9

A Kantian Solution to the Trolley

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

Problem
Pauline Kleingeld

Introduction

The well-known moral dilemmas referred to as ‘trolley cases’ involve a

runaway trolley that will kill five people if it continues on its course;
their lives can be saved, but, in the para­digm scenarios, only at the cost
of the life of another person. The philosophical interest of these cases
lies in the fact that most people (including many philosophers) regard it
as morally permissi­ble or even required for a bystander to save the five by
diverting the trolley onto a different track where it will kill one person,
whereas most regard it as impermissible to save the five by pushing a
person off a footbridge into the path of the oncoming trolley. Which
moral principle, if any, can account for the salient difference between
these cases? This question has proved extremely difficult to answer, and
this diffi­culty is known as the ‘Trolley Problem.’ At least in part because
a solution to the problem seems impossible to many (though not all),1
attention has recently been shifting towards psychological explanations

1 Some have argued for a solution based on the Doctrine of Double Effect. Much of the
related debate focuses on whether its core distinction between ‘intended’ and ‘foreseen but not
intended’ consequences can distinguish effectively between the Bystander and the Footbridge
case. In this chapter, I do not take a stance on whether the Trolley Problem can be solved by the
Doctrine of Double Effect, nor do I assess Frances Kamm’s Doctrine of Triple Effect
(Kamm 2007). The argument of this chapter does not rule out the possibility of there being
more than one solution.

Pauline Kleingeld, A Kantian Solution to the Trolley Problem In: Oxford Studies in Normative Ethics, Volume 10.
Edited by: Mark Timmons, Oxford University Press (2020). © Pauline Kleingeld.
DOI: 10.1093/oso/9780198867944.003.0010
 A Kantian Solution to the Trolley Problem 205

of some of the intuitive judgments involved,2 and an increasing number

of authors now condemn the very use of trolley cases in ethics.3
In this chapter, I explore a new avenue for solving the Trolley Problem

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

by focusing on the practical reasoning attributed to agents in trolley
cases. In particular, I develop a solution to the Trolley Problem in terms
of the Kantian notion of ‘using someone merely as a means.’ At first sight,
this may seem to be an unpromising strategy. In her classic article ‘The
Trolley Problem,’ Judith Jarvis Thomson argues, on the basis of a ­scenario
known as the ‘Loop’ case, that a solution in terms of this Kantian notion
encounters insoluble difficulties (Thomson 1985: 1401–3). Her argument
is widely considered convincing. Against this view, I argue that there is a
plausible conception of ‘using merely as a means’ that yields an explanation
of the moral difference between the ‘Bystander’ and the ‘Footbridge’ case
without foundering on the Loop case.
I do not provide a justification of the moral prohibition on using
­persons merely as means, since this requires separate discussion, nor do
I fully settle the issue of whether diverting the trolley in the Bystander
and the Loop case is morally permissi­ble. Thus, I do not claim to deliver
a full solution to the Trolley Problem if this is thought to require a ‘satis-
fying justification for our initial intuitive judgments’ (Hurka 2016: 150).4
Instead, I use a different standard for what counts as a solution, also
common in the literature, according to which a solution requires a prin-
cipled account of the salient moral differ­ence between the Bystander and
the Footbridge case. Thomson claims that the Kantian prohibition on
‘using someone merely as a means’ cannot provide such an account; my
aim in this essay is limited to showing how that challenge can be met.

2 Some authors explain the so-called ‘deontological’ responses to trolley dilemmas as result-
ing from emotional reactions (Greene et al. 2001). Others argue that our intuitive moral heur­
istics become unreliable when applied to artificial hypo­thetical situations such as trolley
scenarios (e.g., Gigerenzer 2008: 11; Appiah 2008: 96–101; Bruers 2016).
3 Several authors criticize trolley dilemmas for abstracting from the de­tails and uncertain-
ties that characterize real-life situations of moral decision making. For different versions of this
objection, see, e.g., Wood (2011), Fried (2012).
4 I do not mean to rule out that readers who endorse the methodology of reflective equilib-
rium may come to believe that I accomplish all of this if they find the argument of this chapter
convincing. But my aim is more limited.
206 Pauline Kleingeld

I begin by presenting the Trolley Problem, explaining Thomson’s

challenge to a solution in terms of the notion of ‘using’ someone ‘merely
as a means’ and specifying the desiderata for an account that can meet

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

this challenge (Section 1). I then offer an account of ‘using’ someone
‘merely as a means’ (Section 2). I argue that this account enables a ­solution
to the Trolley Problem by explaining the morally salient difference
between the Bystander and the Footbridge case (Section 3), and I show
that it can also handle the Loop case (Sections 4 and 5).

1. The Trolley Problem and the Loop Challenge

As is by now familiar, Thomson articulated the Trolley Problem in a dis-

cussion of the following scenario devised by Philippa Foot:

[Someone] is the driver of a runaway tram which he can only steer

from one narrow track on to another; five men are working on one
track and one man on the other; anyone on the track he enters is
bound to be killed. (Foot 1978 [1967]: 23)

Foot examined the moral difference between a driver’s steering towards

the one man in this case and the case of a mayor who frames an i­ nnocent
person to prevent riots, or a transplant surgeon who kills a random
healthy person in order to use his organs to save a greater number of
patients. Problematizing some of Foot’s core claims, and in order to
focus on the structure of the moral issue without introducing complica-
tions arising from extraneous differences between the scenarios (such as
differences having to do with what is specific to organ transplantation as
compared to riot prevention or trolley malfunction), Thomson created
an entire family of trolley cases (Thomson 1985; cf. Thomson 1976).
In one variant—the By­stander scenario—the original situation is the
same as in the Driver case described by Foot, except that the driver is
incapaci­tated and a bystander is standing near a switch that will allow him
to divert the trolley onto the track with one workman on it. In another
variant—the Footbridge case—someone is standing on a f­ootbridge over
the track and sees the trolley hurtling towards five workmen. Standing
 A Kantian Solution to the Trolley Problem 207

nearby is a ‘really fat man’ who is leaning over the railing.5 Pushing him
would cause him to fall into the path of the oncoming trolley, which
would kill him but stop the trolley and save the five (and the agent can-

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

not stop it by jumping in front of it; Thomson 1985: 1409). There has
been a proliferation of sometimes unusual scenarios in the discussion to
date, such as cases involving people stuck on over­sized Lazy Susans
(Kamm 2007) and heavy men on remote-controlled roller skates
(Unger 1996: 90). But there is nothing outlandish about the basic philo-
sophical question behind the original scenarios on the basis of which
Thomson formulated the Trolley Problem:6

In both cases, one will die if the agent acts, but five will live who would
otherwise die—a net saving of four lives. What difference in the other
facts of these cases explains the moral difference between them?
(Thomson 1985: 1396, cf. 1401)

One of the possible answers Thomson considers but rejects is the reply
that the agent on the footbridge uses the heavy man merely as a means,
whereas the agent in the Bystander scenario does not. She allows that on
the face of it, the Kantian notion of ‘using’ someone ‘merely as a means’
may seem to capture the moral difference between the cases. As she puts
it, the bystander does not need or use the one workman to save the five,
because the latter’s presence on the track contributes nothing to achieving
the agent’s end of saving the five. By contrast, the agent on the footbridge
does need to use the one if he is to save the five (Thomson 1985: 1402).
Thomson argues that the Kantian notion of ‘using merely as a means’
does not provide a satisfactory answer, however, and to this end she
introduces a modified Bystander scenario called the Loop case. As in the
Bystander case, the tracks diverge, but now they reconnect further along
to form a loop. The one workman, who is stuck on the track beyond

5 Given the negative stereotypes often associated with being ‘fat,’ Thomson’s description
may prejudice responses to the scenario. Hence, the man on the bridge is nowadays sometimes
cast as carrying a backpack of trolley-stopping proportions, or simply as ‘heavy.’
6 In a more recent essay, Thomson argues that the Trolley Problem is not really a problem
after all, because diverting the trolley and pushing the man are both impermissible
(Thomson 2008). I do not find her new argument convincing, for reasons including but not
limited to the problems mentioned by FitzPatrick (2009).
208 Pauline Kleingeld

the fork, is so heavy that his body would stop the trolley, whereas the
five are so thin that all five (and only the five) would be killed if the
trolley were to hit them. The trolley is heading towards the five, but if

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

the bystander diverts it towards the one workman, this will save the five
and kill the one instead. In this Loop scenario, Thomson claims, the
bystander needs the one workman to save the five in the same way that
the agent on the footbridge needs the heavy man to stop the trolley. This
suggests, she claims, that diverting the trolley towards the one to save
the five in the Loop scenario is a case of ‘using him merely as a means’
(Thomson 1985: 1403).
Morally speaking, this would put the Loop case on a par with the
Footbridge case: in both scenarios, the heavy man would be used
merely as a means. As far as the details of the scenario are concerned,
however, the Loop scenario strongly resembles the Bystander case. The
only ­difference is a connecting bit of track in the distance. Thomson
argues that it is implausible to assume that this connecting bit of
track creates a major moral difference between the Bystander and the
Loop scenario. Moreover, she asserts, given the similarities with the
Bystander case, it still seems permissible for the bystander to divert
the trolley in the Loop case, even though this would involve using the
one workman merely as a means. If it is permissible in the Loop case to
divert the trolley despite the fact that this involves using the heavy man
merely as a means, h ­ owever, then what makes it impermissible in the
Footbridge case to push the heavy man in front of the trolley ­cannot
be the fact that this involves using him merely as a means. This in turn
implies that the p­ rohibition against using someone merely as a means
cannot explain the moral difference between the Bystander and the
Footbridge case.
Thomson therefore concludes that the Kantian notion is incapable of
solving the Trolley Problem:

[T]here is no plausible account of what is involved in, or what is neces­

sary for, the application of the notions ‘treating a person as a means
only’, or ‘using one to save five’, under which the surgeon [or the agent
who pushes the heavy man off the footbridge] would be doing this
whereas the agent in this variation [i.e., the bystander who diverts the
 A Kantian Solution to the Trolley Problem 209

trolley in the Loop scenario] . . . would not be. If that is right, then
appeals to these notions cannot do the work being required of them
here. (Thomson 1985: 1403)

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

Some Kantians have replied that Thomson’s intuition is mistaken.
Samuel Kerstein has claimed as much, arguing that diverting the
­trolley in the Loop case is impermissible because it involves using the
one merely as a means (Kerstein 2013: 124). This line of response,
however, is unlikely to satisfy those who, like Thomson herself in her
1985 paper and again in her most recent discussion of the issue (2016:
128–32), are wondering just how the addition of a bit of track causes
such a massive moral difference. We can press the issue by assuming
that the ­additional bit of track includes a drawbridge in the looping
section between the one and the five. If one assumes that diverting the
trolley would involve using the heavy man merely as a means in the
Loop case but not in the Bystander case, one must also maintain that,
in this modified Loop scen­ario, diverting the trolley would involve
using the workman merely as a means when the distant drawbridge is
down but not when it is up (assuming the other details of the case stay
the same). This result may seem far-fetched, and Thomson indeed says
that it would be ‘­thoroughly odd’ if diverting the trolley turned out to
be right or wrong depending on a loop connection that may or may
not obtain (Thomson 2016: 131). Moreover, given that the bridge is
located beyond the heavy workman and his body will stop the trolley
no matter whether the bridge is open or closed, it is unclear how the
position of the bridge could affect the permissibility of diverting the
trolley. A reply to Thomson’s challenge should address the Loop
­scenario’s similarity to the Bystander case and explain the significance
of the extra bit of track.
A full reply to Thomson’s challenge would therefore need to include at
least the following elements:

1. a plausible conception of ‘using someone merely as a means’;

2. an indication of the salient moral difference between the Bystander
and the Footbridge case in terms of this notion—that is, a basic
solution to the Trolley Problem; and
210 Pauline Kleingeld

3. a successful extension of this basic solution to the Loop case,

including an explanation of the significance (or lack thereof) of
the extra bit of track.

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

In the remainder of this chapter, I develop these elements in turn, start-
ing with the notion of ‘using someone merely as a means.’

2. ‘Using Someone Merely as a Means’

In the Kantian literature, the prohibition on ‘using’ someone ‘merely as a

means’ is commonly explained in terms of the violation of a requirement
concerning the consent of the person who is used as a means. This
requirement makes it possible to distinguish between abuse and
­permissible use. There are various accounts of the type and object of the
required consent,7 but it is not necessary to discuss alternative accounts
here. My aim is limited to showing that there is at least one plausible
understanding of the Kantian notion of ‘using someone merely as a
means’ that can meet Thomson’s challenge. The conception of ‘using
someone merely as a means’ that I use for the purpose of this chapter is
the following:

Using merely as a means: an agent uses another person ‘merely as a

means’ if and only if (1) the agent uses another person as a means in
the service of realizing her ends (2) without, as a matter of moral prin-
ciple, making this use conditional on the person’s consent, where (3)
the required consent is actual genuine consent to being used by the
agent in a particular manner, as a means to the agent’s end.

I have argued elsewhere (Kleingeld 2020), with the necessary exegetical

details, that this account of the prohibition is plausible as an inter­pretation
of Kant’s texts, and indeed more plausible than alternative i­ nterpretations. In
the present essay, I bracket exegetical debates. Because the canonical
exposition of the prohibition on using others ‘merely as means’ is found

7 See, e.g., Kerstein (2013), Korsgaard (1996), O’Neill (1989), Parfit (2011), Wood (2008).
 A Kantian Solution to the Trolley Problem 211

in Kant’s work, however, I will mention a few key passages as I present

the three elements of the account.

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

2.1 The Use-as-Means Condition

The classic formulation of the prohibition on using persons ‘merely as

means’ is found in Kant’s Formula of Humanity in his Groundwork for
the Metaphysics of Morals:

So act that you use the humanity, whether in your own person or in
the person of any other, always at the same time as an end, never
merely as a means. (G 4: 429)8

Note, first, that Kant speaks of ‘using’ rather than of ‘treating’ someone
merely as a means. In current debates, it is customary also to speak of
‘treating’ a person merely as a means, and even to use these terms
­alternatingly. ‘Treating’ has a much broader scope than ‘using,’ however.
It also includes mere attitudes toward others that do not amount to any
actual use of them as means, such as regarding others as being of poten-
tial use.9 ‘Treating’ seems to have been introduced into the discussion by
influential but infelicitous translations of the Formula of Humanity.10
Kant’s terminology is that of ‘using’ persons (merely) as means. On the
account I shall be employing in this chapter, the first condition without
which we cannot properly say that an agent uses a person ‘merely as a
means’ is that the agent indeed uses the person as a means to her end.

8 Kant’s writings are cited by the abbreviated title, using the Akade­mie volume and page
numbers (Kant 1900–). Translations are found in the Cambridge Edition of the Works of Kant
(Kant 1992–2016). Abbreviations: CLMP = Collins Lectures on Moral Philosophy; CPrR =
Critique of Practical Reason; FLNL = Feyerabend Lectures on Natural Law; G = Groundwork
for the Metaphysics of Morals; MM = Metaphysics of Morals; MM Vig = Lectures on the
Metaphysics of Morals Vigilantius.
9 Treating others merely as means does violate the demands of the Formula of Humanity,
but it involves a different type of violation: the agent fails to promote their humanity at the
same time as an end, as Kant puts it (G 4: 430).
10 Kant uses ‘(ge)brauchen,’ which means ‘use,’ but both H.J. Paton and Lewis White Beck
use ‘treat’ in their translations. Kant uses ‘treating’ at least once (‘behandeln,’ G 4: 433). This is
compatible with ‘using’ being the dominant term, since ‘using’ is a form of ‘treating.’ But since
not all ‘treating’ is ‘using,’ I employ Kant’s terminology of ‘using’ when discussing the Formula
of Humanity.
212 Pauline Kleingeld

There is another terminological issue that requires comment at the

outset, namely the appropriateness of speaking of using ‘a person,’ given
that the Formula of Humanity speaks of using ‘the humanity in a person’

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

merely as means. By ‘humanity’ Kant understands that which distin-
guishes humans from other animals, and this is their capacity to act in
accordance with rational principles and to set ends independent­ly of
inclination (G 4: 428–31, 434). He argues, however, that in a human per-
son this capacity is inseparably connected with the body and that using
part of an entity constitutes using that entity (MM 6: 279, CLMP 27:
387). In other words, using a person’s body or using a person’s rational
capacities counts as using a person. Similarly, when using a particular
program on my computer, I can say that I am using my computer. Kant
indeed often speaks of ‘using a person merely as a means’ or ‘using
another human being merely as a means’ (G 4: 429), and in this chapter I
will do the same.
But when does an agent use another person as a means? This question
can be answered in two different ways. We can describe the agent’s
action either from the agent’s own perspective, in terms of her end, the
means by which she intends to promote or achieve it, and her associated
line of reasoning; or we can describe the action from an external
­narrator’s perspective, focusing on the agent’s observable behavior and
another person’s causal contribution to the realization of the agent’s pre-
sumed end.
Authors discussing the Trolley Problem often describe the agent’s
action from an external perspective. In order to determine whether the
one person is used merely as a means, they examine whether that
­person’s presence is required for (or contributes to) achieving the agent’s
end. Thomson, for example, suggests that the one workman is used
merely as a means ‘only if, had the one gone out of existence just be­fore
the agent started, the agent would have been unable to save the five’
(Thomson 1985: 1402). Since this criterion is external to the agent’s own
means–ends reasoning, it may describe an agent as using another per-
son merely as a means even if that person does not serve as a means in
the agent’s own practical reasoning at all.
Kant, by contrast, takes the agent’s own practical reasoning to be the
relevant point of reference for determining whether the agent is using
 A Kantian Solution to the Trolley Problem 213

another person as a means. In the Groundwork, his examples of agents

who use others merely as means are phrased in terms of the agent’s prac-
tical reasoning: in terms of what the agent ‘has in mind’ or ‘wants’ and

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

what ‘principles’ he is acting on (G 4: 429–30).
For the purposes of this chapter, I will similarly take the relevant
point of reference to be the agent’s own practical reasoning. An agent
uses another person as a means if and only if she wants to reach a ­certain
end, believes that she can reach or further this end mediately, by using
another person as a means, and uses the person for the sake of reaching
or furthering her end.

2.2 The Consent Proviso

Not every case in which an agent uses someone as a means also

involves the agent’s using someone ‘merely’ as a means. We need a
second ­condition to pick out the proper subset, and this is the role
of the ­consent requirement. On the account I will be employing in
this ­chapter, an agent who uses someone as a means uses this person
‘merely’ as a means if and only if she fails to honor the moral constraint
that her use of o ­ thers must be conditional on their consent (and the
relevant type of consent is actual consent, as explained in Section 2.3).
Making one’s use of others conditional on their consent involves com-
mitting oneself not to use others as means unless they consent to being
used as such. The agent ought to honor this constraint as a matter of
moral principle, that is, as a matter of recognizing others as ‘ends in
themselves’ (as Kant puts it), rather than as a matter of following mere
personal inclination.
Kant formulates the consent proviso in terms of a limiting condition
on the agent’s maxims. Maxims, in his moral theory, are the major
premises in the agent’s practical reasoning: the principles on the basis of
which the agent acts (G 4: 421n.; CPrR 5: 19). Although the Formula of
Humanity does not contain an explicit reference to the agent’s maxims,
Kant repeatedly emphasizes that it articulates a limiting condition on
maxims. He claims that ‘the formula says’ that rational beings ‘must
serve in every maxim as the limiting condition of all merely relative and
214 Pauline Kleingeld

elective ends’ (G 4: 436, emphasis added; also 437). He rephrases this

idea by stating that rational beings must serve as the ‘highest limiting
condition in the use of all means’ (G 4: 438).

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

Kant explains the nature of this limiting condition in terms of the
consent of the person who is used as a means. In the Groundwork, he
gives the example of a person who wants to obtain money by getting a
loan with a false promise to repay it. He explains that the false promisor
wants to use the lender ‘merely as a means’ by saying that the lender
‘cannot possibly consent to [the false promisor’s] way of behaving toward
him’ (G 4:429). This suggests that if an agent uses another as a means to
her end without having made this use conditional on the other’s con-
sent, then she uses the other merely as a means.

2.3 The Relevant Type of Consent Is Actual Consent

In the literature, the relevant type of consent is variously conceived as

hypothetical, rational, possible, or actual consent. The textual evidence
suggests, however, that the relevant type of consent is actual consent. Kant
argues that actual consent is required whenever one person wants to use
the property or powers of another (MM 6: 285, 271–2). As an example, he
mentions that the use of servants requires their actual ­consent: a servant
‘must also be an end . . . and not merely a means. He must also want it’
(FLNL 27: 1319; cf. MM 6: 283). Furthermore, Kant introduces the loan
example in the Groundwork by saying that the lender does not actually
consent to the use the false promisor wants to make of him (‘without [the
lender] at the same time containing in h ­ imself the [false promisor’s] end,’
given that the latter misrepresents his real end, G 4: 429).
For a full explanation of the requirement that agents use others as
means only on condition that the others actually consent, much more
would need to be said, of course, but for the purposes of this chapter it is
not necessary to do so. None of the agents in the trolley cases under
consideration obtain the actual consent of the men who are about to be
run over by the trolley. If these agents use the men in these scenarios as
means, they clearly do so regardless of whether the men actually con-
sent, so a discussion of the finer details of applying this requirement is
unnecessary.
 A Kantian Solution to the Trolley Problem 215

Is this account a plausible conception of ‘using someone merely as a

means,’ as Thomson requires? It straightforwardly explains the notion in
terms of its constituent elements, and it distinguishes effectively between

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

using a person ‘merely as a means’ and using a person ‘as a means, but
not merely as a means.’ Furthermore, it matches widely shared moral
convictions concerning the requirement of informed and voluntary
consent in cases of using others as means. For example, informed con-
sent requirements in medical research contexts impose on researchers
an obligation to make their use of human test subjects conditional on
those subjects’ actual consent (to being used in the manner specified
and in the service of a particular research project). This widely shared
moral conviction matches the Kantian prohibition as I have described it
here. Thus, this conception of the prohibition on using others ‘merely as
a means’ seems to have enough initial plausibility, even without a full
defense of the prohibition as such (which, as mentioned, it is not the
aim of this chapter to provide), to warrant examining its potential for
handling the Trolley Problem.

3. The Basic Solution to the Trolley Problem

Returning now to the trolley scenarios, we can draw out important

implications of this account of ‘using merely as a means’ for the assess-
ment of the cases at issue. Clearly, on this conception of the notion,
whether the agent is using the workman ‘merely as a means’ cannot be
established by examining the causal relation between the death of the
one man and the survival of the five. We need to focus our attention on
the agent, namely on whether she uses the one as a means without
­making this use conditional on the workman’s actual consent.
None of the agents in the Bystander, Footbridge, and Loop cases have
obtained the actual consent of any of the men involved. This entails that
if the agent uses the one man as a means to her end of saving the five,
then she uses him merely as a means. The crucial question in each of the
three trolley scenarios, therefore, is whether the agent uses the one man
as a means, and this, as noted above, depends on her practical reasoning.
In what follows, I first show how the proposed account of ‘using
merely as a means’ explains the morally significant difference between
216 Pauline Kleingeld

switching the trolley towards the track with the one workman (in the
Bystander case) and pushing the heavy man off the bridge (in the
Footbridge case). I then turn to the Loop case in the next section.

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

As initially suggested by Thomson, and as widely agreed in the Trolley
Problem literature, the agent who diverts the trolley in the Bystander
scenario does not use the one workman as a means (e.g., Thomson 1985:
1402).11 The agent is assumed to have the end of saving more rather
than fewer lives. She finds herself forced to choose: either she lets the
trolley continue, in which case five workmen die, or she diverts it, in
which case one workman dies. She realizes that she will achieve her end
by steering the trolley onto the sidetrack, and this is why she decides to
divert it. The fact that the workman dies as a consequence of her
­diverting the trolley does not mean that she uses him as a means to her
end. At no point does he enter into her practical reasoning as a means to
her end of saving more rather than fewer lives. In fact, she could save
even more lives if he were not there. Since the agent who diverts the
trolley in the Bystander scenario does not use the one workman as a
means, this trivially entails that she does not use him merely as a means.
In the Footbridge case, by contrast, an agent who deliberately pushes
the heavy man off the bridge as a means to her end of saving the five
does use him merely as a means. Here the conditions for using a person
‘merely as a means’ are satisfied. First, it is clear that the agent uses the
heavy man as a means. Her reason for pushing him is that his body will
stop the trolley and she will thereby reach her end of saving the five. In
fact, she regards the heavy man as ‘the one’ (as compared to ‘the five’ on
the tracks) because she can use him to stop the trolley. Given his weight,
his expected impact on the trol­ley, and her own ability to push him over
the edge, she regards the heavy man as ‘the one’ qua potential trolley
stopper. If he were thin, or if he could not be moved, she would not
consider pushing him. Furthermore, if there were a thin man standing
next to the heavy man on the bridge, only the heavy man could be used
as a means to stop the trolley, and he would suffice. Any agent who

11 As is customary in the debate, I assume in the discussions to follow that the agent is nei-
ther confused, nor acting on an independent desire to kill one workman, nor acting on mis-
taken beliefs (including the mistaken belief that the five are saved by killing the one workman,
rather than by diverting the trolley).
 A Kantian Solution to the Trolley Problem 217

wants to save more lives rather than fewer would therefore face only the
question of whether to push him—not whether to push ‘the two.’ Thus,
the heavy man’s role as ‘the one’ in the Footbridge scenario is defined by

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

his potential use as a means to stop the trolley. And if the agent pushes
him in order to stop the trolley so as to save the five, she necessarily uses
him as a means.12
Second, the details of the Footbridge scenario indicate that the agent
who uses the heavy man as a means to stopping the trolley does so
­without making her use of him conditional on his actual consent.
Because she deliberately uses him as a means to stopping the trolley
without having made this conditional on his actual consent, she uses
him merely as a means.
The basic solution to the classic Trolley Problem in terms of this
account of ‘using merely as a means,’ then, is as follows. The agent in the
Bystander scenario who diverts the trolley to save five does not use the
one workman as a means to her end of saving the larger number. Since
the use-as-means condition is not satisfied, she does not use the one
workman merely as a means. By contrast, the agent in the Footbridge
scenario who deliberately pushes the heavy man necessarily uses him as
a means, and given the details of the scenario, she does so without
­making her use of him conditional on his consent. Thus, she necessarily
uses him merely as a means.
In this way, the proposed conception of ‘using merely as a means’
yields an account of the salient moral difference between the Bystander
and the Footbridge case: the agent in the first case does not use the man
merely as a means, whereas the agent in the second case does.
On the assumption that it is morally impermissible to use another
merely as a means, this establishes that pushing the man off the
­footbridge is impermissible, but it does not establish that it is morally
permissible—let alone required—for the bystander to divert the trolley.
This is because the argument above does not rule out that there are other
considerations that render it impermissible. When a certain action can

12 This analysis of the Footbridge scenario does not change if we recast the heavy man as a
thin man strapped to a backpack (or to a wheelchair, a voluminous suit, or any other item) of
trolley-stopping proportions. If it is impossible to toss only his backpack off the bridge, then
pushing the backpack to which he is attached counts as pushing him (see Section 2.1 above).
218 Pauline Kleingeld

be performed only by using someone merely as a means, as in the

Footbridge case, we can indeed conclude that the principle implies
that the action is impermissible. But it does not follow that any action

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

that is performed without using a person merely as a means is there-
fore ­permissible. The Formula of Humanity not only prohibits using
persons merely as means but also requires promoting them as ends in
themselves. Agents who avoid using anyone merely as a means may
still be violating this further requirement (see Kant’s Groundwork
example of the person who refuses to help others in need). For the
purpose of this chapter, however, it is not necessary to pursue
such cases.

4. Two Construals of the Loop Case

The above account of ‘using someone merely as a means’ has an impli-

cation that is crucial for the analysis of the Loop case: one and the
same action does or does not constitute an agent’s using someone
‘merely as a means’ depending on the agent’s underlying practical
­reasoning. This implication does not rise to the surface in the discus-
sion of the Bystander and Footbridge cases. Given the details of these
scenarios and the reason­ing attributed to the agents, the agent in the
Footbridge case does use the one man merely as a means, but the
agent in the Bystander case does not. I will argue that in the Loop
case, by contrast, we can ascribe different lines of practical reasoning
to the agent, such that the agent does or does not use the heavy work-
man merely as a means, depending on what we stipulate. This will
make it possible to formulate the response to Thomson’s challenge in
the next section.
It is not difficult to conceive of a bystander in the Loop case who uses
the heavy man merely as a means. Suppose an agent named Tili is stand-
ing near the switch. Tili sees the trolley approaching the five, notices the
heavy man on the track beyond the fork, and recognizes that she could
use him to prevent the trolley from killing the five. She realizes that her
situation is such that she has to make a decision without having any
 A Kantian Solution to the Trolley Problem 219

further knowledge about the six people involved. I abbreviate the

description of this aspect of the situation as ‘S.’13 She reasons as
follows:14

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

Tili: When forced to choose in S, I will save more rather than fewer
human lives, even if this involves my using others as means to this end
without their actual consent. Now I am forced to choose in S: either I
let the trolley continue towards the five, which will save one life, or I
divert it towards the heavy man so as to use him as a means to stop-
ping the trolley, which will save five lives. Five is more than one, so I
shall save the five and divert the trolley in order to use the heavy man
as a means to stopping it.

On the basis of this line of reasoning, Tili diverts the trolley towards the
heavy man. In situations like these, she reasons from the general action
principle stated in the first sentence (‘when forced . . .’), and if she ever
found herself in the Footbridge scenario, this principle would lead her
to push the heavy man. Reasoning now from this principle in the Loop
case, she decides to divert the trolley towards him in order for his body
to halt it. This means that she uses the heavy man as a means to her end,
and she also does so without his actual consent. Thus, she uses the heavy
man merely as a means in the full sense of the account presented above.
The details of the Loop scenario, in particular the scenario’s structural
similarities to the Bystander case, also make it possible, however, for us
to imagine a bystander who diverts the trolley without using the heavy
man as a means, and hence without using him merely as a means.
Suppose an agent named Manuela15 reasons from an action principle
that includes the morally required limiting condition: ‘When forced to

13 This is meant to bracket those complications that the trolley scenarios were specifically
designed to bracket, such as special relationships between the agent and the workmen, specific
characteristics of some of the workmen involved, etc.
14 The focus here is on patterns in moral reasoning and justification, not psychology. This is
why I abstract from the psychological fact that most real agents would feel conflicted, pan-
icked, and horrified.
15 I introduce a second agent (rather than two possible patterns of reasoning for one agent)
so as to avoid any suggestion that the agent is ‘tinkering’ with her maxims to produce a morally
220 Pauline Kleingeld

choose in S, I will save more rather than fewer human lives, provided I
do not use anyone as a means to this end without their actual consent.’
Unlike Tili, Manuela would refrain from pushing the heavy man in the

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

Footbridge scenario. But what about the Loop case?
Manuela knows that if she lets the trolley continue, five people will
die. She also knows that if she diverts the trolley, it will hit and kill the
heavy man and stop there. That is, she knows that if she diverts the trol-
ley, this will lead to one death. She reasons as follows:

Manuela: When forced to choose in S, I will save more rather than

fewer human lives, provided I do not use anyone as a means to this
end without their actual consent. Now I am forced to choose in S:
either I let the trolley continue towards the five, which will save one
life, or I divert it towards the heavy man, which will kill him but save
five lives. Five is more than one, so I shall save the five by diverting the
trolley.

On the basis of this line of reasoning, Manuela diverts the trolley. Her
end is to save more rather than fewer human lives. She knows that if she
diverts the trolley, one person dies, and that if she lets it continue, five
die. She will achieve her end by diverting the trolley to the side where
only one man will die, and this is why she diverts it. Her reasoning
includes the morally required proviso, and at no point does the heavy
man enter into her reasoning as a means. It is not that she neglects to
take note of important facts about the case: she knows that his body will
stop the trolley and that this will kill him. But this prediction as such
does not include any instrumental role for him, and she could have
made it even if she had not yet seen the five ahead. Since she does not
use the heavy man as a means, she does not use him merely as a means.
So in terms of the above account, Manuela diverts the trolley without
using the heavy workman merely as a means.
There is a related context in which similar distinctions are drawn.
This is the discussion of whether a bystander in the Loop case who

desirable result. I stipulate that Tili and Manuela really have adopted their respective action
principles and really are committed to acting accordingly in the relevant circumstances.
 A Kantian Solution to the Trolley Problem 221

decides to let the trolley continue towards the five would be using the five
as means to save the one—say, in acting on the principle that it is better
to do nothing than to kill one. Commentators generally deny that this

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

inactive bystander would be using the five as means, for the very same
reason that I have offered in support of the claim that Manuela does not
use the heavy man as a means. They emphasize that the five do not enter
into the inactive bystander’s practical reasoning as means.16 They say
that the fact that the five die as a consequence of the bystander’s decision
to let the trolley continue does not entail that she uses them as means.
By parity of reasoning, then, the fact that the heavy workman dies as a
consequence of Manuela’s diverting the trolley does not entail that she
uses him as a means, let alone that she uses him merely as a means. Her
end is to save more rather than fewer human lives, and she attains this
end by diverting the trolley to the side where—as she predicts—one man
will die.
The similarities between the Loop and the Bystander case make it
possible to conceive of an agent who diverts the trolley in the Loop case
without making instrumental use of the heavy man. An important dif-
ference between the Bystander and the Loop case is, of course, that in
the Loop scenario the agent would not be able to save the five if the one
workman were not there. But the truth of this counterfactual does not
invalidate the analysis of the actual case at hand.
The significance of the extra bit of track, then, is that it makes it
­possible to construe the agent’s reasoning in the Loop scenario as similar
to that in either the Bystander case or the Footbridge case. This makes
the Loop case fundamentally different from both. In the Bystander
­scenario, the agent who diverts the trolley does not use the one work-
man as a means, and hence she does not use him merely as a means. In
the Footbridge case, the agent who deliberately pushes the heavy man
uses him merely as a means. In the Loop scenario, by contrast, the agent
who diverts the trolley towards the heavy man does (e.g., Tili) or does

16 For example, Kerstein (2013: 123) writes: ‘that you allow the five to be killed does not
entail that you are using them.’
222 Pauline Kleingeld

not (e.g., Manuela) use him merely as a means, depending on their line
of reasoning.17
The fact that the Loop case allows for such different lines of reasoning

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

explains why the scenario has generated such radically different moral
assessments in the literature. Some authors, including Thomson in her
1985 essay, assimilate the Loop case to the Bystander scenario and
regard it as permissible to divert the trolley towards the one. Others,
including Kerstein (2013) and Michael Otsuka (2008), assimilate the
Loop case to the Footbridge scenario and regard it as impermissible to
divert the trolley. The argument of this section suggests that it is a mis-
take to assume that the Loop scenario should be assimilated to only one
of the two. The scenario permits both construals. Thus, the account of
‘using merely as a means’ applied in this chapter helps to explain why
the Loop scenario generates so much disagreement, and I take this to be
one of its strengths.
The main result of this section, however, is the conclusion that it is
possible for a bystander to divert the trolley in the Loop case without
using the heavy man as a means, and hence without using him merely as
a means. According to Thomson, quoted above, on any plausible account
of what is involved in ‘using someone merely as a means,’ any bystander
who diverts the trolley in the Loop scenario would be doing so. The
argument of this section shows that this is not true, and this makes it
possible to respond to Thomson’s Loop challenge.

5. Extending the Basic Solution to the Loop Scenario

Recall that Thomson’s challenge is as follows: if it is permissible to divert

the trolley in the Loop case, even while using the heavy man merely as a
means, then what makes it impermissible to push the heavy man in the

17 One might wonder whether it is possible to generate different construals for the
Bystander case, too. I do not think it is, at least if we assume that the Bystander does not act on
the basis of mistaken beliefs (see note 11). But even if it were possible, then this would not be a
problem for the argument of this chapter. The Trolley Problem would then require a solution in
terms of the salient moral difference between the Footbridge case and the two Bystander cases
(in its classic and loop versions). That is, the solution would then consist in the salient moral
difference between the Footbridge and the Loop case described in Section 5.
 A Kantian Solution to the Trolley Problem 223

Footbridge case cannot be the fact that this involves using him merely as a
means. This in turn implies that the Kantian prohibition on using some-
one ‘merely as a means’ cannot solve the Trolley Problem. In the previous

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

section I argued that diverting the trolley in the Loop scenario does not
necessarily involve using the heavy man merely as a means. But what dif-
ference does that make? If Tili diverts the trolley, she uses the heavy man
merely as a means, and if it is nevertheless permissible for her to divert the
trolley—as Thomson (1985) assumes it is—then the Loop case still seems
to undermine the basic solution of the Trolley Problem in Section 3 above.
In order to answer Thomson’s Loop challenge, we need to look more
closely at the relation between the permissibility of an action and the
practical reasoning that underlies it. I will argue that the Loop challenge
can be answered by combining the following two claims, both of which
are familiar Kantian principles: (1) A morally permissible action remains
permissible even when it is performed on the basis of a morally
­impermissible action principle; and (2) if an action can be performed
only on the basis of morally impermissible action principles, then the
action is impermissible.
In the previous section I showed that Tili and Manuela can divert the
trolley on the basis of practical reasoning of very different moral quality.
It is a familiar idea within Kantian ethics that a permissible action can
be performed on the basis of permissible or impermissible action prin­
ciples (‘maxims’). A helpful illustration of this point is Kant’s ­well-known
Groundwork example of a shopkeeper who charges children the right
price on the basis of an impermissible action principle, namely on the
basis of the maxim of only pursuing his own long-term interests (G 4:
397).18 Charging children the right price (the action), Kant says here, is
‘in accord with duty,’ regardless of whether the shopkeeper’s maxim sat-
isfies moral requirements. The action is required by the duty of honesty,
and a shopkeeper who charges children the right price still performs an
action that is ‘in accord with duty,’ even if his practical reasoning is
guided by the principle of self-interest. To add another example: resusci-
tating the victim of an accident does not become impermissible if the

18 Kant argues elsewhere that acting on the maxim of self-interest is impermissible (MM 6:
451; cf. also MM Vig 27: 621).
224 Pauline Kleingeld

agent uses the victim merely as a means to practicing her life-saving

skills, on the basis of an egoistic principle (say, in order to get a higher
salary after she passes a life-saving test, and she would not resuscitate

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

anyone if it weren’t instrumental to reaching her goal of increasing her
salary). If the duty to help others in need requires that she resuscitate
the victim, then she does the right thing, even if she does it for the
wrong reason.
Thus, a morally permissible or required action does not become
impermissible when performed by an agent who reasons from an
impermissible maxim. Nor does an impermissible maxim become
­permissible if it happens to lead an agent to perform a permissible
action. The distinction between the moral status of the action and the
moral status of the underlying maxim (and the associated reasoning) is
crucial to Kantian moral theory.19
This distinction explains how the permissibility of diverting the t­ rolley
in the Loop scenario is compatible with the impermissibility of using the
heavy man merely as a means. The permissibility of diverting the trolley
concerns the action (in abstraction from the reasoning that underlies it),
whereas the impermissibility of using the heavy man merely as a means
concerns the agent’s underlying reasoning. Tili and Manuela perform
the same action in the Loop case: they both divert the trolley. The
­permissibility of diverting the trolley does not vary with the different
moral quality of their practical reasoning. If diverting the ­trolley is per-
missible, it is permissible for Tili too, even though she uses the heavy
man merely as a means in the process. And this clearly does not mean
that it is permissible for her to use the heavy man merely as a means.
We are now in a position to answer the question that still remains:
suppose that diverting the trolley in the Loop case is permissible, and
hence permissible for Tili, then how can the prohibition against using
others merely as means explain the impermissibility of pushing the

19 The terminology with which authors draw related distinctions varies. For example,
Christine Korsgaard distinguishes between ‘act’ and ‘action’ (Korsgaard 2009: 12), and
T.M. Scanlon distinguishes between the ‘permissibility’ and the ‘meaning’ of an action
(Scanlon 2008: 89–121, esp. 101). Non-Kantian authors have drawn similar distinctions (e.g.,
Philip Pettit 2015, ch. 5).
 A Kantian Solution to the Trolley Problem 225

heavy man off the footbridge? The answer turns on a crucial difference
between the Footbridge and the Loop case.
In the Footbridge case, the action of pushing the heavy man off the

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

bridge (given the details of the case) necessarily involves the agent’s using
him merely as a means, as I argued in Section 3. For an action to be
­morally permissible, it should of course at least be possible for an agent
to perform it without violating moral constraints. Thus, if an action can
be performed only on the basis of morally impermissible action prin­
ciples, then it is impermissible. This is precisely the situation in the
Footbridge case, since pushing the heavy man necessarily violates the
prohibition on using others merely as means (given the details of the
scenario and the kind of reasoning that can be attributed to the agent
given these circumstances). Pushing the heavy man in front of the
­trolley is therefore impermissible, and what explains its impermissibility
is indeed the fact that it (necessarily) involves using him merely as
a means.
In the Loop case, by contrast, diverting the trolley does not necessarily
involve the agent’s using the heavy man merely as a means, as I argued
in Section 4. Thereby the Loop scenario opens up the possibility that
diverting the trolley is permissible or even morally required. And if
diverting it is indeed permissible (which is an issue I do not settle in this
essay), then diverting is permissible even for an agent such as Tili who
uses the heavy man merely as a means. She would perform a permissible
action, even though she would do so on the basis of an impermissible
action principle (maxim).
In sum, despite initial appearances to the contrary, the following two
claims are compatible: (1) diverting the trolley in the Loop case is
­permissible, even for an agent who uses the heavy man merely as a
means in doing so; (2) what makes it impermissible to push the heavy
man in the Footbridge case is the fact that this (necessarily) involves
using him merely as a means. The Loop case does not undermine the
basic solution to the Trolley Problem after all.
It seemed to Thomson, in her 1985 paper, that diverting the trolley in
the Bystander and the Loop case is permissible. It is not the aim of this
chapter to determine fully whether diverting in either case really is
­permissible, but it seems to me that Thomson’s impression was correct
226 Pauline Kleingeld

and, moreover, that Manuela’s reasoning stands a good chance of

­satisfying Kantian criteria. For reasons mentioned above, however, a full
de­ ter­
min­ation of what is permissible in these cases requires more

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

­argument, to assess whether the agent would violate moral requirements
other than the one discussed here.20 It is not necessary to settle this for
the more limited purpose of this essay. My aim here has been to offer an
explanation of the salient moral difference between the Bystander and
Footbridge case that does not founder on the Loop case. I have argued
in defense of the following account of the relevant difference: the agent
in the Footbridge case who pushes the heavy man off the bridge to stop
the trolley necessarily uses him ‘merely as a means,’ whereas the agent
who diverts the trolley in the Bystander case does not use him ‘merely as
a means.’ Regardless of whether diverting the trolley in the Bystander
case turns out to be permissible, all things considered, this difference
will remain.21

References

Appiah, Kwame Anthony. 2008. Experiments in Ethics. Cambridge, MA:

Harvard University Press.
Bruers, Stijn. 2016. “Can Deontological Principles Be Unified? Reflections
on the Mere Means Principle.” Philosophia 44: 407–22.
FitzPatrick, William J. 2009. “Thomson’s Turnabout on the Trolley.” Analysis
69: 636–46.

20 Further issues that deserve consideration at that point include the distinction between
killing and letting die, and the question of whether killing one is morally worse than ‘doing
nothing’ even if doing nothing means letting five people die. I doubt that it follows from
Kantian principles that killing one is morally worse than letting five die, assuming that no per-
son is used merely as a means in the process—especially if we take pains to avoid loading the
term ‘killing’ with moral content in a question-begging way. But pursuing this issue lies beyond
the scope of this chapter.
21 I thank the participants in the 2019 Workshop in Normative Ethics and two anonymous
referees for Oxford Studies in Normative Ethics for very helpful comments on this chapter. I also
thank audiences at Cardiff, Oxford, and the 2013 Eastern APA (especially Kenneth Walden), as
well as Joel Anderson, Tom Bates, Katharina Bauer, Jochen Bojanowski, David Copp, David
Enoch, Govert den Hartogh, Florian Merwede, Hanno Sauer, Oliver Sensen, Paulien Snellen,
Jens Timmermann, Marijana Vujosevic, Allen Wood, and members of the ethics research
seminar at the University of Groningen for perceptive comments on a much earlier version of
this chapter. I am grateful to the Dutch Research Council (NWO) for research support.
 A Kantian Solution to the Trolley Problem 227

Foot, Philippa. 1978 [1967]. “The Problem of Abortion and the Doctrine of
Double Effect.” In: Virtues and Vices and Other Essays in Moral Philosophy,
19–32. Oxford: Blackwell. [Originally in Oxford Review 5: 5–15.]

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

Fried, Barbara. 2012. “What Does Matter? The Case for Killing the Trolley
Problem (or Letting It Die).” Philosophical Quarterly 62: 505–29.
Gigerenzer, Gerd. 2008. “Moral Intuition = Fast and Frugal Heuristics?” In:
Moral Psychology, ed. Walter Sinnott-Armstrong, vol. 2, 1–26. Cambridge,
MA: MIT Press.
Greene, Joshua D., R. Brian Sommerville, Leigh E. Nystrom, John M. Darley,
and Jonathan D. Cohen. 2001. “An fMRI Investigation of Emotional
Engagement in Moral Judgment.” Science 293: 2105–8.
Hurka, Thomas. 2016. “Trolleys and Permissible Harm.” In: F.M. Kamm,
The Trolley Problem Mysteries, ed. Eric Rakowski, 135–50. Oxford: Oxford
University Press.
Kamm, F.M. 2007. Intricate Ethics: Rights, Responsibilities and Permissible
Harm. Oxford: Oxford University Press.
Kant, Immanuel. 1992–. Cambridge Edition of the Works of Immanuel Kant,
eds. Paul Guyer and Allen W. Wood. Cambridge: Cambridge
University Press.
Kant, Immanuel. 1900–. Kant’s gesammelte Schriften, ed. Preussi­sche [later
Deut­sche] Akademie der Wissenschaften. Berlin: Georg Reimer, later
Walter de Gruyter.
Kerstein, Samuel. 2013. How to Treat Persons. Oxford: Oxford University Press.
Kleingeld, Pauline. 2020. “How to Use Someone ‘Merely as a Means.’ ”
Kantian Review 25: 389–414. DOI: 10.1017/S1369415420000229.
Korsgaard, Christine M. 2009. Self-Constitution: Agency, Identity, and
Integrity. Oxford: Oxford University Press.
Korsgaard, Christine M. 1996. Creating the Kingdom of Ends. Cambridge:
Cambridge University Press.
O’Neill, Onora. 1989. Constructions of Reason. Cambridge: Cambridge
University Press.
Otsuka, Michael. 2008. “Double Effect, Triple Effect and the Trolley
Problem: Squaring the Circle in Looping Cases.” Utilitas 20: 92–110.
Parfit, Derek. 2011. On What Matters. Oxford: Oxford University Press.
Pettit, Philip. 2015. The Robust Demands of the Good: Ethics with Attachment,
Virtue, and Respect. Oxford: Oxford University Press.
228 Pauline Kleingeld

Scanlon, T.M. 2008. Moral Dimensions: Permissibility, Meaning, and Blame.

Cambridge, MA: Harvard University Press.
Thomson, Judith Jarvis. 1976. “Killing, Letting Die, and the Trolley

Downloaded from https://academic.oup.com/book/33455/chapter/287727751 by University of Groningen user on 07 March 2023

Problem.” Monist 59: 204–17.
Thomson, Judith Jarvis. 1985. “The Trolley Problem.” Yale Law Journal 94:
1395–415.
Thomson, Judith Jarvis. 2008. “Turning the Trolley.” Philosophy and Public
Affairs 36: 359–74.
Thomson, Judith Jarvis. 2016. “Kamm on the Trolley Problems.” In:
F.M. Kamm, The Trolley Problem Mysteries, ed. Eric Rakowski, 113–34.
Oxford: Oxford University Press.
Unger, Peter. 1996. Living High and Letting Die: Our Illusion of Innocence.
Oxford: Oxford University Press.
Wood, Allen W. 2008. Kantian Ethics. Cambridge: Cambridge
University Press.
Wood, Allen W. 2011. “Humanity as an End in Itself.” In Derek Parfit, On
What Matters, vol. 2, 58–82. Oxford: Oxford University Press.