or might he try to ensure that
no substantial phase of his life was too miserable, even if very great satisfactions
at other times were to compensate for this? . . . I am suggesting that looking after
each substantial phase might be the sensible thing to opt for in its own right, not
merely as a means to maximizing the aggregate.11

Distribution of good within a life intuitively matters even in this somewhat
stronger way, whether we focus on our own lives or on the lives of others.
And notice that it does so in a way that supports or re¬‚ects at least a
weak egalitarianism of substantial periods of a life. Notice this intuitive
rationale hovering within Mackie™s case. It seems that each substantial
period of a life has a kind of equal normative standing of the sort that

11 Mackie (1984: 93).

concerns us. But I will not rest heavily on this point. We haven™t reached
our conclusion yet.
The second step in our main argument is to notice that there are a
number of different ways in which temporal distribution might matter,
or, perhaps it is more accurate to say, a number of different explanations
of how and why it does matter. So perhaps the appearance of an intuitive
egalitarian rationale that Mackie provides is misleading. There are three
discernible classes of possible views about the proper distribution of well-
being through life.
The ¬rst class, which I will call Timeless views, have perhaps the
greatest abstract similarity to the Traditional View, and re¬‚ect the rationale
that Mackie invokes. On a Timeless view, the contribution of the well-
being within a period of life to the goodness of the life doesn™t depend
on when in life that particular period occurs.
Egalitarian treatments of periods are such views. They endorse the
thought that each period has a moral status, independent of its temporal
place in a life, comparable to the moral status that individual people are
customarily granted in familiar normative theories, whatever their spatial
position in the world.
But notice that if a period matters just because it is a period of a life,
we may be unable to resist the claim that any noticeable period matters.
There is no obvious reason why a period of one length would be salient
but a shorter period would not. Years or days, youth or age, seem quite
arbitrary periods.
This is not to say that we do not treat different periods differently.
Someone might be willing to have a horrible two weeks in order to gain
a later but somewhat greater bene¬t, but not a horrible ¬ve years in order
to gain a correspondingly larger bene¬t. But there may be no coherent
rationale for such differences. And we can closely approximate a ¬ve-
year horror by a succession of two-week horrors separated by days off.
So if periods matter, we should be prepared for the real possibility that
all periods will matter. An egalitarianism of periods slides quite naturally
toward the distributional signi¬cance of relatively short periods, against
Mackie™s perhaps more intuitive claims about substantial periods.
For this reason, as well as for reasons of familiarity from the literature, it
is natural to consider ¬rst two alternative classes of distributional concep-
tions. Both essentially involve time, but in different ways. They involve
what I will call either objective or perspectival temporal patterning.
Temporal patterning in general occurs when the proper valuation
of a life depends on the temporal pattern “ for instance, the temporal

order “ of the well-being in that life. Objective temporal patterning,
which the views that I will call Objective hold salient, is perhaps the most
obvious and familiar sort. Objective temporal patterns in a life don™t vary
as now varies. Objectively, 2005 precedes 2006. But it won™t always be
now 2005, nor will 2006 always be in the future. Objective temporal pat-
terns are patterns that a life can maintain throughout and beyond its own
history and that don™t themselves change when now moves on. Perhaps
the best way to understand this is to consider notable examples.
Michael Slote and David Velleman are proponents of Objective
views.12 Both suggest that a life that gets better is, other things equal,
better than a life that gets worse.13 An alternative Objective view might
be that what happens in the prime of a life and not in the last days
of decrepitude is especially relevant to its evaluation. Ed Diener reports
a paradoxical “James Dean Effect”, in which people rate a wonderful
life that ends abruptly as better than one with additional mildly pleasant
years;14 this also is an Objective view. After a life is over, it either will
or will not have these temporal patterns, from the point of view of the
universe. Before or during a life, as long as we assume that there are deter-
minate facts about the future, it will be such that it is either going to have
these patterns or not. The relevant patterns are in that sense objective;
they do not change with now.
Some interesting dif¬culties attend Objective views. First, while objec-
tive patterns have some intuitive salience, still, as previously noted, dif-
ferent people have different preferences for the temporal pattern of their
lives, and they may well not have stable preferences for that pattern over
time. And of course there are familiar reasons to be wary of accounts of
individual good that rest on uncorrected and unre¬ned individual pref-
erences, anyway. The James Dean Effect may be another reason of just
this sort. So some art would be required to deliver a plausible Objective
account that at all closely respected individual preferences, even grant-
ing that we must discount some individual preferences as irrational. A
nonegalitarian Objective view would more seriously con¬‚ict with many
individual preferences than would a Timeless egalitarianism. A second
dif¬culty is the plausible suggestion of Derek Par¬t and Frances Kamm

12 Slote (1982); Velleman (1991).
13 Slote (1982: 23“24); Velleman (1991: 50). But notice that Velleman only claims that this
would seem so to most people, and also quali¬es the claim in certain ways by an insistence
on the importance of one™s narrative of one™s life. He also argues that it is wrong to conceive
of the value of a life as some sort of summation of the value of its periods.
14 Diener, Wirtz, and Oishi (2001).

that temporal patterning of this sort may be more intuitively salient for
some goods than for others “ for instance, more salient for achievements
than for pleasures.15
But, there are other ways in which objective temporal patterning seems
intuitively to matter to morality, and this may seem to lend indirect support
to Objective views. Kamm, in her treatment of mortality, has revealed a
number of these.16 Hedonistic value theories traditionally follow Lucretius
into the claim that death is no special harm that having been born is not.
But Kamm more closely tracks our commonsense intuitions. She pre-
sumes that death is a special evil, asymmetrical with having been born,
and asks why. Certainly death deprives us of familiar goods like more life,
happiness, pleasure, and desire satisfaction. But so does not having existed
prior to the moment of our birth, since if our lives had stretched in¬nitely
into the past we would have had in¬nitely greater opportunities for hap-
piness, pleasure, and desire satisfaction. The most natural explanations
of this asymmetry in our attitudes toward birth and death, if we accept
these attitudes as normatively probative at even the deepest level, involve
objective temporal patterning. Kamm suggests as relevant in particular
an Insult Factor, whereby death involves a loss of goods to a person who
already exists, and an Extinction Factor, whereby death ends permanently
all signi¬cant portions of a life. Objective temporal patterning might also
explain why we think that the satisfaction of desires after death may at
least reasonably be supposed relevant to well-being in a way that their
unknown satisfaction prior to one™s existence is not. So objective tempo-
ral patterning is not an isolated phenomenon in commonsense morality.
The third class of views about the normative salience of the distribution
of good or well-being within a life also involves temporal patterning, but
of a different sort. These views, Perspectival views, hold that perspectival
temporal patterning “ that is, temporal patterning dependent on when
now is and hence that changes with time “ is relevant. Par¬t has developed
some classic cases, cases that may also undercut the traditional conception
of prudence, that serve to reveal the most intuitive forms of Perspectival

I am in some hospital, to have some kind of surgery[,] . . . completely safe, and
always successful. . . . The surgery may be brief, or it may instead take a long time.
Because I have to co-operate with the surgeon, I cannot have anaesthetics. I have

15 Kamm (1993: 36“37).
16 Ibid., 13“71.

had this surgery once before, and I can remember how painful it is. Under a
new policy, because the operation is so painful, patients are now afterwards made
to forget it. . . . I have just woken up. I cannot remember going to sleep. I ask
my nurse if it has been decided when my operation is to be, and how long it
must take. She says that she knows the facts about both me and another patient,
but that she cannot remember which facts apply to whom. She can tell me only
that the following is true. I may be the patient who had his operation yesterday.
In that case, my operation was the longest ever performed, lasting ten hours. I
may instead be the patient who is to have a short operation later today. It is either
true that I did suffer for ten hours, or true that I shall suffer for one hour.
I ask the nurse to ¬nd out which is true. While she is away, it is clear to me
which I prefer to be true. If I learn that the ¬rst is true, I shall be greatly relieved.17

There are complications that attend this case. For instance, some people
follow Locke into the claim that amnesia interrupts personal identity
over time. So Par¬t presents an alternative case without amnesia but with
fuzzy memory instead.18 Also, one reason some may want to have an
operation in the past is so that their life always gets better. But with
greater speci¬cation of Par¬t™s case, we could assure that the overall pattern
of improvement in the life as a whole is steadily upward in any relevant
sense, however the local trouble about the operation is resolved. Despite
these complexities, for our purposes the key point is Par¬t™s conclusion,
which is that the preference elicited by these cases, which most of us
share, is not irrational even though it is in con¬‚ict with the traditional
maximizing conception of prudence.19 It is a perspectival preference.
There are some alternative explanations of our intuitive response to this
case that, if properly available, could eliminate the con¬‚ict between the
Traditional View and this response. For instance, it might be that the past
is unreal, indeed more unreal than the future, or that desires cannot work
backward. But no such explanation is suf¬ciently convincing or widely
held. If the Traditional View is incorrect because of Par¬t™s case, then
that suggests a corresponding Perspectival view, according to which the
normatively relevant quality of one™s life turns on whether one™s suffering
is in the past or in the future.
The three classes of views that might explain our ¬rm sense that tem-
poral distribution of well-being within a life is normatively signi¬cant are
now on the table. The third step of our main argument will involve two

17 Par¬t (1984: 165“166).
18 Ibid., 167.
19 Ibid.

subarguments, which conclude that Objective and Perspectival views are
incorrect. This will force us, since distribution within a life does matter,
to Timeless views like an egalitarianism of periods.
We will begin with Perspectival views. Certainly the pattern of prefer-
ence and intuition that Par¬t has uncovered is deep and real. But it is also
relevant that Perspectival views are in con¬‚ict with traditional philosoph-
ical conceptions of normative rationality in a striking way, conceptions
that encompass both the Traditional View and Timeless views.
It has generally seemed to philosophers that a preference for goods
rooted simply in their relation to now is a paradigm of irrationality. This
very traditional and standard conception is, I believe, correct. The prefer-
ences or intuitions that Par¬t has uncovered are not intuitively rational in
the strong normative sense of that word, though perhaps because we all
have them to one degree or another they are not wildly irrational. This is
for exactly the reasons the tradition would have offered, and indeed that
Mackie™s case invokes, that they treat the relevantly alike differently.
But Par¬t™s cases may disturb our traditional certainty on this issue.
For one thing, we may wonder if there can be nearly universally shared
preferences of this general sort that are not fully rational.
But in fact we already know that there can be. Perspectival temporal
patterning of the sort that Par¬t has uncovered, a preference for future
good just because it is future, seems very likely to have an evolutionary
explanation. It is quite plausible that we have evolved to be more con-
cerned about harms and goods to come than harms and goods gone by. In
the common “ and hence most evolutionarily salient “ case, we can affect
by our action and choice only harms and goods that are to come, and not
harms and goods gone by. Still, it would be wrong to conclude quickly
that because evolution makes these tendencies inevitable in us, it also
makes them normatively probative. Some things for which we have plau-
sibly evolved tendencies “ for instance, certain forms of sexual jealousy or
vengeful anger “ are not by that very reason feelings that morality must
endorse, as even sociobiologists and evolutionary psychologists are usually
quick to admit. And certainly we aren™t ineluctably forced to act in accord
with these tendencies, or with those that Par¬t has uncovered. Evolution
is not in itself an adequate rationale. And tradition does offer the plausible
and intuitive rationale for thinking these tendencies are irrational.
Still, consider possible resistance. Evolution may seem to provide more
rationale than I admit. If there is an evolutionary explanation for the pref-
erences in question, then it must be that we have such preferences because,
other things equal, our ancestors were more likely to survive, thrive, and

leave offspring than human beings who lacked such preferences. So it
would seem that such preferences are generally in human interest.
But this resistance is mistaken. Individual interests often come apart
from evolutionary interests. People can choose on grounds of self-interest
to have no genetic descendants, and to ignore the descendants of their
kin. People retain their interests once they can no longer breed, and even
once they are no longer much use to their genetic descendants. And the
pain felt by children is contrary to their interests even if they grow out of
it and it never interferes with the descent of their genes. Evolution is not
Even in light of traditional rationales, and even if evolution is an insuf-
¬cient rationale, we still may feel uncertain in the face of Par¬t™s cases. The
best way to undergird the traditional charge against Perspectival views is
to consider analogous hypothetical cases, in which our intuition will not
be tempted by familiarity into thinking a phenomenon is inevitable and
natural in some normatively relevant way.
So imagine a set of sea creatures that evolved to have a tendency to
favor bene¬ts received on their right sides, as we favor bene¬ts to come.
And imagine that we are distributing bene¬ts to one of these creatures.
How should we do it?
Perhaps we would begin by taking their tendencies and preferences
seriously. But we have already discussed some of the problems this presents.
Bare uncorrected preference, we know, is not normatively probative, and
may differ over an individual™s life.20 And it is important to remember
that cases we use to test our intuitions about the relevance of these spatial
distributions of well-being must incorporate into the relevant well-being
any necessary corrections for the effects of anticipation and memory,
and also for the effects of frustration felt if particular desires for spa-
tial patterns aren™t met. We must presume that the bene¬t we are dis-
tributing to our creatures is some sort of basic normative good, and
we must choose between a pattern that ¬ts their chiral preference and
another with an equal or greater sum of good.21 With those corrections
introduced, their preferences do not seem so clearly decisive, and indeed
seem irrational. Good is not more normatively signi¬cant if it is on the

20 It might, for instance, differ over an individual™s life in our case as they undergo various
forms of anti-evolutionary training to overcome their innate tendencies “ in other words,
as they grow up and mature.
21 Or (if it doesn™t make sense to talk about sums of well-being) in which we must choose
between a pattern that ¬ts their chiral preference and the same pattern mirror-inverted but
with perhaps some extra well-being thrown in here or there.

right. There are few things of which we have greater normative certainty
than that.
Still, the fan of evolution may resist. If these preferences evolved, then
it may seem that the preferences must be in the interest of the crea-
tures in question. But this argument is no more decisive than before. By
some odd twist of circumstance, only what happens on the right sides of
these creatures is relevant to their breeding and their protection of their
descendants, as only certain time intervals matter for us. The rest of such
a creature hangs out unprotected by evolutionary history on the left side.
Still, what™s on their left sides plausibly matters to their interests.
One way to bring out the relevant strangeness of these chiral pref-
erences is this: Notice that if we distribute goods in accord with these
tendencies, which we will assume are ¬xed over time, that will reveal
some practical strain in those preferences and in our consequent action.
As a single individual moves around and about, what is on the right can
become largely what is on the left.22 A distribution that was correct no
longer is correct; one that satis¬ed no longer satis¬es. Perspectival tem-
poral patterning, which pivots on the now, exhibits similar strains, since
the now changes. While today you may wish that you had the longer
operation yesterday rather than the shorter one tonight, tomorrow you
may have a different and more objective view about what would have
been best. And if you don™t care about what happened in your past at
all, still you may wish before any operation to have a shorter operation
later, and yet retain the perspectival wishes that Par¬t uncovered.
I believe that we need to be wary of arguments against normative
theories rooted in practical self-defeat arguments,23 though it may be
relevant that Par¬t himself favors certain arguments of this general class.24
But my point in stressing these practical strains here, and also the analogy,
is merely to generate a proper sense of the strangeness and irrationality of
the preferences that Par¬t™s cases have uncovered in us.
Even if we have such preferences, which we do, and even if they
evolved, which they did, that does not mean that they must be re¬‚ected
in any proper normative theory. Nor does it mean that they are supported
by a properly re¬‚ective normative intuition that has faced relevant facts.
The fact that we have these preferences is aptly explained by evolution

22 Their left hand of course stays on the left.
23 For worries about self-defeat arguments, see Mendola (1986).
24 Par¬t (1984: 3“114).

in a manner that is completely independent of their normative probity.
And what relevant rationales we have rule against them. The very familiar
and traditional argument that it is paradigmatically irrational to treat likes
differently without a suitably differentiating reason shows that Perspectival
accounts of well-being are problematic, and this is reinforced by practical
self-defeat arguments and by my analogy, both of which show that they are
strange. The traditional intuitive pull of Timeless views or the Traditional
View, which remains even in the face of these cases, is one re¬‚ection
of these facts. Suitably re¬‚ective and informed commonsense intuition
certainly does not rule in favor of Perspectival views, and on balance rules
against them.
What about Objective views, which endorse objective temporal pat-
terning that does not pivot on the now? First, distinguish between Objec-
tive views that are temporally symmetrical “ for instance, views that hold
that what matters is what happens in the temporal center of life “ and
those that are temporally asymmetrical “ for instance, views in which a
life that improves is better than a life that declines at the same rate. Then
consider cases, similar to those that we recently discussed, that uncover
our views about analogous spatial asymmetries.
Imagine a creature that has evolved tendencies to pursue an arrange-
ment of things A, and hence a world, that is an exact objective mirror
inversion of another possible arrangement B, with everything switched
on an east-west axis running through, say, its place of birth, even when
B is available to it much more easily, more certainly, and at less cost. The
creature likes good things to be piled up in the East. If one hand hurts, it
strives to keep it in the West. Again, presume that this is for evolutionary
reasons. Such a tendency or preference, even if evolved, like a preference
for a world just like this one but displaced three feet toward Polaris in
absolute space, seems a paradigm of irrationality. Indeed, it seems to be of
roughly the same sort of paradigmatic irrationality as Par¬t™s famous cases
of intrinsically irrational desires. For instance, there is Future-Tuesday
Indifference, in which one prefers agony on Tuesday to mild pain on any
other day.25 (Par¬t™s Within-a-Mile Altruism, which leads one to care
greatly about the well-being of all those within a mile but not at all about
those who are a mile and a quarter away, is perhaps analogous to Per-
spectival views, since what™s within a mile changes as one moves around.)

25 Ibid., 124“125.

We can imagine someone with these tendencies; we can imagine some-
one evolved to be like that. But that is insuf¬cient to provide such hypo-
thetical tendencies intuitive probity. They do not suffer from the same
kind of practical incoherence as the chiral preferences we previously dis-
cussed, since what is in the East doesn™t shift as we move around in the way
in which what is to the right can shift. For this reason, not Within-a-Mile
Altruism but rather Within-a-Mile-of-Cleveland Altruism is a proper ana-
logue of objective asymmetrical patterning. But such tendencies are still
quite strange and paradigmatically irrational.
We are a temporal analogue of this spatial inversion case. Just as we
evolved to have Par¬t-style perspectival tendencies, so we evolved, and
for much the same reasons, to focus especially on what happens later in
a life at the expense of what happens earlier, to prefer a life that gets
better to a life which gets worse, even if we correct for the pleasures
and pains of anticipation and memory and equalize sums of well-being
over a life. Our imaginary creature has preferences between different lives
that re¬‚ect a tendency to favor bene¬ts to the East. That tendency and
those preferences are not normatively probative. Neither are our objective
asymmetrical temporal preferences.
There may even be a way in which our objective asymmetric prefer-
ences are worse than those of the creatures we just considered. Those with
asymmetrical temporal preferences of the ever-upward sort can always be
led to delay some grati¬cation up until the moment of their unforeseen
death in a strange and ultimately unintuitive way. It is not that they will
need to delay all grati¬cation, of course, but that they will need carefully
to husband resources to assure a continuous and ever-upward trend on
into the unforeseen future and hence on beyond their uncertain death.
They must strive to escape any ¬nal cadence, even though the time of
their death may be very indeterminate. Indeed, they may even need to
choose today a somewhat lesser well-being over a greater well-being avail-
able at the same cost so that tomorrow things can still get better. This is
at least an analogue of practical self-defeat. These preferences, like those
of our analogous creatures, are strange. Evolved preferences like these are
not automatically probative, and the traditional rationale against asym-
metrical temporal preferences remains. Treating different cases differently
without reason, as in Future-Tuesday Indifference as well as in objective
asymmetrical temporal patterning, is paradigmatically irrational.
There is no more difference in the quality of a life, all other things
being equal, due to temporal inversion in good things in that life than to
objective mirror inversion in such good things. This is perhaps obscured

by the fact that it is hard to remember that we must control for effects
of anticipation and memory and for the satisfaction of preferences about
the shape of a life when considering sums of well-being to be ¬xed in
appropriate temporal inversion cases.
We have yet to consider objective symmetrical temporal patterning,
which would be preserved under temporal inversion. But no patterns of
this sort have much to be said for them even on immediate and unre¬‚ective
grounds. And indeed, we have already noted a troubling spatial analogy for
these cases in Within-a-Mile-of-Cleveland Altruism. We should conclude
that no Objective view is appropriate.
So we come to the end of our main argument. Distribution within a
life does matter, so the Traditional View is incorrect. But there are three
ways in which it can matter, and we have seen it cannot properly matter
in the ways favored by Perspectival or Objective views. So we are left
with Timeless views like an egalitarianism of periods, which, unlike the
Traditional View, grant periods of lives equal normative status of the sort
favored by many for persons as wholes. And this view has an intuitive and
traditional rationale, which Mackie™s discussion invoked. Periods matter.
And we have already noted that this suggests that short periods matter too.


We have traversed our ¬rst stage in defense of ¬‚ecks, from lives to periods,
and indeed to short periods. It appears that a properly chastened and
re¬‚ective commonsense intuition would favor short periods of lives as the
basic moral patients, the basic subjects of properly egalitarian moral and
rational concern. Or something like that.
I aim in this section to improve on this approximation. But ¬rst note
a possible generalization. Asymmetric objective temporal patterning, and
speci¬cally temporal order, is relevant in many ways in commonsense
morality. I have attacked the normative signi¬cance of temporal order
only for one speci¬c set of issues. Nevertheless, my argument in the
last section suggests a generalization that would have other important
ethical implications. Consider this possible generalization of my main
claim: Moral claims that depend crucially on temporal order are incorrect.
And then notice that deontological views often violate this constraint in
various ways.
First, this general constraint supports traditional hedonist arguments
against the special intrinsic evil of death, contrary to the normative asym-
metry of death and birth embraced, for instance, by Kamm. It undercuts

intuitions that projects completed after death matter to individual well-
being in a way that desires satis¬ed before birth do not. And it would rule
out McMahan™s recent suggestion of Time-Relative Interest Accounts of
the wrongness of killing and of the badness of death.26
Second, the constraint suggests that it is wrong to focus normative
judgment on the causal and temporal consequences of acts in particu-
lar, as opposed to the alternative states of affairs identi¬ed with alterna-
tive actions independent of temporal order. This may be of signi¬cance
because of Newcomb™s problem and the con¬‚ict between causal and evi-
dential decision theory.27
Third, certain sorts of harms and bene¬ts are such only by essential
reference to temporal order. They are essentially a making worse or a
making better. There are some harms, which we might call modal harms,
which are such independent of temporal order. Presume that if you do
X, then Y occurs also, and if you don™t do X, then Y will not occur,
with temporal order irrelevant. If Y leaves someone worse off than they
otherwise would be, then we might call doing X a “modal harm”. Modal
harms are de¬ned as such against a modal and not a temporal baseline “
what happens if you don™t do X, not what was true before you did X.
But modal harms and bene¬ts do not necessarily deliver all the features of
harm and bene¬t on which deontological critics of consequentialism rely.
For instance, it may be that modal baselines are indeterminate, so that it is a
¬rmer and more determinate fact that you ran over somebody than that if
you had tried you would have avoided them.28 And the acting/refraining
distinction is not, even if modally available, necessarily the same as the
harming/not helping distinction, which more crucially involves temporal
order. Consider Heidi Malm™s case: If you are headed downstream in a
canoe and will run someone over if you fail to paddle aside, your failing
to paddle may be your killing someone.29 Modal harm and bene¬t seem
insuf¬cient to deliver the difference between killing and letting die. Of
course, MAC can deliver this difference, but only by appeal to a particular
sort of general group act that may in fact have more bene¬cent relevant
In general, a sensitivity to temporal asymmetries, of the sort we have
seen here to have plausibly evolved in us but to be without probative

26 McMahan (2002).
27 For discussion of such issues in a normative context, see T¨ nnsj¨ (1998: 140“152).
a o
28 Mendola (1987).
29 Malm (1989).

normative weight, runs very deep in deontology. It may ultimately tell
against certain features of our commonsense morality. But of course these
matters are complex, and these remarks merely suggestive.
It is a different extension of our argument that is most relevant in this
context. So far in this chapter, I have concentrated on defending as prop-
erly intuitive the claim that brief periods of lives are a central locus, if
perhaps not exactly the most central locus, of proper distributional con-
cern. I have concentrated on the ¬rst stage of our two-stage discussion.
Brief periods are in a certain sense midway between the lives that con-
cern many contemporary theorists and the ¬‚ecks that are our ultimate
concern. This has allowed us to engage at once contemporary literature
and intuitions, on the one hand, and our own speci¬c concerns, on the
other. But we must close the gap. We must traverse the second stage of
our journey, from short periods to ¬‚ecks. Some brief temporal periods
are longer than even the ¬nite phenomenal moments that ¬‚ecks persist,
and entire phenomenal moments of experience contain many ¬‚ecks.
Still, it would seem that any rational considerations that constitute brief
periods as distributionally signi¬cant also constitute briefer but phenom-
enally distinguishable periods as signi¬cant also.
And it is important to remember that real painfulness in our world does
not come one ¬‚eck at a time, or for that matter one moment at a time.
Almost all real pains involve many ¬‚ecks at any single moment and indeed
at more than one moment. So there is a limit to the practical signi¬cance
of the focus of HMP on ¬‚ecks rather than moments or even somewhat
longer periods.
My third point is that in the context of the other elements of HMP,
some structural relationships among the worst ¬‚eck, worst moment, and
worst life further undercut the depth of the differences at issue here. If
we ignore for the moment details introduced by ties between ¬‚ecks of
value and by the lengths of lives, then the worst-off life by the direct
light of HMP would be that which contained the worst-off moment,
irrespective of what else goes on in it. And the worst-off moment by
the direct light of HMP would be that which included the worst ¬‚eck.
This argument deploys structural relationships that depend on maximin
distribution and not merely a focus on ¬‚ecks. And to tie the value of a life
so closely to its worst ¬‚eck, especially in the context of a hedonist value
theory, may seem itself dangerously unintuitive. But in section VI and
Chapter 8, we will explore the detailed interaction of a focus on ¬‚ecks,
maximin, hedonism, and plausible duties regarding murder and suicide.
We will see that this interaction is properly intuitive. Still, there are some

complications that can unlink ¬‚ecks and moments and lives. There are
some cases in which HMP does suggest that we should bene¬t a better-off
life (in its worse moments) at the expense of a worse-off life (in its better
moments). It is better to make a relatively worse moment in a better
life better at the expense of a relatively better moment in a worse life.
And a similar structure informs the relationship between moments and
¬‚ecks. The worst moment by direct application of HMP might properly
be made worse to the bene¬t of a worse ¬‚eck in a better moment. But
yet these particular complications are relatively unimportant and limited,
and also hard to bring to direct confrontation with common sense. If
I have convinced you that short periods of lives are intuitively close to
the proper primary locus of distributional concern, then that should be
adequate to show that the apparent intuitive dif¬culty of HMP that is
now under consideration is in fact not a dif¬culty at all, but rather almost
an intuitive advantage.
My fourth and perhaps most important point is that we have already
considered some direct arguments that spatial distribution of well-being
within a moment of a person™s life matters. It makes no coherent sense to
build up suffering on your left side to allow for greater happiness on your
right side. It is intuitively absurd to maximize the sum of well-being in
a moment of your life by trading a searing pain in your left index ¬nger
for a mild pleasure across the rest of your body. Your ¬nger intuitively
deserves some distributional concern. That is a quite intuitive and com-
monsensical point, though one ethicists have generally overlooked. If the
spatial distribution of well-being within a moment of life matters, and it
intuitively seems that it does, then this suggests an egalitarian spatial dis-
tribution, for reasons analogous to those we considered in the last section.
And once spatial bits matter, it is hard to believe that tiny spatial bits do
not matter.
There is another conception of phenomenal hedonic value that is a
lot like the conception I have suggested here, in fact a conception that I
once favored,30 which is inconsistent with some of these claims. It is not
unreasonable to suggest that your experience of physical pleasure and pain
presents itself as a general overall feeling about the state of your body. That
would involve no troubling extension of distributional concern beyond
moments of lives to mere ¬‚ecks. But I have moved to the conception
deployed here as a result of what seems the most plausible account of the

30 Mendola (1990b).

nature of phenomenal hedonic value. And though that forces us to ¬‚ecks,
we have now seen that such a conception is not forcefully unintuitive.
Of course, my arguments here have not been suf¬cient to force us on
intuitive grounds to a focus on ¬‚ecks as the basic locus of distributional
concern. But we can properly conclude that the focus of HMP on ¬‚ecks
lies at least in the rough direction suggested by a properly re¬‚ective com-
mon sense, where a rough direction is all common sense can provide on
these complicated matters. We cannot properly expect more detailed res-
olution from our vague common sense than that. And indeed, I merely
insist on a doubly negative claim. We do not know on intuitive grounds
that ¬‚ecks don™t matter. So my claims here are even weaker than the claims
I made for hedonism on the basis of intuition in Chapter 4.
You may retain contrary intuitions about ¬‚ecks. But it is important to
remember that the intuitive detailed normative implications of HMP rest
partly on MAC. As we will see in the next part, that matters.


We next will trace the interaction of commonsense intuition and maximin
principles for ranking outcomes. The literature on maximin presumes that
lives, and not moments or ¬‚ecks of lives, are the central moral locus, the
proper basic unit of distributional concern. But in this section I will pre-
scind from that detail, and trace the interaction of intuition and maximin
whatever the central moral patients happen to be. We will return to the
interaction of maximin and a focus on ¬‚ecks in section VI.
HMP differs from classical utilitarianism in its extreme concern for the
worst-off, and there are apparent intuitive objections to any view of that
sort, whether the worst-off be moments or ¬‚ecks or lives. Maximin prin-
ciples provide an obvious route to the uni¬cation of plausible concerns
with maximization and with distributional equity. And yet it has seemed
that maximin principles are not really a plausibly intuitive way to dispose
of the objection from distribution to traditional utilitarianism. Almost no
one believes that direct application of such principles is remotely plausible.
John Rawls was the most prominent proponent of maximin principles.31
But even he cushioned the effects of maximin in certain ways that are
not available to maximin consequentialism. For Rawls, maximin princi-
ples govern only expectations for primary social goods of representative

31 Rawls (1971).

members of basic social groups in just societies, and indeed merely in
regard to economic expectations inside a ¬xed context of political liber-
ties and equality of opportunity. Rawls, himself, suggested that maximin
principles will be in con¬‚ict with common sense when applied in a more
straightforward way.32
Still, maximin consequentialism would, in fact, be found intuitive by
properly informed and re¬‚ective common sense. Or, at the very least, we
do not know on intuitive grounds that it is mistaken. We will get to that
conclusion in two steps. The ¬rst step will be to notice the truth of a
certain conditional claim. But it will take a moment to do even that. My
initial point is this:
There are certain concrete cases that seem intuitively to favor maximin,
at least relative to traditional competitors, and certain concrete cases that
seem intuitively to count against it. Yet any legitimate difference between
those cases must rest on the deep moral signi¬cance of the difference
between acting and refraining from action, or, to be more exact, between
failing to aid and harming. But the moral signi¬cance of that distinction
is itself quite controversial. It is not beyond commonsense dispute, and is,
indeed, one of the central outstanding issues between consequentialism
and its deontological and other critics.
In its simplest forms, consequentialism does not support any moral
difference between failing to aid and actively harming when they have the
same consequences. Even in sophisticated forms, consequentialism often
does not support any deep such difference. Many, of course, consider
this yet another objection to consequentialism. But my ¬rst goal in this
section is merely to link two standard objections to standard forms of
consequentialism, and to support a conditional claim: If consequentialism
can legitimately undercut the apparently intuitive objection that there is
a deep difference between failing to aid and harming, then it can evade
the objection from distribution to classical utilitarianism by adopting a
maximin principle.
We will return to the antecedent of this conditional in the second step
of the argument of this section. But it is easy to pre¬gure: I believe that
the difference between acting and refraining doesn™t have the deep moral
signi¬cance suggested by some elements of common sense. Nor does that
between harming and not aiding. I believe that there are some differences
here, just not deeply signi¬cant ones. That, in fact, is the implication of the

32 Ibid., 153“157.

treatment of these very distinctions we developed in Chapter 3, though we
will need to come back to that point to see it properly. I believe that these
differences aren™t signi¬cant enough to support a differential treatment of
the cases we will face here. Because there is no suf¬cient moral difference
between the cases that suggest maximin and those that seem initially to
intuitively count against it, we are forced to choose between our intuitions,
between those in accord with maximin and those that count against it.
I furthermore believe that the very facts that show that there is not the
requisite difference between the cases also favor the intuitions that suggest
maximin and undercut the intuitions that apparently constitute objections
to it.
But my ¬rst job here is merely to establish the truth of a conditional,
and not the truth of its antecedent. My ¬rst job is merely to probe the
interaction of, on one hand, the intuitive status of maximin and, on the
other, the signi¬cance of the difference between harming and failing to
aid. Maximin will be seen to be suitably intuitive if that distinction is not
of suf¬cient normative signi¬cance. So let me begin my argument for the
truth of that conditional. For the moment, try to forget about Chapter 3
and the antecedent of the conditional.
We will begin by considering several cases that pull in different direc-
tions. The ¬rst case favors maximin relative to familiar competitors, at
least if such a case can tell us anything about how to evaluate alterna-
tive outcomes. Consider the intuitive force of Ivan™s claim in The Brothers
Karamazov that he, unlike God “ that he against God “ would not con-
sent to the torturing of one child to create great happiness for the rest of
mankind. The suffering of that consequently worst-off individual intu-
itively trumps a concern for the well-being of the rest. Ivan says to his
“[I]magine that you yourself are building the edi¬ce of human destiny with the
object of making people happy in the ¬nale, of giving them peace and rest at last,
but for that you must inevitably and unavoidably torture just one tiny creature,
[a] child who was beating her chest with her little ¬st, and raise your edi¬ce on
the foundation of her unrequited tears “ would you agree to be the architect on
such conditions? Tell me the truth.”
“No, I would not agree,” Alyosha said softly.33

Many share Ivan™s and even Alyosha™s intuition that there would be
something morally and humanly reprehensible in such an act.

33 Dostoevsky (1990: 245).

Of course, the child isn™t just worst-off. The case involves torture,
indeed torture of an innocent. So many ethicists will claim that this case
doesn™t reveal anything about the relative evaluation of outcomes. The
outcome Ivan considers may be a better one, they think; it is just not
appropriate to get there by torturing someone. Part of my job must be to
show that the torture isn™t a distracting and misleading feature of the case,
or, rather, that it isn™t a distracting and misleading feature of the case if
there isn™t a deep distinction between harming and failing to aid. Please
suspend for the moment your belief that I cannot possibly do that.
First consider other cases, in certain ways closely analogous cases, that
seem to pull against maximin. The most thorough negative discussion of
maximin is Larry Temkin™s. Like Rawls™s, Temkin™s discussion is speci¬-
cally focused on the relative justice of circumstances, as opposed to other
virtues. But this seems a detail that we can properly ignore here.
Temkin begins with a situation he calls S, with one well-off group,
one not so well-off group, and one still less well-off individual. Here
are Temkin™s main contrasting cases and his main conclusion rooted in
intuitive reactions to those cases:
A, B, and C are alternatives to S. In A, the worst-off person would remain
unchanged, the better-off group would be slightly lowered, and everyone else
would be dramatically raised. In B, the worst-off person would be slightly raised,
but everyone else would be lowered, and while the others who initially fared
poorly would still not be as badly off as the worst-off person, each of them would
lose more than the worst-off person gained. In C, the worst-off person would be
raised slightly more than in B, the better-off people would be raised signi¬cantly,
and the others would be lowered to the worst-off person™s new level, each (again)
losing more than the worst-off person gained.
On a Maximin Principle of justice focusing on the worst-off person, B and C
would both be more just than A, with C most just. Many ¬nd this unacceptable.34

I admit the force of these intuitions. Indeed, let me be yet more con-
cessive. Temkin™s analysis plausibly captures not only our initial intuitions
about these cases, but also the intuitive reasons for these reactions. Max-
imin is intuitively incorrect for two closely related reasons, according to
Temkin. Here is the ¬rst:
Although considerations of justice may focus our attention on the worst-off
person, and we may even be most concerned about her situation, surely her plight
is not our only (signi¬cant) concern regarding justice. . . . [I]t is implausible that

34 Temkin (1993a: 103).

we should be deeply and genuinely concerned about bene¬tting the worst-off
person, but not concerned at all about bene¬tting those who fare very poorly but
are (ever-so-slightly) better off than she. In bringing about B or C rather than A,
members of S would be directly harming, as well as failing to signi¬cantly bene¬t,
many who are themselves quite badly off “ all for the sake of slightly bene¬tting
the worst-off person.35

Here is Temkin™s second reason:
Let A— and C— be large populations. . . . Let A— be perfectly equal with everyone
faring very well, C— very unequal with the worst-off group faring very poorly.
Finally, assume that in Rawlsian terms A— is perfectly just, and C— terribly unjust.
Clearly, the committed judgment of Rawlsians . . . would be that regarding justice
A— would be much better than C— . But notice, A— and C— might be represented
by A and C . . . , except for the worst-off person. Yet . . . focusing on the worst-off
person, Maximin would rank A and C the exact reverse of A— and C— . This seems
implausible. . . . [F]ocusing on the worst-off person . . . may seriously distort our
judgment because the worst-off person™s condition may not accurately re¬‚ect the
situation™s overall justness. . . . [It] may simply by an anomaly or ¬‚uke.36

So where are we? Temkin™s cases and the general rationales he cites
have some genuine intuitive force. But we also face Ivan Karamazov™s in
some ways very similar case, which seems to pull our intuition in the other
direction. We must reconcile these apparently con¬‚icting intuitions.
So our question becomes: What is the difference between Temkin™s
cases and Ivan™s? Temkin himself notes what I take to be the key surface
difference between the cases:
Although some might think it would be unjust to bring about A if one started in
B or C, many believe A would be more just than B or C, and that starting in S
justice dictates bringing about A rather than B or C.37

In other words, it apparently matters what situation we start with, because
it apparently matters whether a situation is brought about by one™s action
or merely remains the case because of one™s inaction. It apparently matters
whether someone™s horrible situation is a result of active torture by Ivan
or is merely something we ignore in order to favor the interests of better-
off others. It also apparently matters whether we will have to actively
harm those others so as merely to aid the truly worst off. The intuitive
difference between the cases rests, as I suggested earlier, on the apparently

35 Ibid.
36 Ibid., 103“105.
37 Ibid., 103.

intuitive difference between harming and failing to aid, indeed perhaps
on more than one such difference.
Temkin™s discussion is accompanied by relevant footnotes that reinforce
this point:
If we learned that A™s situation resulted from the (concealed) presence of a tortured
slave, whose tormented struggles enabled everyone else to fare so well, we might
completely revamp our judgment regarding A™s justness.38

So in effect he grants the intuitive force of Ivan™s case. And he says this
about the difference:
The relevance of the starting point to our judgments may re¬‚ect two important
strands of moral thinking. The ¬rst corresponds roughly to a distinction like
that between the good and the right to which many nonconsequentists adhere.
The good or just situation cannot always be rightly or justly brought about. The
second corresponds roughly to the asymmetry many see between harming and
not helping. Although it may seem unjust to harm those already worse off than
everyone else so as to bene¬t others, it may not seem (as) unjust not to help those
already worse off than everyone else so as to bene¬t others.39

Temkin is certainly no friend of maximin. His work represents its most
thorough criticism. But the friends of consequentialist maximin should
accept that this part of his analysis of what is going on in these cases
is essentially correct. Our intuitions about these various cases depend
directly for their probity on the presumed deep signi¬cance of the dif-
ference between harming and not helping, or on analogous presump-
tions that otherwise better situations cannot justly be brought about. The
friends and enemies of maximin should agree so far.
But the rub is this. Many consequentialists and at least some strands
in common sense do not accept that these factors have the signi¬cance
that Temkin presumes. We certainly have no normative consensus that
the paradigmatic consequentialist view of these issues, the view that these
alleged differences are not deep differences, is incorrect. If consequen-
tialists can properly insist that there is no deep difference between harm-
ing and failing to aid, then the intuitive case against maximin is not so
Certainly if the moral difference between aiding and not harming is
deep, then maximin is intuitively implausible. I grant that. But what if it

38 Ibid., 105.
39 Ibid., 104.

is not? If it is not legitimate to treat Temkin™s and Ivan™s cases differently
on such a ground, then what are we to conclude? The worst that could
reasonably be said against maximin on that presumption is that common-
sense intuition cannot resolve its plausibility, I believe. But in fact on such
a presumption we can do better than that.
If, for simplicity™s sake, we presume that there is no morally signi¬-
cant difference between acting and refraining, or between harming and
not aiding, that would not intuitively serve to remove moral responsibil-
ity for the effects of one™s action, but rather to extend it to effects of
one™s nonaction. Consequentialists characteristically think we are morally
responsible for more that happens. And it is one™s responsibility for the
suffering of the child in Ivan™s case that so deeply evokes our intuitive
Deontologists think that torturing someone is specially and grievously
wrong in a way that letting someone be tortured is not, and some ethicists
think that it is worse if someone is tortured than if a tree falls on them and
does comparable harm. But consequentialists characteristically argue that
letting someone be tortured is as bad as torturing, and that letting a tree
fall on someone may be as bad as letting them be tortured, not that tor-
turing is as normatively insigni¬cant as some deontologists presume that
letting someone be tortured is, not that letting someone be tortured is as
normatively insigni¬cant as some believe letting a tree fall on someone
is. That indeed is the intuitive implication of a collapse of the customary
distinction between harming and not aiding, to extend responsibility “ in
other words, to undergird our normative response to Ivan™s case and sug-
gest a generalization, and to undercut our normative response to Temkin™s
cases. It is not the torture in Ivan™s case that misleads our intuition, but
rather the lack of responsibility presumed in Temkin™s cases, and indeed
generally presumed by unre¬‚ective common sense when it confronts such
Since this is the key point, we should pause to consider natural resis-
tance. Temkin does at least sometimes suggest a situation of choice among
the options he provides, and speaks of harms from and gains from a given
baseline, and that may invoke responsibility of at least a weak sort. But his
goal is proper intuitions about the justice of outcomes, independent of
whether or not anyone has responsibility for them. And you may think
that the way to uncover these intuitions is to remove as much responsi-
bility from the situations to be considered as possible, not to heighten it
dramatically in the way that torture does. In other words, you may think
that Temkin™s procedure is basically right, but that he should purify his

cases yet further to remove those elements that may even weakly suggest
responsibility for the outcomes in question. It may seem that our intuitive
aversion to maximin in cases when we compare outcomes over which we
know we have no control is quite revealing.
But consequentialists should disagree. They should insist that our real
and unconfused evaluation of outcomes is revealed when the full weight of
our responsibility for anything that happens because of our choice is felt.
Of course, things are not quite that simple and clean. There are com-
plications. First of all, there is a series of differences between Temkin™s
cases and Ivan™s torture that I mostly lumped together in the previ-
ous paragraphs. There is the difference between torturing and letting
someone be tortured by someone else. And then there is the difference
between someone™s being tortured and their being otherwise harmed
in the same way. And then there is the difference between their being
harmed and their starting out in a bad situation from which they might be
Moreover, there can be some moral difference between harming and
not aiding with the same consequences, even if there is not a deep and
signi¬cant difference. As I said, I believe that the facts in the end are
that there is some difference between harming and not aiding, and indeed
that there is some difference along each of the steps we might distinguish
between Temkin™s cases and Ivan™s. Chapter 3 showed why. But, as we will
shortly see, these are not suf¬cient differences, nor indeed the right kind
of differences, for us to conclude that the torture present in Ivan™s case
is intuitively misleading while the lack of responsibility in Temkin™s cases
is not. It is still rather the presumed lack of responsibility in Temkin™s
cases that misleads our intuition in the crucial way. And my own analysis
is not idiosyncratic for a consequentialist. Paradigmatic consequential-
ists who grant some difference between the cases will still attempt to
extend our responsibility for outcomes. According to this whole class of
views, not the many small but real moral differences between torturing
someone and leaving them in some equally painful state, but rather the
general false presumption that we would not be responsible in Temkin-
style cases, is what dominantly drives our false intuition of a normative
difference between Temkin™s cases and Ivan™s. We are responsible for the
effects of our actions and inactions in the most morally relevant sense,
consequentialists think. If properly purged of the errors underlying the
belief that there is a deep moral difference between harming and not
aiding that eliminates our responsibility for the effects of not aiding, an
alleged deep moral difference that does not in fact exist, commonsense

intuition would favor maximin. The general tendency of consequentialists
to extend responsibility suggests that even if there is some morally relevant
difference between Temkin™s and Ivan™s cases, it will be relatively minor,
and that it will indeed be such as to undercut our intuitions about Temkin™s
cases and to support our intuitions about Ivan™s.
With the complications or without, this extension of responsibility
is a controversial feature of consequentialism. Deontologists and others
will object. They do not admit an extension of responsibility beyond the
surface commitments of common sense. But for the moment, remember,
our focus is merely a conditional claim, that if there is no deep difference
between harming and failing to aid, then maximin is intuitive in the
proper way. If this is right, then what have seemed two distinct objections
to standard forms of consequentialism “ the objection from distribution
and deontology™s objection to consequentialism™s characteristic extension
of responsibility “ are connected in an unexpected way.
There is one further complexity that must be addressed even in our ¬rst
and conditional step. Consider Temkin™s ¬rst rationale against maximin,
which stresses that maximin would somewhat sacri¬ce the well-being
of many pretty badly-off persons in order to bene¬t the very worst-off.
We haven™t directly discussed that objection. And his second rationale
against maximin reinforces the worry that it is inappropriate to focus, like
maximin, merely on the worst-off individual. We haven™t yet considered
these central elements of Temkin™s alternative explanation of what is going
on in these cases.
But note that if we prescind from the elements of Temkin™s cases that
evoke questionable intuitions about the normative signi¬cance of the
difference between harming and not aiding, and assume that the agent in
question has moral responsibility for everything in the various options,
some recent and quite intuitively sensitive deontological accounts suggest
a treatment in accord with maximin consequentialism.
For instance, Scanlon has granted that

in situations in which aid is required and in which one must choose between aiding
a larger or smaller number of people all of whom face harms of comparable moral
importance, one must aid the larger number. On the other hand, [the correct
view] does not require, or even permit, one to save a larger number of people
from minor harms rather than a smaller number who face much more serious

40 Scanlon (1998: 238).

In a situation where one is responsible for saving large numbers from small
harms or a few from larger harms, save the few. That may suggest that we
should generally focus on helping the few worst-off people at the expense
of larger numbers of badly-off people.
Naturally, there are complications. Perhaps the course Scanlon suggests
will still seem plausible even if those facing the larger injuries will end up
all-told better-off in one way or another, whether they are injured or not.
But at the very least that possibility leaves our intuitions less clear.
A second complication is that maximin requires other tough conclu-
sions. Imagine that a million people are going to be tortured for ¬fty
years, and one person for ¬fty years and a day. And remember that for
the moment we are forgetting that mere ¬‚ecks are the true basic locus
of moral concern. Maximin suggests that it would be better to reduce
the torture of that one person by a day than to eliminate everyone else™s
torture completely, and this may seem quite implausible.
But note that in cases of that sort that are intuitively clearest, it is cor-
respondingly unclear whether the one day out of ¬fty years would make
any difference whatsoever to the person about to be tortured, who may
seem worst-off in only an abstract and mathematical sense. When the
difference to the worst-off is obviously signi¬cant, then individual intu-
itions may well differ. And of course to focus solely on cases of saving
people from harms and tortures may somewhat mislead common sense
in the ways we have already discussed. If we are not to be misled, the
consequentialist may plausibly insist, perhaps we should consider a case in
which we ourselves would have to administer the extra day of torture to
the one victim in order to relieve the sufferings of the others. Deontolo-
gists should think twice before embracing that course of action, even in
a conditional way.
Of course, there are other cases. Two otherwise identical individuals
A and B are struggling to obtain some scarce medicine, which would
enable A to live to a ripe old age rather than die peacefully now, and B
to live for another month rather than die in severe pain now. Maximin
suggests that it would be better for B to get the medicine, while it may
seem intuitively better for A to get it, since then it will do more over-
all good. There are various elements of this case that, in light of other
elements of HMP, are misleading, and on which a reply might con-
ceivably rest. Still, if we focus on a situation in which you, as a third
party, would have transparently full responsibility for both these out-
comes, in which you would have to torture B to death now so that
the medicine would extend A™s life for so long, we can see that there

are some intuitions engaged even by this case that pull in accord with
I do not claim that maximin can make us happy in the face of all hard
cases and tragic choices. No view can do that without illusion, at least if we
are responsible for all the effects of our choices. And we have no consensus
that we are not. I do not even claim that Ivan™s case, or other similar cases,
suf¬ce to show that a maximin principle is to be intuitively favored over
other relatively close and hence relatively unfamiliar competitors “ say, a
suf¬cientarian principle that requires the minimization of the number of
individuals who fall below some critical threshold of well-being “ though
I also believe that we have uncovered a general pattern of analysis for
cases that pushes towards maximin. Rather, my point is this. Maximin
is at least in adequate accord with our intuition, if consequentialism is
generally right about the nature of our moral responsibilities.
We have traversed our ¬rst step, and can turn now to the second. We
have reduced the issue of whether maximin is suitably intuitive to the issue
of whether or not there is a deep distinction between harming and failing
to aid, at least according to properly re¬‚ective commonsense intuition.
That is one central issue between deontology and consequentialism.
I have been largely pretending that we haven™t already addressed this
issue at some length, but we have. We saw in Chapter 3 that MAC deliv-
ers, in a manner appropriately consistent with commonsense intuition, a
suf¬cient difference between harming and failing to aid. But it is not a deep
difference of the sort relevant to the particular topic at hand. Still, MAC
retains the characteristic consequentialist tenet that we are in the deepest
moral sense responsible for the effects of our actions and our inaction,
because it retains the familiar rationale of all direct forms of consequen-
tialism. The moral propriety of actions rests directly on the value of their
consequences. It is merely that the application of this claim to particular
issues and cases is complicated in MAC by the fact that there are actions
of groups as well as of individuals. MAC delivers the antecedent for our
Still, I have cheated just a little, because torture is worse, according to
MAC, than letting someone be tortured, at least if we have cooperative
practices that are group agents that are properly bene¬cent and support
that distinction. Yet Ivan™s case is properly revealing. Still, the proper
perspective from which to assess our commonsense intuitions about the
ranking of outcomes relevant to basic consequentialist assessment, which
is after all our concern here, is a situation in which we feel full responsi-
bility for the outcomes. That is the kind of responsibility an omnipotent

and omniscient god would have (without recourse to a robust theodicy
and the free will defense), and which Ivan™s case invokes. MAC does sug-
gest that the violation of reciprocity that torturing involves may make it
worse than full responsibility for such an outcome of the sort relevant to
proper consequentialist assessment. But still, invoking torture remains an
appropriate device to trick our recalcitrant intuition into accepting the
responsibility for outcomes that we have in these cases and that we would
at all costs deny. It is a kind of corrective rhetoric. It is not misleading. It
just keeps one™s nose to the relevant grindstone.
This extension of responsibility may seem terrifying. But of course we
are not responsible for all the facts, as some omnipotent god would be.
And we saw in Chapter 3 that our individual responsibility is defused and
delimited in certain ways by the group agents in which we participate.
Temkin provides the most thorough and careful negative treatment of
maximin we possess. And he speaks for many who believe that they know
maximin is false. But they know no such thing, at least on the relatively
abstract structural grounds that we are now considering.


HMP is quite risk-averse, at least in one sense of that ambiguous term. It
evaluates lotteries by reference to their worst possible outcomes, even if
such outcomes are very unlikely. This may seem quite counterintuitive,
but in fact it is not.
That is largely because commonsense normative intuitions do not sup-
port any detailed, coherent treatment of normative risk. We don™t have
very determinate intuitions about how to handle risks in normative sit-
uations, and there are lots of unclarities and confusions in our intuitive
thinking about them.
First point: On the one hand, we reward and blame people “ with
moral praise and opprobrium, among other things “ at least largely on
the basis of the particular outcomes of their actions that happen to turn
up, rather than on the risks they were running. Even when we blame
someone for running a huge risk with someone else, the judgment seems
mitigated if they were lucky. Think, for instance, of the way we feel
about people who pilot a boat drunk and manage to get home safely as
opposed to those who in such a state actually kill someone. Or think of
the different ways we feel about politicians who take dangerous risks in
foreign policy, or doctors who take dangerous risks in treatment, when
they are successful and when they are unsuccessful. And yet, on the

other hand, we seem to have a strong intuition that normative evaluation
should be at least largely independent of just that sort of luck, that agents
should be morally evaluated on the basis of what is within their knowing
Second point: Worst possible outcomes, in particular, receive no very
clear treatment by unre¬‚ective intuition. While we often ignore genuine
but small risks of horrible outcomes in our commonsense normative
evaluations of actions, that may well be because we don™t think of them,
or because we don™t think that the worst risks really differ much between
one relevant option and another. When there is an intuitively signi¬cant
greater negative risk on one option rather than another, what we might
call a salient risk, either a greater probability of a very bad outcome or
a worse very bad outcome, that does seem intuitively to matter, though
not to any clear or certain degree.
Of course, there are other grounds for worry about a highly risk-
averse principle, even if common sense is not clear and decisive on these
matters. If lots of agents act in individually risk-averse ways, then it is
very likely, as the product of many individual likelihoods, that we will
end up with an outcome that is normatively worse than if the agents


. 8
( 10)