58

ISSN 2341-1465  LEAP 2 (2014): 58-86

Rethinking the Good – A Small Taste

L ARRY TEMKIN
Rutgers University

Abstract

This article aims to convey a few of the key claims and arguments of 
my book, Rethinking the Good: Moral Ideals and the Nature of Practical 
Reasoning. The article gives an example of a Spectrum Argument, and 
illustrates that such arguments put pressure on the Axiom of Transitivity, 
which holds that for any three possible outcomes or alternatives, A, B, and 
C, if, all things considered, A is better than B, and B is better than C, then A 
is better than C. The article distinguishes between two different approaches 
to understanding the goodness of outcomes, the Internal Aspects View and 
the Essentially Comparative View. It suggests that two deeply plausible, but 
seemingly incompatible, positions underlying the Spectrum Argument, 
an Additive-Aggregationist Position, and an Anti-Additive-Aggregationist 
Position, reflect the Essentially Comparative View, and that on such a 
view they are not incompatible. The article introduces several widely-held 
views about neutrality and dominance principles, and shows that some of 
these views are incompatible. The article contends that various ideals or 
views that people care about are most plausibly understood as essentially 
comparative, and notes that one such view, a Narrow Person-Affecting View, 
will be especially difficult to reject in at least some cases. It also illustrates 
how such a view, like other essentially comparative views, threatens the 
Axiom of Transitivity. The article concludes by contending that we must 
seriously rethink our understanding of the good, moral ideals, and the 
nature of practical reasoning, while recognizing that the way forward is 
murky, at best.

Keywords: Transitivity, Practical Reasoning, Internal Aspects View, 
Essentially Comparative View, Narrow Person-Affecting View, Spectrum 
Argument, Additive Aggregation, Good, Better than, Ideals.

This article is based on my Fall 2012 LEAP Lecture given at Pompeu Fabra 
University. The Lecture kicked off a symposium on my book, Rethinking the 
Good: Moral Ideals and the Nature of Practical Reasoning (Temkin 2012), 
with responses to the book offered by Oscar Horta and Ingmar Persson, 



 Rethinking the Good – A Small Taste 59

LEAP 2 (2014)

followed by comments from me on those responses.  1 The aim of the Lecture 
was not to give an overview of the book, which would have been impossible 
in the time allotted, but rather, as I told the audience, to give a very crude 
and brief tour of a few of the book’s arguments, just enough to give a sense 
for the sorts of issues the book explores. Correspondingly, this article, like the 
Lecture from which it is derived, is woefully incomplete and superficial. But, 
hopefully, some readers will find it sufficiently important and intriguing to 
turn to the book itself, where a more careful and sustained treatment can be 
found of the issues broached here, as well as many other issues central to our 
understanding of the good, moral ideals, and the nature of practical reasoning.

This article is divided into six sections. In section I, I provide a brief 
introductory remark, and offer a simple example of a Spectrum Argument. 
The Spectrum Argument puts pressure on a widely accepted principle of 
practical reasoning which may be called the Axiom of Transitivity. According 
to the Axiom of Transitivity, for any three alternatives, A, B, and C, if, all 
things considered, A is better than B, and B is better than C, then, all things 
considered, A is better than C.  2 In section II, I offer some background to 
some of the issues I discuss, and make some terminological distinctions. In 
section III, I introduce a distinction between two different approaches to 
understanding the goodness of outcomes, which I call the Internal Aspects 
View and the Essentially Comparative View. I note how two seemingly 
incompatible positions underlying the Spectrum Argument, which I call 
an Additive-Aggregationist Position, and an Anti-Additive-Aggregationist 
Position, can be seen as reflecting the Essentially Comparative View, and that 
on such a view they are not incompatible. I also note various considerations 
against rejecting the Anti-Additive-Aggregationist Position. In section IV, I 
introduce several widely-held views about neutrality and certain widely-held 
dominance principles. I show that some of these views are incompatible. 
In section V, I suggest that various ideals or views that people care about 
are most plausibly understood as essentially comparative. I focus on a 
particularly plausible version of a Narrow Person-Affecting View, and note 

1. I want to thank Paula Casal and José Luis Martí for inviting me to deliver the LEAP 
Lecture, for organizing the symposium, and for arranging for the publication of the symposium’s 
papers. I would also like to acknowledge my gratitude to Horta and Persson for their careful and 
thoughtful attention to my work. 

2. Here, I am using “the Axiom of Transitivity” as shorthand for “the Axiom of Transitivity 
of the ‘all-things-considered better than’ relation”. Elsewhere, I often put my discussions in 
terms of “the Axioms of Transitivity”, where these include the “all- things-considered equally 
as good as” and “all-things-considered at least as good as” relations as well as the “all-things-
considered better than” relation. At times, I may shorten my descriptions and just talk in terms 
of the “better than”, “equally as good as”, or “at least as good as” relations. But, unless noted 
otherwise, if I consider whether one outcome is better, equally as good as, or at least as good as, 
another, I am considering whether the one outcome is better than, equally as good as, or at least 
as good as the other all things considered.



60 Larry Temkin

LEAP 2 (2014)

how this view, like other essentially comparative views, threatens the Axiom 
of Transitivity. In section VI, I conclude with some final remarks.

1. INTROdUCTION ANd A SPECTRUM ARGUMENT

In this article, I will be discussing a number of views that are widely taken 
to be obviously true. At first blush this may seem rather odd. Why labor the 
obvious? The answer, in a nutshell, is that a number of the seemingly obvious 
views aren’t even true, much less obviously so! This follows from the simple 
fact that a number of the so-called “obvious” truths are incompatible with 
each other. Or so I shall argue anyway. Indeed, on reflection, it turns out 
that an awful lot of hard work needs to be done to sort out what we really 
should believe in the domains I shall be canvassing. I can’t do the required 
work here, in this article, but perhaps I can say enough to motivate the 
importance of taking up the task. I tried, in Rethinking the Good, to do much 
of the work in question. The result of that work, I believe, is that we need to 
significantly revise our current understanding of the good, moral ideals, and 
the nature of practical reasoning, and that such revisions will have profound 
practical and theoretical implications. The aim of this article is to provide a 
small taste of the questions addressed in my book, and what is at stake as we 
try to answer them.

Let me begin by presenting two very simple questions, and the answers 
these questions typically provoke.

My first question goes like this. Suppose that you or a loved one are going 
to have to experience a certain intensity of pain, for a certain duration, or 
a little bit less intense pain for twice, or three, or five times as long. Which 
alternative do you think would be better for you or your loved one? 

When I asked that question during my LEAP Lecture, there was total 
agreement amongst the audience of roughly forty people, that the first 
alternative would be better; that is, that an outcome involving a slightly 
more intense pain would be better than an outcome involving a slightly less 
intense pain, if the duration of the pain in the outcome with the less intense 
pain would be two, or three, or five times as long as the duration of the pain 
in the outcome with the more intense pain.

The audience’s responses were very typical. Among audiences around the 
world, involving 1000s of people over many years, virtually everyone thinks 
the better outcome would be the one with a slightly more intense pain that 
lasted significantly less long. Indeed, I estimate that over 95% of the people 
of whom I have asked my question have responded the same way; and, as 
I usually like to put it, only half in jest, if several people in an audience of 
a hundred have answered differently, typically one or two are just being 



 Rethinking the Good – A Small Taste 61

LEAP 2 (2014)

difficult, or figuring it is a trick question, and the other one or two haven’t 
fully understood the question! 

My second question goes like this. Suppose that you, or a loved one, are 
going to live for a long time. Perhaps a very long time. And there are two ways 
your life might go. In one, you will have, on average, fifteen mosquito bites 
a month for the duration of your life and, in addition, at some point in your 
life you will have two years of the most excruciating torture imaginable —
including such things as hot wax under your eyelids, bamboo shoots under 
your fingernails, electrical shocks to your genitals, and so on. You would be 
awake 18-20 hours per day, and during every waking moment your life would 
be much worse than nothing and you would wish you were dead. However, 
after the two years of torture, you would be given a pill so that you didn’t 
remember any of the pain. Further, let us suppose that the torture would 
have no permanent impact on your body or brain, and that there would be no 
other effects of any kind during the remainder of your life, once the two years 
of excruciating pain was over. In the second way your life might go, there 
would be no torture of any kind. However, instead of fifteen mosquito bites 
per month for the duration of your life, you would have sixteen mosquito 
bites per month. Bearing in mind that your life might be very long, which life 
would be better for you or your loved one; the life with fifteen mosquito bites 
throughout and two years of excruciating torture, or the life with sixteen 
mosquito bites throughout? 

To this question, all but one member of the LEAP Lecture audience gave 
the same answer. And I think it is fair to say that many audience members 
were dumbstruck when someone voted for the position that the life involving 
two years of torture would be better than the life involving one extra mosquito 
bite a month, if only the two lives lasted long enough!

As before, the reactions of the LEAP audience were very typical. Of the 
thousands of people to whom I have posed such a question over the years, 
the vast majority of them —again, well over 95% I would estimate— have 
given the same answer to this question. They think that the life involving one 
extra mosquito bite per month would be better, indeed much better, than 
the life involving two years of excruciating torture, and they think this no 
matter how long the two lives might persist. 

As indicated, these two results are very robust. But together, they are 
inconsistent if one accepts the Axiom of Transitivity: that if, all things 
considered, A is better than B, and B is better than C, then all things 
considered, A is better than C. To see this, notice that when I asked my first 
question, I didn’t actually say how intense the two pains were, nor how long 
they lasted. And I didn’t need to! This is because it seems to be a general truth 
that no matter how intense a given pain might be, and how long it lasted, it 
would be better to have that pain than one that was only slightly less intense 
but which lasted much longer. 



62 Larry Temkin

LEAP 2 (2014)

Accordingly, one can imagine a spectrum of lives, each of which would 
be very long and each of which would have, as a persistent background 
condition, fifteen mosquito bites per month. The first life in the spectrum 
would also involve extraordinary pain (the equivalent, let us suppose, of 
excruciating torture) lasting for two years, and each subsequent life in the 
spectrum would involve slightly less intense pain than that involved in the 
preceding life in the spectrum, but the pain would last two, or three, or five 
times as long as the duration of pain in the preceding life of the spectrum. 
Moving from the first member of the spectrum to the last, the pain gets slightly 
less intense though much longer, until eventually the pain has decreased so 
much that its intensity is the equivalent of but one extra mosquito bite per 
month, though instead of only lasting two years, as the pain did in the first 
member of the spectrum, the once a month mosquito-like pain extends 
throughout much, if not all, of the very long life. 

The point, of course, is that in accordance with the answer to the first 
question I asked, most people would agree that, all things considered, the 
first member of the spectrum would be better than the second, the second 
would be better than the third, the third would be better than the fourth, and 
so on. For each pairwise comparison, the life involving fifteen mosquito bites 
per month and a slightly more intense pain lasting a certain duration would 
be better, all things considered, than the life involving fifteen mosquito 
bites per month and a slightly less intense pain lasting two, or three, or five 
times as long. According to the Axiom of Transitivity, it follows that the first 
member of the spectrum must be better than the last. But the first member 
of the spectrum involves a life involving 15 mosquito bites per month and 
two years of excruciating pain the equivalent of torture, and the last member 
of the spectrum just involves 15 mosquito bites per month and many years 
of a minor pain that is the equivalent in intensity to one extra mosquito bite 
per month! Thus, as we have seen, most people would reject the claim that 
the first member of the spectrum would be better than the last. Indeed, I 
have found that most people —though admittedly not all— regard such a 
view as preposterous, if not downright absurd. 

It follows that if people want to maintain the answers typically given 
to my two questions above —answers to which, I believe, most people are 
deeply committed— then they must reject the Axiom of Transitivity.  3 

3. The first Spectrum Argument challenging the Axiom of Transitivity was developed by 
Stuart Rachels (1993). Rachels’s thinking about intransitivity was sparked by my original article 
on the topic, “Intransitivity and the Mere Addition Paradox” (Temkin 1987), but his argument 
against intransitivity was entirely original and at the time it was the strongest argument yet 
posed against the Axiom of Transitivity. Although I have developed and defended Spectrum 
Arguments in my own way over many years now, the basic structure of my arguments remains 
heavily indebted to Rachels’s original argument. Rachels’s published contributions in this 
area include Rachels 1998, 2001 and 2004. Many people have worried about the implausibly 



 Rethinking the Good – A Small Taste 63

LEAP 2 (2014)

This is a very striking result. Because the Axiom of Transitivity is one 
of the key premises underlying Expected Utility Theory, and Expected 
Utility Theory is arguably the central theory underlying game theory, 
decision theory, and much of modern economics. So, rejecting the 
Axiom of Transitivity would entail rejecting, or substantially revising our 
understanding of, game theory, decision theory, and much of modern 
economics. Since, in many ways, those theories are intended to model 
our best understanding of practical rationality, rejecting the Axiom of 
Transitivity would require us to drastically revise our understanding of 
what it is to be practically rational. 

Put differently, the Axiom of Transitivity lies very close to the core of our 
current understanding of practically rationality. We believe that just as it is 
irrational to believe both A and not A, or to prefer A to B or believe that A is 
better than B, all things considered, while at the same time also preferring 
B to A, or believing that B is better than A, all things considered, so, too, we 
believe that it is irrational to prefer both A to B, and B to C, or to believe both 
that A is better than B and that B is better than C, all things considered, while 
at the same time also preferring C to A, or believing that C is better than A, all 
things considered.

As economists would often put it, someone with intransitive preferences 
is irrational and they ought to get their preferences in order! In this context, 
the “ought” is the strong normative “ought” of individual rationality, implying 
that rationality requires that their preferences be transitive. 

It is worth adding that the Axiom of Transitivity is not merely an 
important theoretical assumption underlying our understanding of ideal 
rationality and some important academic fields, it plays an integral role in 
countless cases of everyday practical reasoning, typically without our even 
being aware of the role it is playing. For example, often when we are faced 
with a decision between various alternatives with a number of competing 
factors relevant to our decision, and a significant degree of indeterminacy 
involved regarding how much to weight each factor, we simplify our decision 
procedure by focusing on just two alternatives at a time. 

For instance, suppose we have decided to buy a new car, and based on 
our research we have narrowed our choice down to seven models. At that 
point, we might test drive the first model, and then test drive the second, and 
then, taking account of each of the factors that are important to us and how 
much we care about them —cost, gas mileage, reliability, resale value, ease 

long length of life that might be involved in the kind of Spectrum Argument presented in the 
text. I address such worries in Rethinking the Good, but also show that similar arguments can 
arise involving many different people all living at the same time, rather than a single person 
living at many times (see chapters 2, 5, and 9 for extended discussion and defense of Spectrum 
Arguments). 



64 Larry Temkin

LEAP 2 (2014)

of repairs, handling, storage capacity, power, handling, comfort, looks, extra 
features, and so on— we might determine that, all things considered, the first 
model, A, is better than the second, B. In that case, we remove B from further 
consideration, test drive C, and then decide whether A is better than C. If C 
is better we remove A, from further consideration, test drive d, and proceed 
as before. 

In this way, we might straightforwardly determine which of the seven 
models to buy on the basis of a sequence of six direct pairwise comparisons, 
with the “winner” of each pairwise comparison advancing to a subsequent 
comparison, and the “loser” being discarded from further consideration. 
As long as we are confident in each of our pairwise judgments, we will be 
confident that we have determined the best car for our purposes given our 
preferences. Moreover, given the many different factors we have to pay 
attention to, focusing clearly and carefully on the various models just two at 
a time, we will often be much more confident in any comparative judgments 
we might arrive at as to which of two cars is better, all things considered, 
than we would be in any absolute judgments about exactly how good each 
of the seven cars were, all things considered. 

As indicated, this simplifying decision procedure of focusing on just 
two alternatives at a time is a staple of many practical decisions involving 
multiple options. But, importantly, this decision procedure depends on the 
Axiom of Transitivity for its legitimacy. After all, we can only confidently 
remove B from further consideration after determining that A is better than 
B, all things considered, if we can be certain that it couldn’t be the case 
that there is some third alternative, C, which is both worse than B, and yet 
better than A, all things considered. For if it could be the case that, all things 
considered, A is better than B, which is better than C, which is better than A, 
then there would be no more reason to remove B from further consideration 
just because it is worse than A, than there would be to remove A from further 
consideration given that it is worse than C, or C from further consideration 
given that it is worse than B. It is the Axiom of Transitivity which presumably 
“guarantees” that this unfortunate predicament couldn’t arise. Thus, as 
indicated, the Axiom of Transitivity is presupposed, often implicitly and 
unwittingly, in numerous cases of everyday practical reasoning. Clearly, 
such reasoning is deeply flawed if the Axiom of Transitivity fails to hold. 

I suggest, then, that there is a great deal at stake, both theoretically and 
practically, if the Axiom of Transitivity fails. And for many years, I argued that 
Spectrum Arguments, such as the one given above, as well as various other 
arguments I developed, gave us good reason to conclude that the Axiom of 
Transitivity does fail. That is, I used to claim that we should conclude that 
all things considered better than is not a transitive relation. But my earlier 
claims were too strong, and hence misleading.



 Rethinking the Good – A Small Taste 65

LEAP 2 (2014)

What I now think is that over the years I have developed a series of 
impossibility arguments. The Axiom of Transitivity is one of the key premises 
in my impossibility arguments, but it is not the only one. Accordingly, each 
of the key premises of my impossibilities arguments are in play and, if the 
reactions to the work in this area over the years are any indication, the 
question of which of the premises should be given up is a difficult one about 
which people are deeply divided, and about which there is unlikely to be a 
consensus for years to come. 

A second key premise that is in play in Spectrum Arguments is a position 
I call the First Standard View: Trade-offs between Quality and Number are 
Sometimes Desirable. On this view, in general, it is better to experience more 
intense suffering for a shorter period of time than less intense suffering for 
a longer period of time, if the difference in the intensity of the two pains is 
sufficiently small, and the difference in their durations is sufficiently large. 

A third key premise that is in play in Spectrum Arguments is a position 
I call the Second Standard View: Trade-offs between Quality and Number 
are Sometimes Undesirable Even When Vast Numbers are at Stake. On this 
view, in general, it would be worse to receive a more intense pain of a 
significant duration than a much less intense pain of virtually any duration, 
if the difference in intensity of pains is such that the more intense pain of 
significant duration would have a significant negative impact on one’s life, 
while the less intense pain of longer duration would have little negative 
impact on one’s life.

Each of the Axiom of Transitivity and the First and Second Standard Views 
is powerfully appealing, and I believe that giving any of them up would have 
deeply implausible implications. So my current position is like that of a 
juggler, who is juggling a number of very valuable and fragile balls, and he 
can’t hang on to all of them. He has to let at least one of them drop, but 
can’t decide which one. Initially, he may decide to let the first ball drop, and 
preserve the others. But as the first ball heads towards the ground he thinks 
he can’t possibly let that ball drop, so he quickly reaches out to preserve 
that ball and lets the second ball go, instead. But he then realizes that he 
can’t let that ball drop either, so he seeks to save that one, as well, steeling 
himself to let the third ball drop. But as the third ball gets closer and closer 
to the ground he realizes he can’t bear the thought of losing that ball either, 
so reaches out to save it with the thought that he’ll let the fourth ball go. This 
process continues, till he once again finds himself letting the first ball drop. 
The problem, of course, is that the cost of letting any of the valuable balls go 
seems unacceptably high, so he frantically wants to keep each of them in the 
air, but realizes that that option is ultimately unsustainable. 

To a large extent, my book is about determining what various positions 
stand or fall together, and illuminating both the benefits and costs 



66 Larry Temkin

LEAP 2 (2014)

associated with retaining or abandoning each of the offending premises in 
my impossibility arguments. 

2. SOME BACkGROUNd ANd TERMINOLOGY

Many believe that giving up the Axiom of Transitivity is not an option. They 
believe that it is an analytic truth —literally true in virtue of the meanings of the 
words— that “all-things-considered better than” is a transitive relation. This 
is the view of John Broome (1991 and 2004), and at one time it was the view of 
Tom Nagel, Tim Scanlon, and derek Parfit.  4 I suspect that this, or something 
very close to it, is also the view of many economists, for whom the transitivity 
of the “all-things-considered better than” relation is an unquestioned, and 
perhaps even self-evident, axiom which needs no argument. I think this view 
is mistaken or, more charitably, deeply misleading.

Since people can use words as they see fit, let me first simply grant that 
there may be a use of the words “all-things-considered better than” such that 
it must be a transitive relation, by definition. So, if Broome or others want 
to insist that as they use the notion of “all-things-considered better than” 
the Axiom of Transitivity is analytic, there is no point in denying or trying 
to refute their claim. But then, let me hasten to add that, as Wittgenstein 
might have put it, meaning is use, and there is another, widely accepted and 
more normatively significant, usage of “all-things-considered better than”, 
what I call the reason-involving sense of “all-things-considered better than”, 
according to which to say that A is better than B, all things considered, is to 
say that from an impartial perspective there is most reason to rank A as more 
desirable than B taking full account of all of the factors that are relevant and 
significant for making that comparison.  5 And, as I shall suggest next, on that 
notion of “all-things-considered better than” —the reason-involving one—
even if it is true that “all-things-considered better than” is a transitive relation, 
it is not an analytic truth, rather, it is a truth that turns on substantive facts 
about the nature and structure of the good.

3.  THE INTERNAL ASPECTS VIEW VERSUS THE ESSENTIALLY 
COMPARATIVE VIEW

To see how the transitivity of the “all-things-considered better than” relation 
in the reason-involving sense turns on substantive facts about the nature 
and structure of the good, it will help to consider two alternative models for 

4. Nagel’s, Scanlon’s, and Parfit’s early views on this topic were conveyed to me during 
discussions when I was a graduate student (for more on this see my Preface in Temkin 2012).

5. Ludwig Wittgenstein’s famous contention that “meaning is use” is defended in 
Wittgenstein 1958.



 Rethinking the Good – A Small Taste 67

LEAP 2 (2014)

thinking about ideals in general, and moral ideals in particular, which I call 
the Internal Aspects View and the Essentially Comparative View.

Here is one natural and plausible way of understanding the Internal 
Aspects View. On this view, how good or bad any given outcome is with respect 
to any given ideal depends solely on the internal features of that outcome. 
Likewise, how good or bad any given outcome is all things considered will 
depend solely on how good or bad it is with respect to each ideal. Now this 
will be a function of how much the different ideals matter relative to each 
other, and it may, in fact, be a very complex function reflecting various 
holistic interaction effects between different ideals, but the key point is 
that on the version of the Internal Aspects View that I am now elucidating, 
ultimately there is a fact of the matter about how good or bad each outcome 
is, and that fact depends solely on the internal features of that outcome and 
the internal relations between them. 

So, on the Internal Aspects View, if one wants to assess how good or bad 
an outcome is, all things considered, it will always be sufficient to consider 
that outcome directly, by itself, in terms of all of the factors or ideals that 
are relevant and significant for assessing the internal features of outcomes. 
Thus, for example, if one believes that equality is relevant to the goodness 
of outcomes, one will consider the extent to which equality or inequality 
is a feature of that outcome, and similarly for other relevant ideals such 
as justice, freedom, utility, perfection, and so on. One will then give each 
outcome its due weight, taking account, as necessary, of any relevant 
interaction effects, in order to arrive at an all things considered judgment 
regarding the outcome’s overall goodness.

The Internal Aspects View allows room for epistemological ignorance 
about how good or bad any given outcome is, as well as room for believing 
that facts about the goodness of outcomes may be indeterminate or 
imprecise, but it is natural to assume that each outcome will have a precise or 
imprecise degree of goodness or badness that can, in principle, be accurately 
represented by a number or range of numbers on the real number line. So, for 
example, in principle it might be a fact that, all things considered, any given 
outcome might have 1013 “units” or “degrees” of goodness or, alternatively, 
perhaps there may be no fact as to precisely how good the outcome is, but 
it might still be true that it has between 1003 and 1023 “units” or “degrees” 
of goodness. For simplicity, in what follows I shall ignore the complication 
introduced by imprecision, and assume that each outcome can be given 
a precise number representing its degree of goodness. But the points I am 
making could also have been made in terms of ranges of numbers for those 
who believe that the degree or extent to which an outcome is good or bad 



68 Larry Temkin

LEAP 2 (2014)

is (often) imprecise, and best captured by a range of numbers rather than a 
single number.  6 

The Internal Aspects View is a natural and plausible way of thinking 
about ideals and their relation to the goodness of outcomes. It also supports 
various views that have been thought central to practical reasoning or the 
assessment of outcomes. For example, it clearly supports the Axiom of 
Transitivity, since if the number representing A’s degree of goodness based 
solely on A’s internal features is higher than the number representing B’s 
degree of goodness based solely on B’s internal features —which will be 
the case if A is better than B— and the number representing B’s degree of 
goodness based solely on B’s internal features is higher than the number 
representing C’s degree of goodness based solely on C’s internal features —
which will be the case if B is better than C— then the number representing A’s 
degree of goodness based solely on A’s internal features will be higher than 
the number representing C’s degree of goodness based solely on C’s internal 
features —since “being a higher number than” is a transitive relation— and 
hence A will be better than C precisely as the Axiom of Transitivity requires. 

The Internal Aspects View also supports another principle which many 
economists and others have regarded as a central principle of practical 
reasoning, which is often called the Independence of Irrelevant Alternatives 
Principle (IIAP). On IIAP, to know how A compares with B it is sufficient to 
compare them directly, as how A or B compares with respect to some third 
alternative, C, or some other set of alternatives C through N, is irrelevant to 
how A compares with B. As we have seen, on the Internal Aspects View, any 
outcome A will get a score representing its degree of goodness and that score 
will be based solely on A’s internal features. And similarly for any outcome B. 
A will be better than, equal to, or worse than B, if and only if its score is higher 
than, equal to, or lower than B’s, respectively. Accordingly, how A compares 
to B in terms of goodness follows directly from how good each of them is, 
considered just by itself, and doesn’t depend at all on how either or both of 
them compares to some third alternative or some other set of alternatives. 
Thus, as indicated, the Internal Aspects View supports, or indeed implies, 
the Independence of Irrelevant Alternatives Principle. 

6. Some people reject the numerical model entirely. For example, in discussion, both 
derek Parfit and Ingmar Persson have conveyed their rejection of any sort of numerical model 
for understanding the good. But while there are problems with any numerical model, I think 
this way of thinking about the Internal Aspects View is natural, plausible, and sufficient for my 
present purposes. I might add that a well-worked-out alternative to such a model has not yet 
been given. Moreover, I am skeptical as to whether a coherent non-numerical model can be 
developed which will capture the most important and attractive features of an Internal Aspects 
View. I briefly touch on this issue at the end of my response to Persson’s article (see Temkin 
2014: 151-52). 



 Rethinking the Good – A Small Taste 69

LEAP 2 (2014)

Let me mention just one other principle of practical reasoning which 
has great plausibility and which is supported by the Internal Aspects View. 
It is plausible to believe that if two alternatives, A and B, are equally good, 
then however A compares to some third alternative C, that is exactly how 
B will compare to C. I call this principle the Principle of Like Comparability 
for Equivalents. It is easy to see how the Principle of Like Comparability for 
Equivalents holds if the Internal Aspects View is correct. On the Internal 
Aspects View, for any three outcomes, A, B, and C, how good A, B, and C are 
will depend solely on their internal features, and each of them will receive a 
score representing its degree of goodness. If A and B are equally good they 
will receive the same score, so clearly however A’s score compares to C’s 
score, that is how B’s score compares to C’s score. 

In sum, the Internal Aspects View has great intuitive plausibility, and it 
would support and explain a number of other widely accepted views about 
practical rationality that many have found compelling, including the Axiom 
of Transitivity, the Independence of Irrelevant Alternatives Principle, and the 
Principle of Like Comparability for Equivalents. The problem is that despite 
its great appeal, the Internal Aspects View doesn’t reflect the thinking that 
many people often engage in when assessing outcomes! In particular, as I 
argued in chapter 12 of Rethinking the Good, many of the ideals people value 
most reflect an Essentially Comparative View of moral ideals. This includes 
especially plausible versions of Utility, Maximin, the Pareto Principle, and the 
Narrow Person-Affecting View.  7 On such views, there is no fact of the matter 
as to how good or bad an outcome is considered just by itself with respect 
to the ideal in question, or if there is, that fact has no special significance in 
comparing outcomes with respect to that ideal. Rather, our assessment of 
how good an outcome is with respect to the ideal in question will depend on 
the alternative or alternatives with which it is compared. More specifically, 
on an Essentially Comparative View of ideals, the relevance and significance 
of the factors for assessing how good an outcome is regarding a particular 
ideal may differ depending on the outcome’s alternative(s), so, in essence, 
a given outcome may have one value regarding an essentially comparative 
ideal given one alternative, but a different value regarding that very same 
ideal given another alternative.

7. Roughly, Utility assesses the goodness of outcomes in terms of how much utility, or 
well-being, the sentient beings in those outcomes have, Maximin assesses the goodness of 
outcomes in terms of how well off the worst-off individuals fare in those outcomes, and the 
Pareto Principle claims that in outcomes involving the same people, one outcome will be better 
than another if it is better for at least one person and at least as good for everyone else. I’ll 
discuss the Narrow Person-Affecting View more later. As stated in the text, in Temkin 2012: 
ch. 12, I argue that in many contexts, the most plausible versions of the ideals in question are 
Essentially Comparative. 



70 Larry Temkin

LEAP 2 (2014)

It follows that if an Essentially Comparative View of moral ideals is 
correct —so, for example, in comparing certain outcomes it is legitimate, as 
many believe, to assess them in terms of essentially comparative versions of 
Utility, Maximin, the Pareto Principle, or a Narrow Person-Affecting View— 
then there is no reason to expect the “all-things-considered better than” 
relation to be transitive. This is because if the relevance and significance of 
the factors for assessing an outcome can vary depending on the alternative 
with which it is compared, then it could well be the case that for any three 
alternatives A, B, and C, A might be better than B in terms of all of the factors 
that are relevant and significant for making that comparison, and B might be 
better than C in terms of all of the factors that are relevant and significant for 
making that comparison, and yet A might not be better than C in terms of all 
of the factors that are relevant and significant for making that comparison. 
After all, it could then well be the case that the factors that are relevant or 
significant for comparing A with C, and which might rightly support the 
judgment that A is not better than C, may differ from the factors that are 
relevant and significant for comparing A with B, or B with C, allowing for the 
real possibility that those factors might rightly support the judgment that A 
is better than B, and B is better than C. 

So, in reflecting on whether or not the Axiom of Transitivity holds, a key 
question is whether the nature and structure of ideals reflects an Internal 
Aspects View of the sort sketched above, or an Essentially Comparative View 
of the sort sketched above. And I submit that the answer to this question 
is a substantive matter determined by the nature of the normative domain, 
it is not a terminological matter determined by the meanings of the words 
“all-things-considered better than”! The words “all-things-considered 
better than” can’t dictate the nature and structure of the normative realm. 
If ideals have the structure embodied by the Internal Aspects View as I have 
characterized it, then, indeed, the Axiom of Transitivity will hold. But if at 
least some ideals have the structure reflected by the Essentially Comparative 
View —as might be the case— then it will not. 

I submit, then, that in the face of seemingly compelling arguments 
that put pressure on the Axiom of Transitivity, we must do the hard 
philosophical work of facing those arguments head on and determining 
which, if any, of their premises should be rejected. We cannot confidently 
reject such arguments on the analytic grounds that the Axiom of Transitivity 
is necessarily true in virtue of the meanings of the words “all-things-
considered better than”. 

In light of the foregoing, let us quickly revisit what appears to be going 
on in section I’s initial Spectrum Argument. The First Standard View reflects 
an additive-aggregationist approach that seems relevant and significant 
for certain comparisons. That is, in comparing the first alternative with the 
second, it seems appropriate to basically multiply the intensity of the pain 



 Rethinking the Good – A Small Taste 71

LEAP 2 (2014)

times its duration, in determining which of the two alternatives is better, 
and this yields the plausible judgment that the first alternative (the slightly 
more intense pain of shorter duration) is better than the second (the slightly 
less intense pain of much longer duration). Similar additive-aggregationist 
reasoning seems appropriate in comparing the second alternative with the 
third, the third with the fourth, the fourth with the fifth, and so on. However, 
the Second Standard View reflects an anti-additive-aggregationist approach 
that seems relevant for other comparisons. In particular, in comparing the 
first alternative with the last, most people don’t simply multiply the intensity 
of the pains times their durations. Rather, they judge that where the difference 
in intensity of the pain is such that the more intense pain of a given duration 
has a significantly adverse effect on one’s life, while the less intense pain of 
much longer duration would have little adverse effect on one’s life, then the 
former would be much worse than the latter, even though the sum total of 
pains as determined by their intensities times durations would be greater in 
the latter situation than the former. So, in essence, most people believe that 
one set of criteria is relevant and significant for assessing how bad the first 
alternative is in comparison with the second, but a different set of criteria 
is relevant and significant for assessing how bad the first alternative is in 
comparison with the last. This reflects an Essentially Comparative View for 
assessing outcomes and, as we have seen, such a view opens up the door to 
rejecting the Axiom of Transitivity. 

In response to my Spectrum Arguments, some total utilitarians and 
economists would reject the anti-additive aggregationist reasoning of the 
Second Standard View, and just insist that as long as there are enough extra 
mosquito bites, the life involving 16 mosquito bites per month is worse than 
the life involving two years of excruciating torture and fifteen mosquito bites 
per month. But is such a view really plausible? 

Here are three related cases where most people would oppose simple 
additive aggregation. Most people firmly believe that derek Parfit’s Repugnant 
Conclusion is, indeed, repugnant (Parfit 1984: ch. 17). They believe that an 
outcome, A, of at least ten billion people, all with a very high quality of life, 
would be better than an outcome, Z, with a vast population all of whom 
have lives that are barely worth living, no matter how many people live in Z. 
Similarly, most firmly believe that an incredibly flourishing human life that 
lasted, say, a million years, would be better for the liver of that life than a 
mere oyster-like existence, no matter how many years one might live in an 
oyster-like state.  8 And likewise, most firmly believe that no matter how many 
people would each get one lick of a lollipop, it would be better for that not 

8. I discuss this kind of example, which I call the Single Life Repugnant Conclusion, in 
Temkin 2012: ch. 4. The Single Life Repugnant Conclusion was originally presented by J. M. E. 
McTaggart (1921: vol. 2, 452-3).



72 Larry Temkin

LEAP 2 (2014)

to occur, if it unavoidably involved an innocent person suffering unbearable 
agony for many years followed by a slow, lonely, miserable death.  9 

Notoriously, total utilitarians reject such claims. Insisting that more utility 
is better than less utility, they offer a number of sophisticated explanations 
for why our intuitions about such cases are not to be trusted. For the total 
utilitarian, then, no matter how small the amount of good may be in a life 
that is barely worth living, or in a moment of oyster-like existence, or how 
small the amount of pleasure may be from one lick of a lollipop, if only there 
are enough such lives, moments, or licks, eventually the total amount of 
good or pleasure will be greater, and then be better, than, any finite amount 
of good or pain that might be balanced off against it. 

The utilitarian’s position is admirably consistent, but it reminds one of 
Emerson’s contention that “a foolish consistency is the hobgoblin of little 
minds, adored by little statesmen and philosophers and divines” (Emerson, 
1983). Few are willing to “bite the utilitarian’s bullet” in such cases, and I 
believe they are right not to do so. In evaluating outcomes, we don’t simply 
care about how much utility obtains, we also care about how that utility is 
distributed and the impact that the distribution has on people’s lives. 

4. NEUTRALITY ANd dOMINANCE PRINCIPLES

It is common for philosophers and others to assume that in certain contexts, 
morality requires us to be neutral with respect to people, places, and times. 
So, for example, setting aside the special obligations that one may have 
towards people with whom one stands in certain special relations —such 
as one’s family, friends, students, patients, and so on— it is thought that, 
other things equal, if one could save one person or five, it would be better 
to save the five whether the five were (a) black or white, rich or poor, Hindu 
or non-Hindu, men or women, European or African, and so on (neutrality 
with respect to people), (b) close or far (neutrality with respect to space), 
or (c) living in the present, the near future, or the distant future (neutrality 
with respect to time —we’d also think it wouldn’t matter if the five were 
living in the past if, contrary to fact, we could save people who were living 
in the past).

Now I am aware that certain prevalent theories of modern physics discuss 
the space/time continuum in a way that suggests that space and time are not 
really distinguishable, so that however we treat space we should also treat 
time, and vice versa. But despite this, I have my doubts whether space and 

9. My Lollipops for Life case is presented in Temkin 2012: ch. 2. That case serves as the 
inspiration for the book’s cover art. 



 Rethinking the Good – A Small Taste 73

LEAP 2 (2014)

time should, in fact, be treated the same normatively. Consider, for example, 
the following thought experiment. 

Suppose I learn that our civilization will live in our galaxy another 1000 
years, and then die out. I also learn that in a distant galaxy another advanced 
civilization will exist for the same 1000 years and then die out, and that this 
is also so in a third distant galaxy, and a fourth distant galaxy. I find this all 
quite interesting. It is somewhat pleasing to me to learn that there are, in 
fact, advanced civilizations living in galaxies far away. But suppose I also 
learn that beyond the fourth galaxy there is nothing but cold, empty, space. 
This, too, I find interesting, but I must confess that learning that fact doesn’t 
bother me at all. Indeed, if someone said that events beyond the fourth galaxy 
were about to unfold which would make those distant reaches inhospitable 
to life forms in perpetuity, I wouldn’t think it important for our civilization to 
make significant sacrifices, if it could, to prevent that from happening.

Suppose, on the other hand, I vary the story a bit. As before, I learn that 
civilization in our galaxy will die out in 1000 years; but I learn that after ours 
dies out another advanced civilization will arise and persist for 1000 years 
in a second galaxy, and that this will happen again a third and fourth time. 
But I also learn that after the fourth civilization dies out there will be nothing 
but cold, empty, space, forever. For some reason, that knowledge would 
bother me a lot. Indeed, if I learned that events were about to unfold which 
would make the universe uninhabitable for any life forms 4000 years from 
now, unless our civilization made significant sacrifices to prevent that from 
happening, I would feel quite strongly that we should do so, and I would feel 
that way even if I knew that our civilization was going to die out in 1000 years 
no matter what we did. 

My views here may ultimately be indefensible, but I don’t think they 
are idiosyncratic. They reveal an asymmetry in my thinking about space 
and time. I think it very important that many periods of time are filled with 
flourishing sentient beings. I think it much less important that many areas 
of space are filled with flourishing beings. 

There is much more to be said about this suggested asymmetry between 
space and time, but I shall not pursue this here. Instead, let me turn to 
another set of views that might be held regarding space, time, and people. 
At first blush, I think most people would readily accept the following three 
dominance principles: (1) if outcome A is better than outcome B at every 
point in space, then A is better than B; (2) if outcome A is better than outcome 
B at every moment in time, then A is better than B; and (3) if outcome A and 
outcome B involve the very same people, and A is better than B for every 
person, then A is better than B. 

1, 2, and 3 are exceedingly weak Pareto-like principles. According to the 
Pareto Principle, if two outcomes involve the same people, and the first 
outcome is better for at least one person and at least as good for everyone 



74 Larry Temkin

LEAP 2 (2014)

else, then the first outcome must be better than the second. 1 and 2 apply 
similar reasoning to the domains of space and time, respectively, as to the 
domain of people. In addition, 1, 2, and 3 require that the first outcome be 
better than the second at every point in space, at every moment in time, or 
for every person, respectively. 

Given the widespread appeal of the Pareto Principle, the fact that the 
dominance principles noted above are much weaker —and are therefore 
even more plausible— than the standard Pareto Principle, and the common 
assumption that we should be neutral with respect to people, places, 
and times, I think it is fair to assume that most people would find each of 
the three dominance principles intuitively appealing. Indeed, I suspect 
that many people would think that each of the dominance principles is 
“obviously” true. Yet, it is easy to see that however intuitively appealing the 
three dominance principles may be, at least one of them must be rejected.

Consider diagram One. 

day 1 P
1
 Hell day 1 P

1
 Heaven

day 2 P
1
 Heaven; P

2
, P

3
 Hell day 2 P

1
 Hell; P

2
, P

3
 Heaven

day 3 P
1-3 

Heaven; P
4-9 

Hell day 3 P
1-3 

Hell; P
4-9 

Heaven
day 4 P

1-9
 Heaven; P

10-27 
Hell day 4 P

1-9
 Hell; P

10-27 
Heaven

 : :
 : :
 W

1
 W

2

Diagram One

diagram One represents two possible worlds God is thinking of instantiating, 
W

1
 and W

2
. In W

1
, there will be a single person, P

1
, who will exist on day 1, 

and he will be in Hell. We don’t have to think that P
1
’s life will be infinitely 

bad, we just have to think that it will be very bad. during the course of that 
day, it would be much better for P

1
 if he were not alive. On day 2, P

1
 moves 

to Heaven, where it will be very good for P
1
 that he is alive. For simplicity, 

let us assume that each day in Heaven would be as good for the person 
experiencing it as a day in Hell would be bad for a person experiencing it, 
so that on balance the net value of a life with an equal number of days in 
Heaven and in Hell would be zero. Unfortunately, on day 2 two new people, 
P

2
 and P

3
 are created and put in Hell. On day 3, each of P

1
-P

3
 are in Heaven, 

but six new people P
4
-P

9
 are in Hell. On day 4 each of P

1
-P

9
 are in Heaven, but 

18 new people are created in Hell. And so on. 
W

2
 is just like W

1
 except in reverse. In W

2
, P

1
 will again exist on day 1, but 

this time he will start in Heaven. On day 2, P
1
 moves to Hell, but two new 

people, P
2
 and P

3
 are created and put in Heaven. On day 3, each of P

1
-P

3
 are 



 Rethinking the Good – A Small Taste 75

LEAP 2 (2014)

in Hell, but six new people P
4
-P

9
 are in Heaven. On day 4 each of P

1
-P

9
 are in 

Hell, but 18 new people are created in Heaven. And so on. 
How do W

1
 and W

2
 compare in terms of goodness? Which, if either, is 

the better outcome, all things considered? If one looks at the two outcomes 
day by day, it may seem clear that W

2
 is better than W

1
. After all, on day 1, 

there would be one person in Hell in W
1
 and one person in Heaven in W

2
. 

So, on day 1, W
1
 is clearly worse than W

2
. Similarly, on day 2, W

1
 would have 

one person in Heaven, but two people in Hell, whereas W
2
 would have one 

person in Hell, but two people in Heaven. Given our views about neutrality 
with respect to people, it seems clear that it is worse for there to be twice as 
many people in Hell as in Heaven, than it is for there to be twice as many 
people in Heaven as in Hell, so W

1
 is worse than W

2
 on day 2. Similarly, on 

day 3, W
1
, where there are three people in Heaven but six people in Hell, 

will be worse than W
2
, where there are three people in Hell, but six people 

in Heaven. And so on. The point is that on day 1, W
1
 is worse than W

2
, and 

that on each day after that W
1
 is worse than W

2
, since, on each day after day 

1, there will always be twice as many people in Hell as in Heaven in W
1
, while 

there will always be twice as many people in Heaven as in Hell in W
2
. Thus, 

comparing W
1
 and W

2
 day by day, or moment by moment, the dominance 

principle with respect to time would entail that W
2
 is better than W

1
. 

Is W
2
 better than W

1
? I find that very hard to believe. Suppose we compare 

the two outcomes not moment by moment, but person by person. In W
1
, 

each person spends exactly one day in Hell, and the rest of eternity in Heaven. 
In W

2
, each person spends exactly one day in Heaven, and the rest of eternity 

in Hell. I know which of these worlds I would want for myself, a loved one, 
or anyone else who was not pure evil! I would want W

1
, and I would want it 

because it would be better for each person who ever lived. Notice, since in 
this example we are assuming that the very same people would live in each 
world, and we know that each of them would be better off in W

1
 than W

2
 

(indeed vastly so, since it is much better to spend only one day in Hell and 
the rest of eternity in Heaven, than to spend only one day in Heaven and the 
rest of eternity in Hell), then the dominance principle with respect to people 
would entail that W

1
 is better than W

2
.

In this example, we see that two intuitively plausible and seemingly 
“obvious” dominance principles are in fact incompatible. In this case, at 
least, we must choose between the dominance principle with respect to 
time and the dominance principle with respect to people. As I have already 
made clear, I know how I would choose in this case. I think W

1
 is clearly and 

unequivocally better than W
2
. 

Notice, if one adopted a purely impersonal view of morality, according 
to which it didn’t matter how any particular sentient beings fared, or how 
benefits or burdens were distributed within or between lives, but it only 
mattered how many benefits or burdens obtained in an outcome, then it 



76 Larry Temkin

LEAP 2 (2014)

might be plausible to maintain that W
2
 is better than W

1
, in accordance 

with the dominance principle with respect to time, or, alternatively, that W
1
 

and W
2
 were equally good, since each would ultimately involve an infinite 

number of days lived in both Heaven and Hell of the same orders of infinity. 
But my own view is that one lesson to be learned from diagram One is that 
in assessing the goodness of outcomes we should not merely focus on the 
impersonal questions of how much well-being there is in the two outcomes, 
or how many benefits and burdens obtain in total. Rather, in some cases, at 
least, we must focus on the question of how the well-being or benefits and 
burdens are distributed, and, in particular, on how the sentient beings are 
affected for better or worse in those outcomes. 

5. ESSENTIALLY COMPARATIVE IdEALS

I claimed earlier that a number of ideals people attach great value to have an 
Essentially Comparative structure, including the Pareto Principle, the most 
plausible versions of Maximin and Utility, and the Narrow Person-Affecting 
Principle. I defend this claim in Temkin (2012: ch. 12) for each of the ideals in 
question, but for the purposes of this article let me just focus on the Narrow 
Person-Affecting Principle.

In any choice situation between possible outcomes, let us call those 
people who do exist, or have existed, or will exist in each of the outcomes 
independently of one’s choices, independently existing people. By contrast, 
let us call those people whose existence in one or more possible outcomes 
depends on the choices one makes in bringing about an outcome, 
dependently existing people. Bearing these distinctions in mind, we can now 
state the Narrow Person-Affecting View.

The Narrow Person-Affecting View: In assessing possible outcomes, one 
should (1) focus on the status of independently existing people, with 
the aim of wanting them to be as well off as possible, and (2) ignore the 
status of dependently existing people, except that one wants to avoid 
harming them as much as possible. Regarding clause 2, a dependently 
existing person is harmed only if there is at least one available alternative 
outcome in which that very same person exists and is better off, and the 
size of the harm will be a function of the extent to which that person 
would have been better off in the available alternative outcome in which 
he exists and is best off.  10

10. derek Parfit presented a position which he also called a Narrow Person-Affecting 
View in Parfit 1984: ch. 18. The view as I present it here is different in important respects than 
Parfit’s, but I have retained the name Parfit uses, because I think the view I have described 
reflects a fundamental approach to assessing outcomes that is best described as a Narrow 
Person-Affecting View. I believe that my version of the Narrow Person-Affecting View is 



 Rethinking the Good – A Small Taste 77

LEAP 2 (2014)

As stated, the Narrow Person-Affecting View reflects an important extension 
of Jan Narveson’s claim that “Morality has to do with how we treat whatever 
people there are…. [We] do not … think that happiness is impersonally good. 
We are in favor of making people happy, but neutral about making happy 
people” (Narveson 1973: 73 and 80). Specifically, the first clause reflects the 
view that we are neutral about making people exist, while the second clause 
reflects the important qualification that if we are going to make a particular 
person exist, her interests have to count the same way as every other existing 
person’s, in that we must equally seek to make that person, like every other 
existing person, as well off as possible.

Now, in fact, that there are lots of ways in which the Narrow Person-
Affecting View needs to be qualified and limited in scope, which I won’t go 
into here (Temkin 2012: ch. 12.3). Nevertheless, when properly interpreted, 
the Narrow Person-Affecting View reflects a deeply plausible and widely-
accepted view for a certain range of cases. 

To illustrate the Narrow Person-Affecting View, it will be useful to 
consider a range of cases to which it might be applied, and to contrast it with 
some other principles that might be appealed to in assessing outcomes: the 
Impersonal Total View, the Impersonal Average View, and the Wide Person-
Affecting View. Roughly, we might say that the Narrow Person-Affecting 
View assesses outcomes by considering how the particular people in those 
outcomes fare, relative to how they fare in any available alternative outcomes 
(here, and below, “people” refers to any sentient beings). The aim is to make 
sure that each particular person who does, or will, exist independently of our 
choices, or who will exist as a result of our choices, fares as well as possible. In 
contrast, the Wide Person-Affecting View assesses outcomes by considering 
how the people in those outcomes fare, but it is not concerned with how 
any particular people fare in one outcome relative to how those very same 
people might fare in any available outcomes.  11 A precise characterization 
of the Wide Person-Affecting View is elusive, but one natural and plausible 
way of interpreting it implies, among other things, that if the people in one 
outcome, A, are all better off than the people in another outcome, B, whether 
or not they are the same people or there are the same number of people, then 
A is better than B; if, for each distinct person in B, there is corresponding 

more plausible than Parfit’s original version, and in conversation Parfit has indicated that 
he agrees. 

11. The notion of a Wide Person-Affecting View was introduced by Parfit 1984: ch. 18. 
Unfortunately, as Parfit originally presented the position, he combined two elements which are 
best kept distinct. The first reflects the view that in assessing outcomes we want to assess them 
in terms of the extent to which the people (sentient beings) in those outcomes are affected for 
better or worse. The second concerns the very distinct question of whether causing someone 
to exist benefits that person. I use the notion of a Wide Person-Affecting View to reflect the first 
element only. Parfit now shares my view (Temkin 2012: note 41, section 12.4).



78 Larry Temkin

LEAP 2 (2014)

distinct person in A, at least one of whom is better off and the rest of whom 
are at least as well off, then A is better than B as long as anyone else existing 
in A has a life that is (sufficiently) worth living; and if A and B have the same 
number of people, and for each person in B there is a corresponding person 
in A who is equally well off, and vice versa, then A and B are equally good. 
Finally, the Impersonal Total and Average Views imply that regardless of 
whether or not they have the same people or the same number of people, 
one outcome will be better than (equal to) another if and only if the one 
outcome has a higher (the same) total or average amount of utility or well-
being, respectively. 

Consider diagram Two.

1000

 P
1 

P
1 

P
1 

P
2 

P
1 

P
3 

P
1 

P
4 

P
1 

P
4

 A A A B A C A C A d

 I II           III                IV               V                VI

Diagram Two

1200
1100 1100

800

1100

800

600

In I, there is a large population, A, say of 10 billion people, on a given 
planet, P

1
, all of whose members are at level 1000. Assume that I is the 

initial outcome, and that the A people are thinking about transforming their 
outcome into one represented by II. In II, those very same people all exist 
and are better off, at level 1200. II would be judged a better outcome than I 
on all of the different approaches for assessing alternatives. Specifically, II is 
better than I on the Impersonal Total and Average Views, since the total and 
average amounts of wellbeing are greater in II than in I, on the Wide Person-
Affecting View, since it is better for people, as everyone in II is better off than 
everyone in I, and on the Narrow Person-Affecting View, since it is better for 
the particular, independently existing A people who exist in both outcomes. 

Suppose, instead, that the A people could transform I into an outcome 
like III. In III, the A people have all been lowered to level 600, but a new 
population of 10 billion people, B, would also come to exist at level 600 on a 
second planet, P

2
. In this scenario, III would be ranked better than I on the 

Impersonal Total View, since the total wellbeing would be greater in III than 
in I. But III would be ranked worse than I on the Impersonal Average View, 



 Rethinking the Good – A Small Taste 79

LEAP 2 (2014)

since the average level of wellbeing would be less in III than in II. III would 
also be ranked worse than I on the Wide Person-Affecting View, since the 
people in I are better off than the people in II. Finally, III would also be ranked 
worse than I on the Narrow Person-Affecting View, as the independently 
existing A people are better off in I than in III, and the principal aim of the 
Narrow Person-Affecting View is to make the particular existing people as 
well off as possible (making people happy) rather than to add more people 
to an already large and well-off population (making happy people). 

While total utilitarians would rank III better than I, if outcome I were 
one’s starting point, many people, and perhaps most, would rank I better 
than III, and they might do so on any combination of the grounds suggested.

Suppose next that the people in I could bring about IV. IV involves a new 
group of 10 billion people, C, living on a different planet, P

3
. Unfortunately, 

the conditions on P
3
 are not quite as favorable as those on P

1
, so the C people 

would only be at level 800. But we may presume that level 800 is still quite 
high, so that everyone on P

3
 would have lives well worth living. In addition, 

there might be resources on P
3
 which could be used in trades with those on 

P
1
, so that everyone in P

1
 would be raised up to level 1100. 

IV would be worse than I on the Impersonal Average View. Many find this 
hard to believe. If there is an objection to IV, it would seem to rest on the 
fact that IV involves inequality while I is perfectly equal, not on the fact that 
the average level of well-being is lower in IV than in I. After all, IV is better 
off than I for everyone who lives in I, and in addition IV involves a very large 
group of people all of whom have lives that are well worth living.  12 

On reflection, I believe most people would judge IV better than I, and this 
would be supported by the Impersonal Total View —since the total wellbeing 
is greater in IV than in I— by the Wide Person-Affecting View —since IV is 
better for people than I, as for each person in I there is a corresponding 
person in IV who is even better off, and any additional people in IV have lives 
that are well worth living— and by the Narrow Person-Affecting View, since 
the particular independently existing A people are better off in IV (being at 
level 1100) than in I (being at level 1000). 

Next, suppose that the people in outcome I could bring about either IV 
or V. In V, the A people have to make extra sacrifices to enable the C people 
to live on a different, more hospitable, fourth planet P

4
. The result would be 

12. The strongest arguments against the average view involve alternatives where people’s 
lives are well below the level at which life ceases to be worth living. Surely, one wouldn’t improve 
an outcome where billions of people were living in the worst hell imaginable in any respect, 
merely by adding billions of more people whose lives were almost, but not quite, as badly off. 
But, of course, the addition of all those extra people living hellish lives would raise the average 
level, even if only by a small amount. For further discussion of this kind of case, which Parfit 
called Hell Three, and other reasons to be skeptical of average views, see Parfit 1984: 422; Temkin 
2012: section 10.4; Temkin 1993: section 7.5. 



80 Larry Temkin

LEAP 2 (2014)

that the C people would be at level 1100, but the A people would only be at 
level 800. Interestingly, as alternatives to I, IV and V would likely be regarded 
as equally good on all four of the principles we have been discussing. IV and 
V are equally good on the Impersonal Total and Average Views, as they are 
equally good in terms of total and average wellbeing. They are equally good 
on the Wide Person-Affecting View, since in terms of how people in those 
outcomes fare (rather than in terms of how the particular people fare in one 
outcome rather than another), they are equally good for people. Finally, they 
are equally good on the Narrow Person-Affecting View, since on that view 
one doesn’t have to bring about the dependently existing C group, but if one 
is going to bring a particular group into existence —and, by hypothesis, the 
very same C people would be brought into existence in both IV and V— then 
their interests have to be given the same weight as those of the independently 
existing people, A. Hence, on the Narrow Person-Affecting View, there would 
be nothing to choose between outcome IV, where the independently existing 
A people would be at level 1100 and the dependently existing C people would 
be at level 800, and outcome V, where the independently existing A people 
would be at level 800 and the dependently existing C people would be at 
level 1100. 

Finally, suppose that the option facing those in I is not IV or V, but IV or VI. 
Here, the option is between populating planet P

3
 with 10 billion people, the C 

people, who would all be well off, but “only” at level 800, but where this would 
enable the A people to raise their level from 1000 to 1100, or populating a more 
hospitable but more distant planet P

4
, with an entirely different group of 10 

billion people, the d people, but where the cost of populating the more distant 
planet would be to lower the level of the A people from 1000 to 800. IV and VI 
would be equally good on both the Impersonal Total and Average Views, since 
the total and average levels of wellbeing are equal in both outcomes. Likewise, 
IV and VI , would be equally good on the Wide Person-Affecting View, since, 
overall, people fare equally well in both outcomes. However, importantly, if 
one’s initial starting place was I, then IV would be decidedly better than VI on 
the Narrow Person-Affecting View. This is because, insofar as we are concerned 
with “making people happy, rather than making happy people”, IV is a clear 
improvement, while VI is a clear worsening of the outcome. 

That is, on the Narrow Person-Affecting View, IV is better for the 
independently existing A people (they are at level 1100 rather than level 
1000), and it in no way harms the dependently existing C people, since their 
lives are well worth living, and, in this choice situation, there is no available 
alternative in which they would be better off. VI, on the other hand, is clearly 
worse for the independently existing A people (they are at level 800 rather 
than level 1000), and this worsening of the outcome cannot be made up for 
by the neutral factor of adding extra “happy” d people. 



 Rethinking the Good – A Small Taste 81

LEAP 2 (2014)

Let me acknowledge that the Narrow Person-Affecting View is not plausible 
in cases like Parfit’s Non-Identity Problem (Parfit 1984: ch. 16) As stated, it is 
also implausible in a host of other cases, many of which will readily occur to 
the reader. However, despite this, I believe that the Narrow Person-Affecting 
View is plausible, and relevant and significant for comparing outcomes in 
a large range of cases, including those just discussed. Thus, in considering 
cases like those represented in diagram Two, I believe that many people 
would judge that if one’s initial outcome were like I, then II would be better 
than I, III would be worse than II, IV would be better than I, IV and V would 
be equally good, and IV would be better than VI, and I believe that many 
would base their judgments partly, if not wholly, on narrow person-affecting 
grounds (or a position very much like it in spirit if not exact detail).

As should be clear, the Narrow Person-Affecting View is an Essentially 
Comparative Ideal. On such a view, assessing how good an outcome is depends 
not solely on its internal features, as is the case on the Internal Aspects View, 
but on whether the particular people in that outcome exist in available 
alternative outcomes, and if so, on how they fare in the available alternatives. 

Assuming that there would be no morally relevant differences between 
the different people in my examples other than narrow person-affecting 
considerations, on the Internal Aspect View IV, V, and VI would be equally 
good, since their internal features are identical, except for which particular 
people exist in which outcomes and which particular levels they are at. Hence, 
in accordance with the Principle of Like Comparability for Equivalents, on 
the Internal Aspects View, however one of them compared with some other 
alternative, that is how each of them would compare with that alternative, 
and this would be so regardless of whether or not any other outcomes were 
available. But, as we have seen, in accordance with the Narrow Person-
Affecting View, many people would judge IV as better than I, if outcome I was 
the initial starting point and those were the only alternatives, but they would 
judge V as worse than I, if outcome I was the initial starting point and those 
were the only alternatives. Similarly, in accordance with the Narrow Person-
Affecting View, many people would judge IV as equally as good as V, if those 
were the only alternatives, and V as equally as good as VI, if those were the 
only alternatives, but, contrary to both the Principle of Like Comparability 
for Equivalents and the Axiom of Transitivity for Equally as Good As (each 
of which is entailed by the Internal Aspects View), they would deny that VI 
is equally as good as IV. Likewise, in accordance with the Narrow Person-
Affecting View, and contrary to the Axiom of Transitivity for Better Than, it is 
plausible to contend that if outcome I were one’s initial starting point, then 
IV would be better than I if those were the only alternatives, and I would be 
better than V if those were the only alternatives, but IV would not be better 
than V if those were the only alternatives. 



82 Larry Temkin

LEAP 2 (2014)

Could we abandon the Narrow Person-Affecting View and simply adopt 
Impartial Views or the Wide Person-Affecting View instead? Not without 
abandoning a view that underlies many judgments people make in assessing 
outcomes. And not easily. To buttress this claim, let us consider two further 
cases, of a different sort, the first of which is exemplified by diagram Three. 

–1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11 –12 –13 –14 –15 –16 –17 –18 –19 –20 →

T
1

T
2

T
3

T
4

T
5

T
6

T
7

T
8

T
9

T
10

 T
11

T
12

T
13

T
14

T
15

T
16

T
17

T
18

T
19

T
20

P
1

P
2

P
3

P
4

P
5

P
6

P
7

P
8

P
9

P
10

P
11

P
12

P
13

P
14

P
15

P
16

P
17

P
18

P
19

P
20

O
1

–11 –12 –13 –14 –15 –16 –17 –18 –19 –20 –21 –22 –23 –24 –25 –26 –27 –28 –29 –30 →

T
1

T
2

T
3

T
4

T
5

T
6

T
7

T
8

T
9

T
10

T
11

T
12

T
13

T
14

T
15

T
16

T
17

T
18

T
19

T
20

P
1

P
2

P
3

P
4

P
5

P
6

P
7

P
8

P
9

P
10

P
11

P
12

P
13

P
14

P
15

P
16

P
17

P
18

P
19

P
20

 

O
2

Diagram Three

Suppose that one of two outcomes was going to come about. In O
1
, there 

would be one person living on planet one, P
1
, at time one, T

1
, and that person 

would be at level –1, which is below the level at which life ceases to be worth 
living. It would be better for that person if he or she never existed. There would 
also be one person living on planet two, P

2
, at time two, T

2
, and that person 

would be even worse off at level –2. There would be a third person living on 
planet three, P

3
, at time three, T

3
, and that person would be even worse off at 

level –3, and so on. Hence, there would be an infinite number of people living on 
different planets and at different times, and each person, after the first, would 
be worse off than those that preceded him or her. In addition, let us assume 
there would be no other morally relevant factors or events obtaining in W

1
.

In the second outcome, O
2
, there would again be one person living on 

planet one, P
1
, at time one, T

1
, but this time the person would be at level –11. 

There would also be one person living on planet two, P
2
, at time two, T

2
, and 

that person would be even worse off at level –12. There would be a third person 
living on planet three, P

3
, at time three, T

3
, and that person would be even worse 

off at level –13, and so on. As before, there would be an infinite number of 
people living on different planets and at different times, and each person, after 
the first, would be worse off than those that preceded him or her, and there 
would be no other morally relevant factors or events obtaining in O

2
. Finally, for 

any level –n, it is worse for someone to be at level –(n – 10), than to be at level –n. 



 Rethinking the Good – A Small Taste 83

LEAP 2 (2014)

How do O
1
 and O

2
 compare? As described, there might be some reasons 

associated with how we think about cases involving infinity, for claiming that 
O

1
 and O

2
 were equally good. On the other hand, I think there would also be 

powerful reasons for thinking that O
1
 was better than O

2
. If we, or God, had 

to choose which of the two outcomes to produce, or we learned that one of 
the two outcomes was going to be instantiated, at first blush it seems that 
we should produce or hope that it is O

1
 rather than O

2
. Other things equal, 

it seems we should be confident that O
1
 would be at least at good as (and 

probably better than) O
2
. 

The preceding ranking of O
1
 and O

2
 would be supported by both Impersonal 

Principles, as well as any plausible Wide Person-Affecting View. Insofar as one 
merely focuses on the impersonal value in each outcome, or on how people 
fare in each outcome without regard to how any particular people fare, then 
it seems clear that O

1
 is at least at good as O

2
. Moreover, I think O

1
 would be 

at least at good as O
2
 if completely different people lived in O

1
 than in O

2, 
or if 

anyone who lived in both outcomes, lived on the same corresponding planet 
and at the same corresponding time in both outcomes, such that if a given 
person, John, lived in both outcomes, then whatever planet P

n
 and time T

n
 

that he occupied in O
1
, he would also occupy P

n
 and T

n
 in O

2
. 

Suppose, however, that I now tell a different story regarding the members 
of O

1
 and O

2
. Suppose it is true that every person who would exist in O

2
, if O

2
 

obtained, would also exist in O
1
, if O

1
 obtained, but that each of them would 

be worse off in O
1
 than in O

2
. Specifically, let us assume that the very same 

individual, I
1
, who would exist at T

1
 in O

2
, would exist at T

21
 in O

1
, that the very 

same individual, I
2
, who would exist at T

2
 in O

2
, would exist at T

22
 in O

1
, that the 

very same individual, I
3
, who would exist at T

3
 in O

2
, would exist at T

23
 in O

1
, and 

so on. It would then be the case that every single person who would exist in O
2
 

would also exist in O
1
 and would be ten units worse off, where, as before, for any 

level –n, it is worse for someone to be at level –(n – 10), than to be at level –n.
Given that scenario, it seems clear that if we, or God, had to choose which 

of the two outcomes to produce, or we learned that one of the two outcomes 
was going to obtain, we should produce or hope that it is O

2
 rather than O

1
! O

1
 

is worse than O
2
 for every person who lives in O

2
, and, in addition, there are 20 

different individuals who exist in O
1
 but not in O

2
 (those who would be living at 

times T
1
 through T

20
 in O

1
), whose lives are below the zero level —they would 

rationally prefer that they had never been born. Surely, if we were aiming to 
choose the better outcome, and we knew that we or our loved ones might 
actually be occupants of one of the two worlds, we would choose O

2
, and we 

would make a similar choice on behalf of any strangers who were not pure evil. 
It seems clear, then, that our judgments about how outcomes like O

1
 and 

O
2
 compare would not, and should not, be influenced solely by impersonal 

or wide person-affecting considerations. In some cases, how the particular 
people are affected for better or worse depending on the alternatives is rightly 



84 Larry Temkin

LEAP 2 (2014)

relevant to our assessment, as is implied by the Narrow Person-Affecting 
View. Thus, in some cases at least, cross-world identification of particular 
individuals is both relevant and necessary for accurately comparing 
outcomes, as is permitted on the Essentially Comparative View of ideals, but 
is prohibited by the Internal Aspects View. 

Let us apply the preceding reasoning to a final case, represented by 
diagram Four. 

–1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11 –12 –13 –14 –15 –16 –17 –18 –19 –20 →

T
1

T
2

T
3

T
4

T
5

T
6

T
7

T
8

T
9

T
10

T
11

T
12

T
13

T
14

T
15

T
16

T
17

T
18

T
19

T
20

P
1

P
2

P
3

P
4

P
5

P
6

P
7

P
8

P
9

P
10

P
11

P
12

P
13

P
14

P
15

P
16

P
17

P
18

P
19

P
20

O
3

–1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11 –12 –13 –14 –15 –16 –17 –18 –19 –20 →

T
1

T
2

T
3

T
4

T
5

T
6

T
7

T
8

T
9

T
10

T
11

T
12

T
13

T
14

T
15

T
16

T
17

T
18

T
19

T
20

P
1

P
2

P
3

P
4

P
5

P
6

P
7

P
8

P
9

P
10

P
11

P
12

P
13

P
14

P
15

P
16

P
17

P
18

P
19

P
20

O
4

–1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11 –12 –13 –14 –15 –16 –17 –18 –19 –20 →

T
1

T
2

T
3

T
4

T
5

T
6

T
7

T
8

T
9

T
10

T
11

T
12

T
13

T
14

T
15

T
16

T
17

T
18

T
19

T
20

P
1

P
2

P
3

P
4

P
5

P
6

P
7

P
8

P
9

P
10

P
11

P
12

P
13

P
14

P
15

P
16

P
17

P
18

P
19

P
20

O
5

Diagram Four

O
3
, O

4
, and O

5
 are just like O

1
 in diagram Three. In each outcome there is 

one person on P
1
 at T

1
 at level –1, a second person on P

2
 at T

2
 at level –2, a 

third person on P
3
 at T

3
 at level –3, and so on. If one asked how O

3
, O

4
, and O

5 
compared, it would be natural to assume that they were all equally good, all 
things considered. And if there were different people in O

3
, O

4
, and O

5
, then 

it seems clear that they would all be equally good. 
Suppose, then, we make the assumption that the people in O

3
 would be 

different people than those in O
4
, and similarly that the people in O

3
 would 

be different people than those in O
5
. In that case, there would be no narrow 

person-affecting considerations that were relevant for comparing O
3
 with 

O
4
, or for comparing O

3
 with O

5
, and there would be good grounds for 

judging that O
3
 and O

4 
were equally good, and similarly that O

3
 and O

5 
were 

equally good. does it follow from this that O
4
 and O

5 
must be equally good, 

as it must if the Internal Aspects View is correct, since such a view entails 
both the Principle of Like Comparability for Equivalents and the transitivity 
of the “equally as good as” relation? It does not! Because consistent with 
the forgoing relations between O

3
 and O

4
, and O

3
 and O

5
, O

5 
may stand in a 



 Rethinking the Good – A Small Taste 85

LEAP 2 (2014)

different relation to
 
O

4
, one that is similar to the relation in which O

2
 stood to 

O
1 

in diagram Three. 
After all, even it is true that the people in O

3
 are different from the people 

in both O
4
 and O

5
, it doesn’t follow from that that the people in O

5
 are different 

from the people in O
4
. They may not be! Suppose, then, that the person who 

would occupy P
1
 and T

1
 and be at level –1 in O

5
, would occupy P

11
 and T

11
 and 

be at level –11 in O
4
, the person who would occupy P

2
 and T

2
 and be at level 

–2 in O
5
, would occupy P

12
 and T

12
 and be at level –12 in O

4
, the person who 

would occupy P
3
 and T

3
 and be at level –3 in O

5
, would occupy P

13
 and T

13
 and 

be at level –13 in O
4
, and so on. It would then be the case that everyone who 

exists in O
5
 also exists in O

4 
and is ten units worse off, and that, in addition, 

there would be 10 different individuals who exist in O
4
, but not in O

5
, whose 

lives would be below the zero level and who would rationally wish that they 
had never been born. In this case, as above, it seems clear that O

4
 would be 

a worse outcome than O
5
, and mainly in virtue of narrow person-affecting 

considerations.
We see, then, that in accordance with the Essentially Comparative View, a 

factor that is relevant and significant for comparing O
4
 and O

5
 —specifically, 

the fact that everyone who exists in O
5
 also exists in O

4
 where he or she is 

worse off— is not relevant or significant for comparing O
3
 with O

4
, or O

3
 with 

O
5
. This explains how it can be the case that in terms of all of the factors that 

are relevant and significant for making each comparison, O
3
 and O

4
 might be 

equally good, and O
3
 and O

5
 might be equally good, but O

4
 and O

5
 might not 

be equally good.
More generally, as we have seen, once we accept an Essentially 

Comparative View of ideals, as it seems we must if we are to account for the 
judgments to which many are committed regarding diagrams Two, Three, 
and Four, then there is no reason to expect the “all-things-considered better 
than” or “all-things-considered equally as good as” relations to be transitive, 
or, alternatively, no reason to think that such relations even apply to various 
alternatives we may have expected them to for the purposes of practical 
reasoning.  13 

13. In my book, I discuss various ways of preserving the Axioms of Transitivity in the face 
of my arguments, which have the implication that there is no single set of alternatives that are 
being compared in the cases I discuss, or no single relation that the different alternatives are 
being compared in terms of, so that there is, strictly speaking, no violation of the axioms of 
transitivity in the cases I discuss, rather those axioms don’t even apply to the cases I consider. 
I suggest that even if such a move can be plausibly defended, it has significant practical and 
theoretical difficulties akin to those that would accompany the rejection of the Axioms of 
Transitivity (see Temkin 2012: ch. 13). 



86 Larry Temkin

LEAP 2 (2014)

6. CONCLUdING REMARk

As promised at the beginning, this article barely scratches the surface 
of some of the issues raised in Rethinking the Good. Moreover, the further 
one explores such issues, the more one realizes how this domain is fraught 
with complications, unresolved difficulties, and impossibility results whose 
premises are exceedingly difficult to abandon. The book seriously challenges 
us to rethink our understanding of the good, moral ideals, and the nature of 
practical reasoning in many ways that have deep practical and theoretical 
implications. But beyond that, I’m afraid, it offers little guidance, and I have 
little sense, of where we go from here. I wish it were otherwise.

BIBLIOGRAPHY

Broome, J., 1991: Weighing Goods, Oxford: Basil Blackwell.
— 2004: Weighing Lives, Oxford: Oxford University Press. 
Emerson, R. W., 1983: “Self- Reliance”, in Essays and Lectures, ed. Joel Porte, New York 

Library of America: 259-82.
McTaggart, J. M. E., 1921: The Nature of Existence, 2 vols., Cambridge: Cambridge 

University Press.
Narveson, J., 1973: “Moral Problems of Population”, The Monist 57.
Parfit, d., 1984: Reasons and Persons, Oxford: Oxford University Press.
Rachels, S., 1993: “A Theory of Beneficence”, unpublished Philosophy, Politics and 

Economics thesis, Oxford University.
— 1998: “Counterexamples to the Transitivity of Better Than”, Australian Journal of 

Philosophy 76: 71-83.
— 2001: “A Set of Solutions to Parfit’s Problems”, Noûs 35: 214-238.
— 2004: “Repugnance or Intransitivity”, in The Repugnant Conclusion: Essays 

on Population Ethics, ed. J. Ryberg and T. Tannsjo: 163-86, dordrecht: kluwer 
Academic Publishers.

Temkin, L. 1987: “Intransitivity and the Mere Addition Paradox”, Philosophy and 
Public Affairs 16: 138-187.

— 1993: Inequality, New York: Oxford University Press.
— 2012: Rethinking the Good: Moral Ideals and the Nature of Practical Reasoning, 

Oxford: Oxford University Press.
— 2014: “Intransitivity and the Internal Aspects View”, Law, Ethics and Philosophy 

2: 139-152.
Wittgenstein, L., 1958: Philosophical Investigations, 2nd ed., ed. and trans. by G. E. M. 

Anscombe, Oxford: Blackwell.