Andreas Føllesdal

© 1985 05 06

THE THEORY OF REFLECTIVE EQUILIBRIUM

Second Year Paper

Philosophy Department, Harvard University

1.1 Introduction
In this paper I shall present and discuss the theory of reflective
equilibrium, a theory of justification in ethics. The aim of such a
justification is practical: to provide a justification of judgments, actions or
policies concerning particular issues, by giving reasons accepted by the
group we are concerned with justifying something for. The reflective
equilibrium which provides such justification is the temporary end
product of a process, the method of reflective equilibrium, which lets us
arrive at agreement by rational argument Why we seek to justify our
behavior, and why we seek agreement by argument instead of settling the
issues by force are important questions, but beyond the scope of the
theory - and of this paper.
The theory is not claimed to be an alternative to the ways ethical
argument and justification have been done before. Rather, I regard it as
an explication of the practices we and those before us have engaged in -
an account which lets us step back and consider what we are doing, so
that we may succeed more often in the future.
I first give a summary of the theory: What reflective equilibrium is, and
what the method consists of. I then give a brief sketch of the historical
development of this theory since the 1950s. In chapter 2 I discuss the
nature of considered moral judgrnents, and in chapter 3 the method of
reflective equilibrium. I conclude the presentation with a chapter on the
justification provided by reflective equilibrium. Throughout the
presentation I seek to clarify and answer objections which have been
raised against various formulations of the theory.

I .2 Main features of the theory of reflective equilibrium

The theory of reflective equilibrium has three main features.
i) The aim of arriving at reflective equilibrium is practical: To attain

-1-
agreement on specific moral issues.
Such agreement will allow for practical justification of actions and moral
judgments. The theory presupposes that there is a shared desire among
the parties to arrive at such an agreement on the issues it is applied to.
2) The theory claims that there is a rational procedure by which this aim
sometimes can be reached.
Arguments can be conducted according to standards of practical reason
accepted by all parties. These standards are the same as those used in
scientific arguments. Thus, if an agreement is achieved, it is not mere
agreement, in the sense of being generated by some irrational
mechanism. Rather, the agreement is in same sense objective. The theory
assumes that the parties wish to proceed according to such a procedure.
3) The starting points of the procedure are nonfoundational considered
judgments: judgments concerning particular cases (“particular
judgments”), moral principles, and “second—order judgments”:
judgments concerning standards of reasoning, formal requirements on
principles, the procedure itself etc. None of the considered judgments are
regarded as impervious to change through the procedure.
Let me briefly sketch how these features fit together. As I shall use the
term, “reflective equilibrium” (“RE”) is a relation holding between the
considered moral judgments (!CMJs?) of one or more persons. The
relation is one of consistency and coherence, in a sense which we must
return to discuss. Suffice it now to say that judgments in RE are ordered
in such a way as to constitute a theory: the set of judgments has a logical
structure. Thus the set is temporarily stable, at least in one sense: We are
not aware of any internal inconsistency which compels change on logical
grounds.
Depending on the scope of the RE, the judgments involved may include a
small or large part of the person’s total body of moral judgments. A total
RE will include all the moral judgments of a person. Our main concern
will be with limited REs, subsets of these judgments on a specific area in
moral philosophy. The judgments involved in a limited or total RE will
prirnarily be principles and particular judgments concerning ethics. This
may be called a partial RE. Å wide partial RE may also include judgments
belonging to other theories: “background theories” of the person, of
social psychology, biology etc, and theories of (moral) epistemology,
standards of relevance for moral arguments, theories about the nature
and force of moral utterances, and standards of coherence and
consistency. We shall sometimes want to talk about an ideal RE. Such a
state is a wide and total RE where the set of judgments is coherent and
consistent with the person’s complete ideal theory of the world,
accounting for all her experiences.1 . All the judgments are stable: they
will not be withdrawn because they conflict with other judgments or with
our experience. They are based on true beliefs, and true accounts of the
world, however “true’ is defined.
The “Method of RE” refers to the way the structure between judgments is
achieved. Starting with a set of CMJ, at different levels of generality,
individual judgments are rejected or modified, and new ones are added,
according to rules of consistency and coherence until the set is in
reflective equilibrium. Thus the method of RE has a normative effect:
-2-
changes in judgments may occur in order to improve consistency and
coherence. In ethics such changes are likely to occur: when the attempts
to produce consistency are made available to the person making the
judgments, this may cause major changes in the set of judgments. Thus,
these judgments are very theory-dependent.
The set of CMJs in RE — and each CMJ — is justified, in a practical sense,
for the person at that time. If several persons agree on the judgments, the
RE will be justified for them. They will be able to justify their actions and
judgments to each other, on grounds they all assent to. Particular
judgments and principles can be justified “from above”: from principles
and second—order principles, respectively. Principles and second—order
principles can also be justified “frorn below”: by the sets of particular
judgments and principles that follow from them,
To clarify my terminology, let us apply it to Rawls’ theory of justice.
Rawls’ theory constitutes a limited RE: the judgments involved mainly
concern the social justice of the major domestic institutions of societies.
Re is assuming that the RE is somehow shared by the citizens. The set of
judgments in RE does not include a theory of moral truth, moral
epistemology or a theory of theory—constraints. Still, the theory relies on
judgments properly belonging to such theories. The RE is not claimed to
be ideal: the judgments are only claimed to be accepted in our time, by
many members of Western democratic societies. Thus the RE is liable to
change, as is the rest of our judgments, in light of new knowledge,
increased reflection and new proposed theories. Rawls is claiming that
his theory is a better RE, in the sense of being closer to an ideal RE, than
the other prevailing political theories — particularly utilitarianism (cf
1971 381).

1.3 Recent history of the theory of reflective equilibrium

Features of the theory of RE can be traced to Plato. However, I shall be
concerned here only with the developrnent of the theory in seven articies
and books ranging from the early 1950s to 1979, using the vocabulary I
have sketched.

John Rawls: “Outline of a Decision Procedure for Ethics,”

1951
Rawls’ main concern in this paper is to defend and work out the second
feature of the theory: The rational decision procedure for ethics which
can adjudicate between interests and proposed principles. This makes
moral judgments similar to scientific knowledge and inductive logic: They
are capable of the same sort of justification. That is: they can be shown to
be objective, in the sense of being supported by a reasonable and reliable
method (177, 19—96). Moral principles are held to be justified both from
above and from below.
The practical aim, search for agreement, is noted (186), but it is not clear
that this paper distinguishes this aim from the more ambitious search for
moral truth, as later papers will. Thus Rawls talks of attaining moral
knowledge at one place (177).

-3-
All the CMJs are taken to be particular moral judgments: they concern
particular cases, to which ethical principles are later applied (182—83,
186). The feature of anti—foundationalism is not clearly present in this
paper, especially not concerning the second—order principles presented
in the conception of competent judges. Rawls acknowledges that some
particular judgments may be rejected in light of convincing principles
(188). However, the attitude seems to be that there are deviations and
mistakes among the particular judgments that are discovered when
comparing alternative principles (189). That is their deviant nature seems
to be logically prior to and independent at the principles chosen — this is
not necessary for the theory of RE.
Israel Scheffler: “On Justification and Commitment,” 194.
With regards to the first feature of the theory, Scheffler points out that
the justification given (by RE) is relative to a person and a time (182): the
justification is not ultimate and final, but temporary. This, no doubt, stems
from his claim that justification primarily is relevant for behavior we are
responsible for: when assessing people’s responsibility, what matters is
what they knew (or reasonably should have known) at the time.
The systematic and holistic character of the method is pointed out. The
procedure seeks to maximize initial commitment to the system of
judgments adhering to coherence rules, while making changes in
accordance with a “principle of least action”. Scheffler also introduces
the claim in ethics that these second—order principles are regarded as
considered judgrnents.
The CJ are clearly held to be non—foundational. This implies that even
the coherence rules of a system and the method of RE itself —
“maximizing commitment” — need justification, since they are
theoretically subject to modification. He also notes that even though
congruence between RE is not guaranteed, striving for wide RE may
serve to settie disagreements (190), since he allows for major, systematic
changes in considered judgments.
Morton White: Toward Reunion in Philosophy, 1956
Morton White develops t relation between science and ethics, and
regards them as one unified structure. He extends Quine’s application of
Dunem’s thesis, so that the system holistically tested includes ethical
statements, and the experiences the system is tested against include
moral feelings. He concludes that there s no radical epistemological
difference in kind between science and ethics.
This position is certainly well worth developing, and seems to be a
promising way to integrate the theory of RE into an all—encompassing
theory. However, it is not required by the theory of RE. If the conception
of moral truth underlying this understanding of ethics must be either like
Tarski’s, er a constructivist conception, the theory of RE need not be
committed to White’s epistemological claim. We shall see that the theory
of RE presented in this paper is not committed to those alternatives, but
it certainly regards the method of RE as similar to the methods of
research in science.
White also holds that definitions are to be included among the CMJs.
Moreover, the definitions are not held to by synonyms. Rather, they

-4-
establish which reasons and arguments should be accepted in support of
ethical decisions (259). A definition is justified in the same ways as other
second—order principles: By our certainty of its truth, our conviction that
it does not contradict other judgments, and the fact that its acceptance
helps organize our experience (267).

Richard Brandt: Ethical Theory, 1959

Brandt’s “Qualified Attitude Method” is very similar to the method of RE.
Especially noteworthy for us is that Brandt explicitly includes ethical
principles among the CMJs (246). He also holds that the method itself can
be justified, not only by the reader recognizing it as the one she actually
uses, but also in the sense that arguments can be given for its various
features.

John Rawls: A Theory of Justice 1971

The practical aim of the method is made explicit in this back: The goai is
to arrive at reasonably reliable agreement in judgments in order to
provide a common coneeption of justice. Moreover, the equilibrium
arrived at is regarded as temporary, restricted by our present knowledge
etc (548).
In the book, Rawls claims that the moral facts are determined by the
principles of justice developed in the theory. He therefore seems
committed to a theory of truth holding that moral truth is not logically
prior to and independent of the theory—construction. This is a position
that is not required by the theory of RE.

John Rawls: “The Independence of Moral Theory, 1974

Rawls here refines or abandons the position about moral truth held in
1971: He holds that the theory of RE makes no claim about the existence
of objective moral truth, in the sense of whether moral truths “exist”
logically prior to the theory.2 2 If the theory of RE is regarded as
developing towards a complete integration with a holistic view of science,
as White envisioned, this position constitutes a demarcation against a
complete assimilation. This is not to say that such integration cannot be
done. Rather, the idea is that since claims to that effect are likely to be
contested, and a claim about integration is not necessary for the practical
aim of the theory, such a clai2 should not be made.
In this paper Rawls also holds that there are CMJs at all levels of
generality, concerning particular situations and institutions, about broad
standards and first principles, as well as formal and abstract conditions
on moral conceptions. The non—foundational nature of the CMJs is noted
— and he even seems to hold that the total set of particular judgments
can be rejected (8).

Norman Daniels: “Wide Reflective Equilibrium and theory

acceptance in ethics”, 1979.
-5-
Daniels elaborates on the role of wide reflective equilibrium, showing
how that can strengthen the coherence restraints, helping to settle
disagreements and allowing drastic theory—based revisions. He notes
that the theory of RE is a theory of justification, and not of truth — a point
many critics have missed. Moreover, he holds that the question of
whether a converging and stable wide RE has attained moral truth is an
idle question — which serves to underline the practical aim of the
method.

2 Considered moral judgments

The method of RE is both applied to, and yield, a kind of judgments called
“considered”: They are judgments made under certain conditions, namely
those conducive to the exercise of our intellectual and moral powers, not
unduly influenced by emotional agitation, strong personal bias etc. Thus
the judgments we start out with are not completely “pre—theoretical”:
cur choice is guided by a more or less explicit and worked—out “theory of
reliable judgment.” The theory—laden nature of our judgments is also
apparent e.g. when aiming at a RE concerning political theory, where we
are able to roughly pick out which judgments are relevant — even though,
of course, in the process of adjusting and constructing the RE we may
have to include or exclude judgments in light of the theory. We are
guided, even at this stage, by some vague moral theory. 3

2.1 Considered moral judgments of many kinds

The CMJs can be of several types: principles or general statements:
particular judgments — i.e. about particular states of affairs: and also
“second order judgments”, stating e.g. conditions on theories, methods of
reasoning, rules of inference etc.
In his 1951 paper, Rawls seemed to regard the initial set of CMJ as
consisting exclusively (or mainly?) of particular judgments, concerning
particular cases (cf. 1951 182. There seems to be no explicit position on
this in Rawls 1971). Thus P. Singer (1974 503n) and Lyons (1975 145)
hold that Rawls’ method of RE starts out with only, or primarily, particular
judgments. But this is not the position I am interested in defending, nor is
it, I believe, Rawls’ position in later works (cf. 1974 8). Thus Sidgwick’s
or P. Singer’s se’f—evident first principles might be encompassed in a RE
— at least as very compelling principles, if not acknowledged as infallibly
grasped moral truth.

2.2 The initial credibility of CMJs.

The CMJs we start out with have a weak privileged evidential status.
Israel Scheffler (1954 182) says that we have an initial commitment to
these judgments, Rawls (1974 8) holds that they have an initial credibility.
This need not be taken to mean that we hold these judgments to express
moral truths, but merely that i) we actually make these judgments, and 2)
we see no reason to think them unjustified, sifted as they are according to

-6-
an underlying theory of reliable judgrnents.

2.3 Non—foundational.
The CMJs are not foundational. That is, any considered judgnent is in
principle open for modification or rejection when confronted with other
considered judgments, eg theoretical considerations: If a particular
judgment belonging to the original set of judgments is incompatible with
a proposed general principle, it may be discarded. And a principle with
some degree of initial credibility can be discounted if it yields conclusions
incompatible with one or more particular judgments. Raz (1982 319)
claims that in his 1951—paper, Rawls was holding on to a foundationalist
view of considered judgments. There is some support for this claim: One
of Rawls’ conditions for “explications” of particular judgments, (Le. for
proposed principles covering those judgments) is that the consequences
of the principle is coextensive with the set of intuitive/pre—theoretical
particular judgments (1951 184). Here there seems to be little room for
principle—based modification of the particular judgments. However, at
other places in the paper Rawls seems to hold that judgments are non—
foundational: a principle is more reasonable to the extent that

it shows a capacity to hold its own (that is, to continue to be felt

reasonable), against a subclass of the considered judgments of competent
judges, as this fact may be evidenced by our intuitive conviction that the
considered judgments are incorrect, rather than the principle, when we
confront them with the principle.
Rawls 1951 188

The feature of anti—foundationalism is also expressed explicitly by later

writers — with regards to science, Goodman claimed that no sentence
need be certain: all that is required is that they have some degree of
initial credibility (1952). Israel Scheffler makes that point directly
concerning ethics:

no sentence is ever immune from withdrawal, or, what araount to the

same thing, is a necessarily fixed description of any particular fact of
experience.
(I. Scheffler 1954 181—182)

Brandt (1959 248—49) also makes similar early suggestions.

It has been pointed out that the degree of non—foundationalism increases
when the RE sought is wider, because the theory—based considerations
can force deeper revisions (I. Scheffler 1954 190, Daniels 1980b 26,
Nielsen 982 108).
Marcus Singer (1977 610—611) points to one anomaly in this area: the
non—foundationalism seems to conflict with Rawls’ claim that some CMJs
are “fixed points” (Rawls 1971 318). I do not agree. With Daniels (1979
267), and Elackburn (1981 178), I take this to mean no more than that we
are so committed to these CMJs that we cannot now envision a RE (or
perspective) wnich would not include these judgments. White made this

-7-
point early, concerning a priori statements:

The closest we can come to converging with the traditional view of the
notion is to say that an a priori statement is one that we believe quite
firmly and one that we therefore resolve to make immune to overthrow at
a given moment.
White 1956 288

2.4 Shared considered moral judgments

The central role of RE is for justifying actions. This requires that those to
whom the justification is given, share the CMJs taken as premises in the
justification. Thus, the considered judgments may be held by only one
person, if I seek to justify an action to myself (even though in this case,
we might not want to call it a moral justification, but rather perhaps a
prudential one) or by a group of people.
2.5 Rawls’ set of considered moral judgments
In Rawls’ theory, concerned with public, political issues, the assumptions
in the RE — i.e. the CMJs appealed to — are shared in the society. In a
total and ideal RE the set of considered moral judgments would fit well
with all the considered judgments shared in the society, not only the
moral. But Rawls’ theory is far more restricted. He claims (Rawls 1974 5
—6) only to be concerned with a part of moral theory, not moral
philosophy. That is, the main area is the structure of various moral
conceptions: how “the right”, “the good” and ‘moral worth” can be
arranged to form moral structures concerning domestic distributive
justice. In addition to the particular and general judgments, tne set of
relevant CMJs includes formal constraints on moral conceptions,
constraints on arguments for political conceptions of justice — expressed
as tne original position with a veil of ignorance — and beliefs about social
psychology. But the theory does not deal with problems central to the rest
of moral philosophy, e.g. not the analysis of moral concepts, the existence
of objective moral truths, the nature of persons and personal identity, the
normative and binding nature of morality — and, I assume, the nature of
moral argument. Still, the theory of justice connects with such theories,
and relies to some extent y on their conclusions.4 But the arguments for
those judgments are not part of the theory of concern for Rawls.

2.6 Problems with considered moral judgments.

I shall discuss two objections to taking CMJs as starting points for
developing a moral theory and for justifying the resulting RE, based on
the fact that they are subjective in several senses.
P. Singer and Brandt criticize the appeal to intuitions (i.e. considered
judgments) in ethics, because of certain facts about the genesis of these
moral beliefs: They are strongly affected by our particular cultural
tradition, and are often derived from clearly false religious, biological or
societal beliefs (P. Singer 1974 516, Brandt 1979 21—22). Someone
holding the method of RE might agree with Brandt that

-8-
 What we should aim to do is step outside our own tradition
somehow, see it from the outside, and evaluate it, separating what
is only the vestige of a possibly once useful moral tradition .from
what is justified at present.
Brandt 1979 22

But I would strongly disagree with Brandt when be goes on to claim that
the method of RE “in principle prohibits our doing this. It is only an
internal test of coherence, what may be no more than a reshuffling of
moral prejudices.” The method of RE does not prohibit us from aiming at
stepping outside our prejudices — Rawls’ “original position” is an
attempt. But a proponent of the method may hold that we cannot step
completely outside our tradition: we cannot test or assess data by
judgments that do not reflect a distinctive cultural milieu (cf. Nielsen
1982 110). A claim to the contrary might be held as nearly
incomprehensible:

Not only may one doubt if there are alternatives to relying on our
intuitions, on our considered judgments, and on our common
understandings of, and insights into morality; one may well wonder
how there could be such alternatives.
Shaw 1982 123
A response to Singer and Brandt may go along these lines: Someone
defending the method of RE is certainly not denying that many judgments
in the moral realm (as elsewhere) is caused by mistaken beliefs and
unhappy social circumstances. However, it is only possible to determine
which of our beliefs and judgments are mistaken in light of our own
theories, according to our own beliefs. Thus, when seeking total RE we
are able to reject certain of our judgments as unreasonably biased,
discriminatory or mistaken — i.e. we have reasons to distrust them, but
only on the basis of our other present theories.
Å second major line of objection is that, contrary to the assumption of the
RE—theorist, there is disagreement between people on CMJs. flow, as I
have introduced RE, there is no claim that in fact any single person has
ever had all his judgments in RE, not to mention a group of individuals.
What is required is agreement on a sufficient set of CMJs for the
justification to be a justification, and not just a deduction from
unaccepted premises. Moreover, the theory of RE is not holding that the
CMJs we start out with must be shared. Arguments according to the
method of RE can proceed from shared assumptions, or they can be
arguments ad hominem, appealing to judgments held only by the
opponent. In the first case the arguments will move towards a RE of
shared judgments. If there is such a consensus on the set of considered
judgments we start out with, the RE we end up with will be a social
product, a RE for a group, and not only for an individual (cf. Fleck 1983
19). In the second case no shared RE is guaranteed, even though it may
turn out that some positions are untenable and a shared RE evolves.
Some considered judgments may be shared — i.e. the REs may be
overlapping. This may be sufficient for the purposes. There is a claim,
then, that the method of RE can serve to increase tne (degree of) RE for

-9-
an individual or for a shared RE in a group. Certain arguments can be
made to which a group of people will agree, even though they may
disagree strongly on issues that aren’t among the premises of the
argument. Rawls seems to hold that there is sufficient agreement in RE
for his argument for the two principles for justice to serve as a
justification of them towards citizens of western democratic societies.
Now Nielsen (1982 107) points out that the required intersubjective
agreement fails to hold: There isn’t even consensus within our society on
the disapproval of racial discrimination and religious intolerance, deeply
imbedded considered judgments e.g. in Rawls’ theory. As an empirical
claim this is probably true. But I believe that it does not challenge the
theory of RE: We certainly disagree on some of the considered judgments
Rawls’ argument appeals to. But they are not now in RE. Rawls claim is
that if each of US were to hold a RE developed through the method of RE,
while having Rawls’ arguments available, our respective RES would have
all the required judgments in common. And it seems that some of the
present most blatant areas of disagreement would satisfy this claim. Thus
I think that a defender of the method of RE can claim, with some
confidence, that attempts to maintain e.g. racist judgments in a wide RE
will fail: the assumptions required run counter to biological, psychological
and sociological theories the racist has good reasons to accept. Thus, at
our stage of knowledge, racist beliefs can only be held by refusing to
strive for consistency and coherence among one’s beliefs — Le. by
refusing to develop a RE by rational arguments.
We should no doubt keep a skeptical attitude towards our CMJs. Thus,
with Lyons and Blackburn “We may wonder whether our shared,
considered judgments express more than arbitrary commitments we
happen now to share — even when they are impartial” (Lyons 1975, also
Blackburn 1981 176). This is why the considered ,judgments are regarded
as nonfoundational even when we believe that we have reached a RE. We
can never know that our RE is ideal, or that better alternatives may be
developed. However, even in non—ideal wide RE we should have no
reason to hold that our ,judgments are “arbitrary”: No other theory we
now hold gives us reason to suspect that they are unduly slanted or
biased. Moreover, each judgment will belong in a comprehensive system
of logically connected judgments — thus the ,judgment won’t be arbitrary
in the sense of not being constrained by any considerations or other
choices.

3 The Method of reflective equilibrium

3.1 Rational Adjustment
The method of RE has sometimes been criticized as being very vague and
unspecified (cf. Raz 1982 314): which standards are applied, and how
does the process of adjustment proceed? I shall hold that the method
applies the ordinary standards of rational acceptability (e.g. as discussed
by Putnam 1982 132—136). For two examples of how the method has
been applied, cf. S. Scheffler (1976), who argues against Nozick’s political

- 10 -
theory, and . Nielsen (1982), aganst a Nietzschean position.

Logical consistency is one central standard. Considered judgments that

are found to be logically inconsistent cannot all be included in the
reflective equilibrium. Thus, if principles yield consequences that are in
conflict with other considered judgments, the principle or the particular
considered judgments must be modified. Similarly, if higher—order
considered judgments rule out certain general considered judgments,
some modification must take place. But as several writers have noticed,
logical consistency is a weak constraint — that is, standing alone, it
provides very little guidance for what kinds of modification that should be
undertaken. This is especially so when there is little theoretical structure
among the considered convictions — i.e. when the particular judgments
seldom are instantiations of general principles.

A second standard applied in the method is to increase the logical

structure of the set of considered convictions which we start out with. I
shall call this standard coherence. This standard may require that the
CMJs be made more specific. Thus second order judgments, e.g. on what
constitute moral principles, are made more precise than they were at the
outset, or that they be applied stringently to all cases. Moreover, this
standard requires that new principles be introduced into the set of CMJs,
so that particular considered judgments become logically related to and
can be checked against each other. There are very few restrictions on the
introduction of these principles. Already in 1951 Rawls notes that these
principles have the status of explications (1951 184—185). They are
theoretical constructions. Specifically, no claim is being made that these
principles are 1) analyses of the meanings of the ethical terms used in the
relevant particular judgments; 2) what people have intended to assert all
along; 3) the actual psychological or sociological causes of the particular
considered judgments, held implicitly (e.g. unconsciously) by the judges
(possibly contrary to what Dworkin asserts (1973 22)). The main aim for
introducing these principles, then, is to improve the theoretical structure
of the set of CMJs. A principle introduced in this fashion is appropriate if
it enables a RE, and inappropriate if it does not.
Still, these two standards seem to underdetermine the direction of
modification seriously: starting with some CMJs, it seems that a
consistent and coherent system can be generated in very many ways, by
rejecting any number of the judgments and replacing them with new ones
which fit the principles and standards we pick rather arbitrarily. The third
standard reduces the scope of choice rather much: the method must
maximize credibility. The phrase “maximizing initial credibility” is used by
I. Scheffler, in an attempt to inject some flavor of a “correspondence”
requirement of theories into a coherentist account of ethical justification
(1954 182). The CMJs are judgments we have some confidence in from
the start: we find them credible, and they are felt to be certain by the
person making them. The alternative systems developed in attempts to
achieve RE will vary in the credibility of the judgments encompassed by

- 11 -
them. Some will include principles that yield consequences we do not find
credible, or which require that we reject many CJs we accepted, or which
we simply hesitate to accept. Thus this standard may lead us to refuse
some of the introduced principles. Some will have less strict rules of
consistency or coherence — and thus allow conclusions we lack
confidence in. We are to choose that RE which has the greatest total
credibility. To what extent this “maximization” can be quantified is a topic
I shall not go into. But I believe that this standard does serve to exclude
many proposed theories.
Furthermore, there is a requirement that the resulting system be simple,
a standard easier applied than defined. It seems mainly to play a role
when we are forced to choose between REs that seem equivalent
according to the other three standards.
The third standard illustrates that the method of RE is far from value—
free:
it is guided by the participant’s confidence in and evaluations of CMJs,
and the RE chosen may well be “rigged’ to fit these judgments. We are
taking our convictions seriously.5 But there is nothing in the procedure
itself which guarantees that the result, even for a single individual, must
be one unique RE. There may be no structure that can satisfy the four
constraints even to a minimum degree, or there may be more than one.

3.2 A problem of incommensurable REs.

How can there be more than one RE satisfying these standards when the
standards are to be maximized? Because comparisons between REs are
difficult. There are two reasons. 1) Different REs may be satisfying the
sets of standards to approximately the same extent, but one can be far
simpler yet less coherent than the other. Since we have no rule for
weighing the standards, the method of RE seems to be indeterminate
between the two REs.
2) There are some problems since the standards are applied from within
the REs. In particular, whether the theory maximizes credibility can only
be determined by seeing whether it accounts for those CMJs we have
confidence in in light of the RE itseif. At least two cases are of concern. a)
The scope of relevant CMJs will vary between partial REs: e.g. CMJs
concerning abortion will be part of some REs concerning domestic
distributive ,justice, but not of other REs, depending on the conceptions
of the person they rely on. How can the total credibility of the two REs be
compared? One way out of this problem is to use the partial RE with the
widest scope of relevant CMJs to determine the set of judgments that are
to be compared (— i.e. include the judgments about abortion in our case),
and then see whether the other alternatives, possibly in wide RE, yield at
least as high total credibility of CMJs in that scope.
b) The different REs may embody large “perceptual shifts”. That is, quite
different features of events may be pioked out as relevant aspects by the
C4Js in each RE. There is then a sense in which the CMJs, as well as the
events as described by them, are different for different REs on the same
topic, e.g. justice. I have in mmd cases like legal pornography being

- 12 -
regarded as an expression of the government’s endorsement of
discrimination against women ar as a government—protected instance of
free speech. Another example is taxation to provide food stamps. This
may be regarded as (below) the minimum requirements of justice towards
the poor, or as legal yet immoral theft fram the other citizens. Problems
arise in such cases because we must compare the credibility of the CMJs
belonging to each RE. It does not seem possible to do this fairly for other
REs than our own: our convictions about our moral judgments are the
results of habituation and psychological upbringing, and not something
we can shift away from at will by considering a theory. We may be aware
of this, and therefore not want to reject the other REs because they yield
lower total credibility. We may be left with several somewhat
incommensurable REs, several of which we may regard as morally
acceptable. Total convergence is therefore not guaranteed when applying
the standards.

3.3 The claim of convergence.

But to what extent is this a threat to the method of RE? Starting with I.
Scheffler, many advocates of the theory of RE and similar theories have
noted that it does not guarantee convergence (Scheffler 1954 189—90,
Rawls 1971 5, Daniels 1980b 25, Blackburn 1981 176). The method is
claimed to aid in achieving RE, but not in all cases, and it is not a perfect
procedure in the sense that it guarantees that we end up with an ideal
RE.
The method requires that we start with a set of CMJs. When several
individuals are involved, it requires that the set of CMJs is shared (at
least in the sense that central disagreements can be settled on the basis
of other shared ,judgments). Thus the method is not claimed to be of help
where there is deep and fundamental disagreement between people: It
proceeds from agreement, and is aimed at dissolving disagreement in
certain specified areas.
How does the method remove disagreement? First of all, it serves to
clarify and organize our CMJs, enabling us and our opponents to
determine more exactly whether we actually do disagree: the method
helps to remove misunderstandings, vague formulations and what we
ourselves come to regard as misapplications of principles. The
clarification may help to pinpoint the disagreement to particular issues,
factual assumptions or particular trade—offs. The disagreement may also
turn out to rely on different opinions concerning what are to count as
relevant considerations on the issue. A large and apparently unsolvable
disagreement may thus be identified as concerning a second order
principle. This may be easier to deal with.
Moreover, the method serves to extend the area of agreement. By
applying the method to points of uncertainty, or even disagreement, it
may turn out that one judgment is preferred over other alternatives. This
is because that ,judgment follows from principles we all are prepared to
accept, or because its consequences are more appealing than the
consequences of its rivals, or because it conforms better with the
standards applying to judgments of its kind that we all affirm. The method

- 13 -
may help in locating the points of disagreement. Sometimes they may
turn out to be in social theory, biology etc. — outside the area of moral
theorv. Daniels notes that the method is especially valuable when the
disagreement can be placed in background theories, where the
disagreements may be more tractable (1979 262).
We should remember that the method of RE is not claimed to generate a
shared total RE: The scope of shared considered judgments we end up
with will depend on the scope of judgments we start out with. Thus, if we
start with some shared judgments in political philosophy, the method may
be successful on its own terms even though the participants end up
disagreeing seriously on other moral issues.
Thus the method does not guarantee congruence. In some areas this may
be a serious fault: e.g. in the sciences, there seems to be a very strong
considered judgment that the methods employed should yield one unique
theory of the world. However, in other areas it is not clear that we require
complete congruence between inembers. Even though many moral
systems include as a considered judgment that there is only one correct
or true moral theory, this does not seem to be universal: Thus Blackburn
holds that we don’t need the idea that at the end only one position is
morally acceptable: We may acknowledge several views as at least
blameless. That is, several RE are at least as good as mine (1981 177).
Moreover, in the limited scope of political philosophy, it does not seem
that there is a considered judgment that all moral (or even political)
theories should converge completely: Pluralism with regards to moral
theories is often seen as unavoidable, and sometimes even as a good
thing. Ånd pluralism in political philosophy may also be acceptable, as
long as there is a sufficient overlapping concensus for the purposes we
are pursuing.
On the other side, the method is not holding that lack of congruence can
in fact be determined at any point. I. Scheffler points out that we can
never determine with certainty that the method of RE cannot solve the
disagreement:

We cannot determine with finality at any given time, regarding any

given disagreement, that we have exhausted rational means of
adjudication and gotten down to the rock bottom of all relevant initial
commitments. Theoretically, we may always continue to expand our
attention, originally focused on the circumscribed area of conflict, so
that it takes in more and more of the
totality of our acts. We may hope to encounter some area of shared
commitnent, of systematic sentrality, such that the original
disagreement will be overshadowed.
I. Scheffler 1954 190

Similar points an made by others - e.g. by Nielsen, who claims that we

can never know that a concept (e .g . that of justice) is “essentially
contested”, that there is a necessary breaking off point where rational
argument must come to an end (1977 307); or Blaokburn, who says that
there is no decision procedure for deciding when our divergence of
opinion is “blameless” (1981 177).

- 14 -
3.4 The method of RE va a holistio viev of scienoe.
Several writers have discussed the relation between the method of RE
and holistic views of science which regard science as proceeding
according to a hypothetico-deduotive method. In the 1950s, White
presented the method of RE in ethics as integrated with science, and
similar approaches have been made since. Generalizations and principles
in theories an proposed as hypotheses. They are tested by checking
whether their consequences fit each other and our experiences, and are
modified — somewhere — as required. Thee hypotheses are justified from
below, by having consequences fitting our experiences. Moreover, certain
observations may be rejected is they do not fit our theories - here there is
an aspect of theory-dependency of facts. But there is also a strong aspect
of justification of observation sentences front above: When pressed to
justify them, we will turn to a theory of perception about the causes of
our sensory stimulations, normal conditions of observaton etc to explain
why the sentences are not only believed by us, but are credible — i.e.
causally related to objects out there’.
White, and later Dworkin, points out that in both ethics and science
revisions to accomodate experience may occur at several places: different
principles may be modified in order to generate the required fit. But
Dworkin also claims that the method of RE, as opposed to scientific
procedures, allow some considered judgments to be ignored for the sake
of consistency (Dworkin 1973 32). I believe that this is incorrect:
Certainly, science can dismiss certain data as fallacious, suffering from
errors in measurernent etc and also, science may regard some
observations as inexplicable at the present stage of knowledge (cf.
Daniels 1980b).
Brandt claims that an important distinction between science and the
method of RE is that science is based upon what scientists see, not on
what they think (Brandt 1979 236). However, I believe this distinction not
to be valid. It rests on the claim that what the scientists see — and what
they report — is completely independent of whatever theories they have:
i.e. that certain kinds of experience take place without influence from
theories, expectations, and general world view. Certainly the methods
used, both in science and ethics, rely on theory—laden data. Shaw,
defending Brandt, claims that Quine holds a weak epistemic
foundationalism: The observation sentences are a kind of foundation,
linking science with observation, thus giving science a stronger claim to
truth and objectivity (Shaw 1982). Two comments are appropriate. First,
it would be a misreading of Quine to hold that he regards the
observations sentences as logically prior to theorizing: Not only do we
rely on hypotheses when picking out which noises to regard as words (cf.
Quine 1960 68), but we must decide what to treat as signs of assent and
dissent and use these assumptions as a working hypothesis (cf. Davidson
and Hiinikka (eds.)
1969 312). Secondly, the observation sentences are primarily identified as
sentences to which most of the community will react similarly:

- 15 -
 What qualifies a sentence as observational is not a lack of such
[theoretical] terms, but just that the sentence taken as an undivided
whole commands assent consistently or dissent consistently when the
same global sensory stimulation is repeated.
(Quine 1981 26)

an observation sentence is one on which all speakers of the language

give the same verdict when given the same concurrent stimulation.
(Quine 1968 86—87)

What counts as an observation sentence varies with the width of

cornmunity considered. But we can also always get an absolute
standard by taking in all speakers of the language, or most.
(Quine 1968 88)

There are several similarities here with the method of RE: they both
proceed from considered judgments shared within the relevant group,
according to standards there is consensus on. Two differences remain,
which account for the lack of correspondence: Ethics lacks a generally
accepted ontology to give a causal story for the considered judgments,
and the considered judgments are very theory-dependent. Even though
Quine may be more foundational, the claim to truth (to the extent that he
is making such claims) and objectivity is ultimately based on the
observation sentences — to which the considered judgments correspond.
Even though there are several points of similarity, there are also
important dissimilarities. It is elear, I think, that the considered
judgments are more theory—dependent than observation reports in
science.
4oreover, the RE concerning moral theory does not contain a causal
theory about moral judgments which can account for e.g. the genesis of
our moral judgments, in particular those we challenge. Such an account
seems to belong to sociology or psychology, rather than to moral theory.

Furthermore, the scientific method is often regarded as proceeding from

observation sentences about particular phenomena to high—order
principles: the basis is particular judgments. But the method of RE starts
from considered judgments on several levels of generality.
Finally, and perhaps most importantly, RE is only claimed to be offering a
practical justification of judgments: the judgments are not claimed to be
true in any strong sense, but only to be shared by us now. This is all that
is required for its aim: to settle certain points of disagreement between
us according to rational standards we accept.

3.5 The charge of conservativism

Peter Singer has claimed that the method of RE is a very conservative
process, since it aims at establishing consistency among our particular
moral judgments. We are assuming that at least some of the judgments
we now make are correct in some sense, and strive to construct a theory
incorporating them. But there are, Singer claims, good reason to believe

- 16 -
that none of our particular judgments stand up to criticism — “it would be
best to forget all about our particular moral judgments, arid start again
from as near as we can get to self—evident moral axioms” (P. Singer 1974
516).
The response from the advocate of the method of RE will be that Singer’s
argument may be valid: All, or most of, our particular judgments nay be
faulty, and should be discarded. Moreover, we might agree that under
some circumstances, we are able to determine this by our present lights
and theories about human psychology etc: Drastic revisions ought to take
place. But the method of RE allows for this: these conclusions can be
drawn by applying the method of RE itself — they are not a point against
it (cf. Rawls 1974 7-8). The reason for this is that the method proceeds
fron CMJs held at all levels: not only particular moral judgments (as P.
Singer believes), but also higher—level principles and judgments
concerning methods of reasoning in ethics. Systematic overhauling at
many levels may occur (cf. e.g. I. Scheffler 194 189). Rawls atoane point
even makes the very radical claim that even the totality of particular
judgments can be rejected (1974 8). Moreover, by working to achieve
wide RE, the scope of revision is increased, as Daniels points out (1979,
1980a). Finally, since the method of RE proceeds by appeal to publicly
shared considered judgments, the (wide) RE is not used to express a form
of false consciousness: it is based on shared beliefs about how the social
structures function, a shared political conception of the person and the
legitimate purposes of society: Nothing needs to be hidden (c.f. Rawls
1980 39).

4 Justification by reflective equilibrium

4.1 The practical nature of the justification
The CMJs in RE are heid to be justified in a certain sense. They are
claimed to be justified for certain persons —e.g. those holding the RE - at
a certain time. Thus the justification is person—relative and relative to a
specific time. Are these limitations acceptable?
Yes, because the justification by RE serves a practical purpose. Insofar as
the method of RE is applied to solve disagreement between persons, it is
claimed to be of he’p only when the CMJs we start with are shared. In
these cases there are reasons to believe that the area of agreement will
be increased, hopefully sufficiently for the purposes. Thus Rawls writes
that

The practical aim is to reach a reasonably reliable agreement in

judgment in order to provide a common conception of justice.
Rawls 1971 44—45

Now there are obvious limitations on the RE we end up with. It will only
be the best theory among those alternatives we consider — Options we
are not aware of may of course be better. The result is relative to the
basic azreement we start out with, thus the result will be limited by the
present state of public knowledge, and it may possibly be terribly biased,
- 17 -
but only in ways we are unaware of.
No claim is made that the RE we justify in this way contains moral
judgments that are true. That is beyond the intended goal of the method
of’ RE: Agreement according to rational standards is all we need. On the
other hand, even though we seek agreement, we would not be willing to
agree on a proposed RE we knew rested on false factual claims, or which
we otherwise had good reason to believe was false. Such a theory would
not satisfy the standard of rationality in wide RE, and would not be
stable.
Justification by RE has been presented as only practical, not
epistemological. It is not clear to me whether this detachment from
epistemological issues have been held from the early 50’s Note that RE
might be held to give justification in a stronger sense also, and I think
good arguments can be made for that claim. But for the practical
purposes, this practical justification is all we need.
A particular judgment is justified from above, from the principles and
conditions it instantiates. That is, a particular judgment is justified by
being “deducible” (more or less strictly) from higher—level principles. A
higher-level principle may in its turn be justified from yet higher—level
principles or second—order principles: conditions on principles,
background theories etc. But it is also justified from below, from the set of
particular judgments following from it.
The second—order judgments — the conditions on principles etc — are
justified from below, by the system of judgments they make possible.
4.2 The threat of circularity
Particular judgments are justified by principles, and principles are
justified by particular judgments, yet this conception of justification is not
viciously circular, and thus not an obvious misinterpretation of the
method, as Kulenkampff holds (1979 97). Each single particular judgment
is justified by its relation to (a) higher—level principle(s), while each
single principle or condition is justified by the set of judgments that
follow from it, which have initial credibility.
Tugendhat fails to see that the RE provides justification both for
principles and particular judgments, since he does not notice that
particular judgments are justified from above. Thus he claims that Rawls
unwarrantly assimilates “ist person moral theory” to “3rd person moral
theory”, thereby blinding himself to the fact that “our [particular] moral
judgments are items which, according to their own sense, do not form a
court of appeal, but are in need of a court of appeal” (Tugendhat 1979
81). Our own moral judgments — i.e. in a 1st person theory — can only be
justified by principles, and this task of justifying the particular judgments
themselves is ignored by this account, Tugendhat claims. It is only in a
3rd person moral theor’ that conjectured principles are to be checked
“against the particular judgments of those persons whose sense of justice
is being studied” (1979 81).
It seems that many writers fail to notice that the conception of
justification presented functions even when the principles and the
particular judgments appealed to depend on each other for support:
Coherence and consistency provide justification: i) a principle provides

- 18 -
justification for a judgment even when the principle is only justified by
the set of judgments to which the particular judgment belongs, and 2) the
particular judgments deducible from the principle can serve as a test for,
and justify, the principle, even if they are not chosen or determined
completely independently of it. Nagel and Dworkin disagree with 1), it
seems, when they require that the underlying principles justifying
particular moral intuitions should have independent support (Nagel 1973
3, Dworkin 1973 23). Daniels also seems to disagree with 1) when
discussing the role of background theories: He holds that coherence of
the principles of justice with background theories is in itself no
justification: Increased support can only come if the background theories
are somehow independently supported (1979 259). I certainly agree that
the background theories provide a stronger justification for the principles
of justice if they have independent support (which, by the way, may come
from considered judgments of many kinds, not only the moral ones, which
Daniels seems concerned with), but if my account is correct, coherence
with background theories itseif also provides justification for the
principles.
Marcus Singer, on the other hand, disagrees with 2), since if particular
considered judgments “are to play a role in the justification of the
principles, then one would think that they have to be themselves found,
and not themselves chosen. ... [but] the moral facts [i.e. the particular
judgments] are determined by the principles which would be chosen ...
“(1977 612). The justification provided by RE, however, does not require
that the set of considered judgments justifying the principle is
determined and justified completely independent of the principle.

4.3 Justification of the method of RE

Are the standards of coherence, consistency credibility and simplicity, and
the conception of considered judgments as non-foundational also justified
by the RE? Should they be? Concerning the second question, Tugendhat
claims that Goodman’s view on justification of principles of inductive
inference, referred to by Rawls, is properly compared to the justification
of the procedures of moral argument, and not the justification of
particular and general considered judgments (Rawls 1971 20n, Goodman
1955 65—68, Tugendhat 1979 81/82). Tugendhat criticizes Rawls
because Rawls never seems concerned with this task, to justify the
methods employed. Raz criticizes Dworkin on a similar point, since
Dworkin holds that we have a responsibility towards making our system
of particular judgments coherent (Dworkin 1973 28): Such a
responsibility “should be justified by the method of reflective equilibrium
and not be used to justify that method, since if such a responsibility exists
it is but one of our moral responsibilities” (Raz 1982 323). I think Rawls
and Dworkin can be easily defended by claiming that they are not
concerned with providing a complete moral philosophy: their concern is
primarily a limited RE within moral theory, and it is beyond that scope to
justify all assumptions relied on. In particular, this would hold for the
assumptions underlying the method of RE.

- 19 -
But if we venture beyond Rawls’ and Dworkin’s concern, how can we
justify the method of RE in a more total RE? I find the comparison with
Goodman’s view appropriate. I. Scheffler claims that the other standards
of the method of RE are justified because they maximize initial
commitment (1954 188). He furthermore holds that the standard of
maximizing commitment is itself what “rational justification’ means. This
claim, in turn, can only be tested by seeing whether it fits with our
practice: “Only a thorough testing of the present proposal will reveal
whether or not it is accurate’ (1954 189). The standards of consistency,
coherence, credibility and simplicity are among our considered
judgments, since they seem to govern much of our practice in moral
argument. They are justified by the theories they allow us to construct —
both RE and in other fields of inquiry. Thus, the standards are not
justified by appeal to something outside our structure of ,judgments and
beliefs, but by showing that they are part of this structure. This gives us a
response to e.g. Brandt’s challenge:

How might these philosophers justify their view that a normative

principle should be tested against our intuitions in the sense of our
considered beliefs, or a given intuition be tested by its incompatibility
with a system of normative beliefs…?
Brandt 1979 20

The justification provided by RE is that the method of RE is in fact how

we proceed in moral argument, and that this inethod tends to increase
agreement in certain cases we are interested in. But is this correct? In
particular, when discussing moral issues, do we hold all moral judgments
to be non-foundational? Again, as an empirical claim this seems false. But
I am inclined to think that at least on certain issues, perhaps mainly
concerning justice and interpersonal relations, there are arguments to
the effect that we ought to do so. That is, for the sake of arriving at
rational agreement on certain important issues, we should, in arguments,
be willing to refrain from appeal to convictions we have that aren’t
shared and that we cannot justify to each other. Thus, the non—
foundationalist nature of considered judgments is itself a considered
judgment we attempt to justify.

4.4 The role of the original position

We are now in a position to see how Rawls’ use of the original position
relates to justification by RE. The OP plays a role within Rawls’ RE as a
device for achieving a high degree of coherence within the set of
considered
judgments. This will also strengthen the tendency towards convergence if
accepted by the disagreeing parties. The OP is justified both from above
and below, and justifies principles and particular judgments from above.
It is a heuristic device for bringing together various restrictions and
conditions on oral arguments widely shared among us. For instance, it
presents and specifies ideas we share about treating each other as free,

- 20 -
equal and autonomous moral agents, about the kinds of arguments we
consider relevant for political principles of domestic distributive justice,
and the restraints we find reasonable to put on moral principles. The
considerations presented as the OP allow a “considered rational choice”
of principles of justice. The mechanism is hardly stringent enough to be
called a “derivation”. Nevertheless, the combining and refining of second
—order considered judgments can guide the choice of principles from a
list in part generated by earlier efforts at establishing moral theories. By
careful pruning and rigging different principles can be generated — that
is: different interpretations of the “initial situation” supports different
choices of principles. Rawls’ claim is that the interpretation called 0P, and
its corresponding principles, will be preferred by the method of RE over
certain other combinations.

4.5 Reflective equilibrium and objectivity

The second—order judgments, e.g. as presented in the original position,
allow us to step “partly outside” the set of CMJs we start out with. They
allow us to strengthen the coherence of the RE. Some of them, the
standards of rationality, also enable us to evaluate the consistency
between particular judgments, principles and other second—order
principles (standards for moral arguments etc.) more critically, by
enabling us to stand back frem our moral sensibilities and check whether
we can endorse them (cf. Blackburn 191 171—2, 175—6). It provides an
objective point of view in a certain useful sense: It is impartial, or as c’ose
to that as we can come under our conditions. That is, it restrains the
kinds of arguments offered for moral principles in ways there is
consensus on. (Snare 1976 107—8, Singer 1977 305—6, Rawls 1980
554f). The reasons for principles that can be presented in the OP are
“objective” reasons, in the sense that everyone can (reasonably) accept
these reasons. It is not a completely independent Archimedian point - that
cannot be achieved - and is not what we need, nor what we are looking
for (against Nielsen 1982 110). Insofar as the practical aim of the method
of RE is to arrive at rational agreement among parties, we need some
notion of objectivity, in a sense of being in accordance with reasonable
standards everyone can agree to.
Note that the OP is only claimed by Rawls to be relevant in his context:
domestic social justice in democratic societies. For other problems, e.g.
concerning duties towards animals, or in international contexts, other
interpretations of the initial situation may be appropriate. Thus standards
for acceptance as relevant parties must be worked out — or the whole
notion of “contract” may be irrelevant. This does not show that the OP is
inappropriate in Rawls’ context — that is a question of whether the
constraints and conceptions involved are part of the best RE available to
us concerning Rawls’ issue: domestic social justice.
So the conception of OP is a part of the RE, and the justification provided
for principles from the OP is a part of the justification by RE. This
presentation may guide our response to certain criticisms raised against
Rawls’ use of the OP.

- 21 -
Lyons maintains that the OP goes beyond justification by coherence. He
holds that the contract argument involves a rigorous deduction, and
avoids the circularity of the coherence argument by grounding the
principles of justice on an independent theory oi rational choice. This
makes the contract argument more powerful than the coherence
argument (Lyons 1975 152).6
Marcus Singer charges that the ap is an anomaly in Rawls theory (1977
611-612): Ideally, the principles are to follow deductively from the OP, and
that this justifies them. However, derivation is ,justification only if the
premisses (i.e. the OP) are justified, but they are not, so the principles are
not justified. This misses the point that the “deductive” argument from
the OP is embedded in the RE as a whole. The ap is justified partly from
above, partly from below; it is the mutual support of all the CMJs which
justifies the principles and everything else.
Snare holds that Rawls has two different methods of justification: by RE,
which is explicitly accounted for, and the other one actually at work in
Rawls’ argument for the two principles of justice: from the OP (1976
102). Thus

the appeal to the original position [is] a quite separate method

justifying moral principles. It would not itself be a matter appealing to
our considered convictions even if it can be supplemented by such a
method.
Snare 1976 110

Kulenkampff also seems to misread Rawls to the effect that two attempts
at justification are at work. They yield the same result, but if I understand
him correctly, he holds that the argument from the OP has logical priority
over the argument from coherence (1979 91). But the OP is not a
stronger and independent kind of justification which happens to support
the same set of principles as the RE does; rather, the justification from
the CP is subordinate to the overall RE.

4.6 Reflective equilibrium and moral truth

A series of writers have criticized the theory of RE along these lines:
The proponent ol RE is missing a major point of criticism completely.
Surely the aim of theory construction, especially in ethics, is truth, and
not merely consensus? Consensus, especially arrived at through the
method of RE, does not guarantee truth — in fact, no argument has even
been made yet to the effeot that RE is evidence of truth. Both shared and
“objective” judgments can be false, since considered judgments “justified
for us today” may well be false. The theory of RE is misinterpreting the
central concepts of justification and objectivity, and neglecting central
arguments wflich are required if tfle central claim of the theory is to be
justified, namely that RE constitutes moral truth (cf. Hare 1973 82—83, P.
Singer 1974 493—4).
The first line of defence against this criticism is that tie theory oi: RE is
making no claims about truth, and is in fact not even concerned about

- 22 -
whether the RE contains true CMJs, as long as we are not in doubt about
them now. The aim of the method of RE is primarily practical: achieve
consensus on specific areas of present disagreement that we (or others
applying the method) find important to resolve. For this purpose
justification for specific persons at a specific time is important, as is
objectivity by our standards, and the shared beliefs held to be public
knowledge by us. The practical aim can be achieved without appeal to
“absolute” justification, objectivity as being in accordance with moral
truth, or “really” true beliefs. There is a risk that we will agree on moral
falsehood, however understood, but the RE will be the best result we can
arrive at given our state of knowledge, and given that the appeal must be
to shared considered judgments.
Let us assume that we correctly believe that we have certain true moral
intuitions. What do we do if our fellow human beings are mistaken? As
long as the intuitions aren’t shared, and we cannot argue successfully for
them, we cannot achieve a consensus by them. It seems to me that if we
are in this situation, and regard reaching agreement on the contested
issues as very important, we shall be led to argue by the method of RE —
the alternatives seem to be to live with disagreement or to rely on force.
And to the extent that the true CNJs are shared among members of
society, we are free to rely on these correct intuitions when proceeding
according to the method of RE, and we will hold these CMJs fixed. We
may find, of course, that we do not wish to revise certain CMJs, and we
may also decide that we want to modify our second—order considered
judgment concerning the method of RE about non—foundationalism.
Moreover, the RE established in response e.g. to Rawls’ political concern
is not to be blamed for not containing a theory of moral truth. Such a
theory lies beyond the scope of moral theory, and is to be developed in
moral philosophy (cf. Rawls 1974).
But does the theory of RE commit us to a specific theory of (moral) truth?
Let us distinguish two schools in theories of truth: Realism and
Constructivism. On the realist view, moral facts and principles exist
logically prior to and thus independently (in some sense) of our theories.
It is in virtue of their closeness to these facts that some theories are
better than others. True moral judments are true because they
correspond to such moral facts, and methods of moral reasoning are
correct are appropriate to the extent that they reliably lead to moral
truths (I consider Rawls’ conception of rational intuitionism to belong
here — cf. Rawls 1980 557).
Constructivism, on the other hand, is agnostic with regards to the
existence of logically prior moral truths. A “radical” version of it might
even reject the view that it makes sense to at all talk about anything
outside our (compete) theory, to which it should correspond: Moral facts
and principles, if needed, can be posited within the theory, as our
constructions, in order to enable a consistent, comprehensive and simple
account of our experiences. All constructivist theories, not only “radical”
ones, will agree that there seems to be no practical need to assume
logically independently existing moral principles. Therefore, whether
there are such moral truths is irrelevant for the theory. For practical
purposes, truth might be considered an intra—theoretical notion — i.e.
- 23 -
judgments are true—in-our—RE. If needed, e.g. to account for progress in
moral argument and judgments, we may operate with an ideal theory,
covering all our experiences and structured according to ideal standards
of rationality, towards which we are approaching. We may need to
stipulate that there is but one such ideal theory, but again, we may not.
The judgments contained in this theory are “really” true, and the
justifications given in that theory are “really” objective.
Now, does the method of RE implicitly require that one and only one of
the two general schools of theories is correct? Rawls claims that the
method is independent of epistemology; that is: no plausible theory of
epistemology (or theory of meaning, philosophy of mind or their
combination) implies that any otherwise plausible moral theory is false
are true. Thus the choice of moral theory is underdetermined by the rest
of moral philosophy. And epistemology is relatively independent of the
method of RE: Whether it is correct or not will not settle any
epistemological issues7 The method of RE is not assuming that there is
exactly one correct moral conception, and is therefore neither assuming
nor rejecting the existence of objective moral truths (Rawls 1974). The
method at RE is not assuming that some realist theory of truth is correct.
But others have held that the method of RE, as any theory of justification,
is committed to a realist theory of truth: A method af justification can only
be justified or seen as reasonable by showing haw it (in most cases)
brings us closer to the truth by increasing the correspondence with the
independent facts. only in this way can it be established that previous or
competing views are less reasonable. Two responses seem available. It
should be kept in mind that the method of RE is a method of justification
only in a practical sense — it is not claiming that judments in our present
RE are true. And moreover, it seems to me that the constructivist can
give a satisfactory account of moral theories being better and worse (by
her standards!) by showing both how certain methods of justification are
in fact pursued by us, and that they improve the coherence, consistency
credibility and simplicity of the theory by the theory’s own lights: how the
method aids in removing false beliefs and mistaken judgments, replacing
them with true and better ones — according to our standards, beliefs etc.
Another argument to the effect that the method of RE is assuming the
truth of some realist conception of moral truth is that the best or only
explanation of convergence, e.g. in ethics, is that people converge around
the truth, recognizing the independent moral order (cf. Daniels 1979 278,
Rawls 1983 26). However, I believe that the grounds sketched above (in
3.3) also can explain what convergence may occur between persons
having many shared judgments, without appealing to moral truths (cf.
Rawls 1983 26)
Some writers have objected that the method of RE is incompatible with
certain realist theories of truth. Dworkin holds that the feature of non-
foundationalism in the method rules out a “naturalist” theory of morality.
Such a theory claims that a given moral theory describes an objective
moral reality discovered by a moral faculty. The moral reality causes
concrete (particular) intuitions which are clues to the nature and
existence of moral principles, just as physical observations are clues to
physical laws (1973 27). But it is unclear to me why this theory is
- 24 -
incompatible with non—foundationalism, unless the moral intuitions are
held to be incorrigible. And that is certainly not a feature of physical
observations: they can be set aside if no account can be given; either as
misperception, or as a yet unsolved problem for science. Dworkin seems
mistaken when holding that the method of RE, but not science, is
authorized to set aside former intuitions without giving an account for it
at that time (1973 33). However, be is right to point out that such
explanations would legitimately be expected of science, while moral
“misconceptions” need not have a moral explanation: they can properly
belong to psychology, sociology etc.
The method of RE seems compatible with constructivist accounts of truth.
It might be noted in passing that the method of RE is not by itself such an
account. The theory of RE concerns justification of considered judgments,
just as coherence—theories in science are theories of justification, and
not theories of truth.
Rawls sometimes mentions an ideal RE (1971 ?), which might seem
difficult to relate to the reading of justification by RE as practical: If our
problems and disagreements can be settled in a RE for us, what is the
role of an ideal RE? It may simply serve to remind us that his theory is
not claimed to be an ideal RE, and that it is open to modification as we
develop his proposal. But it may also support the claim that Rawls is
holding some constructivist theory of moral truth if he also holds that we
attain true beliefs in the background theories by the method of RE
applied to non—moral theories. This is a claim outside the scope of this
paper. In any case it seems that Rawls’ claim about the independence of
moral theory concerning theories of truth is acceptable, and that the
theory of RE can be seen as not being concerned with epistemic
conceptions of justification, but only with a practical conception.

5 Conclusion
In this paper I have presented the theory of reflective equilibrium as a
theory of practical justification in ethics. I have tried to discuss the
central features of this theory as I conceive of it, and to show how it in its
present formulation stands up to criticisms presented during the last 10—
5 years.
It is my opinion that the theory meets this criticism well. This is partly
because the criticism has been directed against earlier versions of the
theory - e.g. concerning the possibility of major revision of present
considered judgments. Since some features of the theory were absent, or
at least not clearly stated, in the early papers, some criticism can now be
regarded as not concerning the present theory. This holds, it seems, for
criticism concerning the nature of justification when regarded as
epistemological, not practical. The claim made by the present theory
concerning justification is weaker than the claims made, by justification
in science. The question is then whether the claims made by the theory
are strong enough. It seems to me that the practical justification is strong
enough for reaching the aim of the theory: arriving at rational agreement
on specific moral issues. Moreover, I am convinced that such agreement

- 25 -
is often very important for us. The theory of reflective equilibrium serves
to rehabilitate the realm of ethics as a sphere of reason. It provides us
with an important and useful approach to pressing moral and political
problems facing us today, and should be applied.

References
Blackburn, S.W. 1981: “Reply: Rule—following and moral realism”, S. H. Holtzman and
C. M. Leich (eds.): Wittgenstein: to follow a rule, London.
Brandt, Richard 1959: Ethical Theory, Prentice—Hall, Englewocd Cliffs N.Y.
- 1979: A Theory of the Good and tte Right, Clarendon Press, Oxford.
Daniels, Norman (ed.) 1975: Reading Rawls, Basic Books, New York.
—. 1979: “Wide Reflective Equilibrium and theory acceptance in ethics, The Journal of
Philosophy 76 256-282.
—. 1980a: Reflective Equilibriuin and Archimedian Points”, Can J Phil 10, 83—103.
—. 1980b: “On some Methods of Ethics and Linguistics”, Phil Stud 37, 21—36.
Davidson, D. and J. Hintikka (eds.) 1969: Words and Objections, L.Reidel, Dordrecht
Holland.
Dworkin, Ronald 1973: “The Original Position”, in Daniels 1975, 16—53.
Fleck, Leonard M. 1984: “Rawls, reflective equilibrium, and justification: an
economic model” Paper at the APA Western Div. April meeting, unpublished.
Goodman, Nelson 1952: “Sense and Certainty”, Phil Rev 61 160—167.
-. 1955: Fact, Fiction and Forecast, Harvard University Press, Cambridge Mass.
Hare, R.M. 1973: “Rawls’ Theory of Justice”, in Daniels 1975: 81—107.
Kulenkampff, Arend 1979: “Methodenfragen der Gerechtigkeitstheorie: Uberlegungen
im Anschluss an Tugenhats ‘Comments on sorne methodological aspects of
Rawls’ “Theory of Justice”’”, Analyse und Kritik 1 90—104.
Lyons, David 1975: “Nature and Soundness of the Contract and Coherence argurnents”,
in Daniels 1975 141—167.
Machan, Tibor R. 1976: “A Note on Independence”, Phil Stud 419—422.
Nagel 1973: “Rawls on Justice”, in Daniels 1975 1—16.
Nielsen, Kai 1977: “Rationality, Needs and Politios: remarks on rationality as
emancipation and enlightenment”, Cultural Hermeneutics 4, 281—308.
-. 1982: “On needing a Moral Theory”, Metap 13, 97—116.
-. 1982b: “Grounding Rigtits and a Method of Reflective Equililbrium”, Inquiry 25, 277—
306.
Putnam, Hilary 1982: Reason, Truth and History, Cambridge University Press,
Cambridge.
Quine, W.0. 1960: Word and Object, MIT press, Carnbridge Mass.
-. 1968: “Epistemology Naturallzed”, in Ontological Relativity and other Essays,
Columbia University Press, N.Y. 69—90.
-. 1981: “Empirical Content”, in Theories and Things, Harvard Univeristy Press,
Cambridge Mass. 24—30.
Rawls, John 1951: “Outline of a Decision Procedure for Ethics”, Phil R 60 177—1 97.

- 26 -
-. 1971: A Theory of Justice, Harvard University Press, Cambridge, Mass.
-. 1974: “The Independence of Moral Theory”, Proc An Phil Ass 4.8 (1974—75) 5—22.
-.1980: “Kantian Constructivism in Moral Theory”, J Phil 72 515—572.
-.1983: “Lecture 0: Background Notions”, unpublished.
Raz, Joseph 1982: “The Claims of Reflective Equilibrium”, Inquiry 25, 307—330.
Scheffler, Israel 1954: “On Justification and Commitment”, J Phil 51, 180—190.
Scheffler, Samuel 1976: “Natural rights, Equality and the Minimal State”, Can J Phil 6,
56—76.
Shaw, William H. 1982: “How to do Ethics”, Metap 13,117—130.
Singer, Marcus G. 1977: “Justice, Theory, and a Theory of Justice”, Phil of Science 4.4.
594—618.
Singer, Peter 1974: “Sidgwick and Reflective Equilibrium”, The Monist 58 490—517.
Snare, Frank 1976: “John Rawls and the Methods ot’ Ethics”, Phil and Phen Res 100—
112.
Tugendhat, Ernst 1979: “Comments on sorne methodological aspects of Rawls’ ‘Theory
of Justice”, Analyse und Kritik 1 77—89.
White, Morton 1956: Toward Reunion in Philosophy, Harvard University Press,
Cambridge Mass.

- 27 -
1
The standards applying to the ideal RE may be similar to what Habermas calls “ideal
communicative conditions’ cf. Nielsen 1977 297.
—

2
In his 1980 paper Rawls may be read as reversing his position again: Re asserts that “there are
no moral facts” (1980 564). However, this can be read as “there are no moral facts available as
reasons to the parties in the original position, to which the principles adopted could approximate.
Rather, the principles single out which facts to count as moral facts, i.e. as reasons of justice.

3
It seems that Raz (1982 319—320) is reading Rawls to hold that we identify our considered
moral judgments completely independent of, and prior to, any attempts at formulating a moral
theory. This I hold to be a mistake.

4
— as Machan seems to misread Rawls as claiming (Machan 1976).

5
We are certainly not ignoring them, as Kulenkampff holds when criticizing the method: “Wenn
wir von unseren eigenen ethischen Präferenzen absehen dann müssen wir, denke ich, einrumen,
... ,

dass unter dem Gesichtspunkt der Kohärenz die verschiedenen Gleichgewichtszustände ethisch
durchaus gleichwertig sind.” (Kulenkampff 1979 97)

6
He furthermore holds that if the contract argument was a branch of the coherence argument,
this would strip the contract argument of justificatory force (159). I believe that he is misreading
the theory of RE as including only particular judgments, moreover that his conception of
justification is different from the one we are operating with this helps account for the latter
—

claim.
7
Even though Rawls holds that if we could determine the non—convergence of wide REs, this
would rule out theories holding that there is objective moral truth, c.f. 1974 9.