Meaning, Dispositions and Supervenience
Leonardo Marchettoni
ciclvm@tin.it
Jura Gentium: Centre for Philosophy of International Law and Global Politics,
Department of Theory and History of Law
University of Florence
Abstract: In this paper I raise some doubts about Brandom’s pragmatic strategy of explanation of
norms. I argue that Brandom’s attempt to explain normative statuses through recourse to normative
attitudes does not succeed in preserving a hiatus between norms and regularities of behaviour. Since
calibrating one’s own behaviour as a consequence of normative assessments can be described, at least
in principle, in non-normative vocabulary, the upshot of Brandom’s pragmatism about norms is an
account of normative phenomena—and especially of semantic phenomena—that does not require
reference to normative notions.
My aim in this paper is to discuss some issues about the thesis of the supervenience of
norms that Brandom defends in Making It Explicit (henceforth MIE). The basic idea is that
the pragmatist strategy of explanation of the normative aspects of intentional phenomena
does not succeed in distinguishing itself from naturalistic approaches: pragmatism about
norms explains normativity through a reduction of norms to notions that are in the end—
contrary to what Brandom asserts—non-normative.
The plan of the essay is as follows: in the first section I consider Brandom’s arguments
against the reducibility of attitudes to dispositions and argue that they are less than
conclusive. In the second section I will then contend that Brandom’s thesis of the
supervenience of norms on deontic statuses and hence on normative attitudes is untenable,
because norms are indeed reducible to attitudes and, for what I have argued in the first
section, to dispositions.
1. Attitudes and Dispositions
Brandom’s solution to the problem of rule-following is centred around the idea that we can
explain the existence of rules if we focus on our activity of treating performances as correct
or incorrect. In this way normative statuses are taken to supervene on normative attitudes,
which in turn are deemed to be non describable in purely naturalistic terms.1 Is this
assumption reasonable? To answer this question, we need to examine his account of
normative attitudes.
Normative attitudes are assessments, “assignments to performances of normative
significance or status as correct or incorrect according to some norm” (MIE, p. 35). But
assessments can be understood as dispositions to sanction, that is, to reward appropriate and
to punish inappropriate performances, as Brandom recognizes.2 Therefore, it seems that
normative attitudes are completely explainable in naturalistic terms: they seem reducible to
clusters of dispositions. When we speak of one’s attitudes we are actually speaking of her
dispositions to react to certain performances sanctioning them, positively or negatively. But
1
See MIE, ch. 1.
2
Cf. MIE, pp. 34ff.
if this hypothesis were correct, we could conclude that there is a naturalistic description of
norms that is couched entirely in non-normative terms.
This line of reasoning, however, is quickly dismissed.3 In fact, Brandom offers two
arguments against the reducibility of attitudes to dispositions.
(A) The first argument is based on the observation that the normative character of the
metalanguage in which norm-instituting social practices are specified is irreducible to
naturalistic accounts:
… it is important to realize that it is one thing to understand practical assessment as
sanctioning, and quite another to understand sanctioning in non-normative terms such as
reinforcement. … Defining normative attitudes in terms of dispositions to apply
sanctions does not by itself reduce the normative to the non-normative—it just trades
off one sort of norm for another. At the most basic level, to reward someone is to offer
some good …, and to punish them is conversely to inflict something bad. Benefit and
harm, desirable and undesirable, are concepts that also have normative senses. Indeed,
these senses would seem to be primary, so that some sort of reductive hypothesis would
be needed to naturalize them. (MIE, p. 42)
According to Brandom the reduction of attitudes to dispositions to apply sanctions cannot
be a proper reduction of normative notions to non-normative ones, because the description
of attitudes as dispositions to sanction is not entirely couched in naturalistic vocabulary. In
fact, the concept of ‘sanction’ is a normative notion, something that refers back to a
normative theory of what benefit and harm consists in.
(B) The second argument against the naturalizability of normative attitudes starts by
noticing that positive and negative sanctions need not consist in rewards and punishments.
Indeed, they “may consist in acclaim and censure that itself has only a normative
significance” (MIE, p. 43). A correct performance can be rewarded by release from an
obligation; in the same way, an incorrect execution might be punished by withholding a
license. In such cases there is no direct shift from normative evaluations to bestowals of
benefits or impositions of harms; we simply face “a change in normative status rather than
natural state” (MIE, ibidem). Following this remark it is possible to distinguish between
external and internal sanctions. External sanctions are those sanctions that are expressible
in non-normative terms, like offering something good or inflicting a physical punishment.
Instead, internal sanctions are those sanctions that involve only a change of normative
status. Now, for Brandom
[i]t is possible to interpret a community as instituting normative statuses by their
attitudes of assessment, even though each such status that is discerned is responded to
by sanctions that involve only other normative statuses. … Such an interpretation would
not support any reduction of normative status to non-normatively specifiable
dispositions, whether to perform or to assess, whether individual or communal. (MIE, p.
42)
3
Even though Brandom criticizes explicitly the identification between assessments and
dispositions to sanction, several commentators have attributed to him this view: “So put it appears as
though Brandom is offering a dispositionalist account of the determination of correctness—since the
starting point includes nothing more than behavioural dispositions. Moreover, … nothing is added
that would distinguish the account from dispositionalism” (Hattiangadi 2003, p. 425). However, this
seems to me a hasty conclusion, because to establish the thesis one would have first to address the
arguments Brandom adduces against the naturalizability of normative attitudes. See also Grönert
2005, p. 163-4.
Brandom takes the possibility of a community that dispenses with external sanctions to
show the non reducibility of normative statuses to non-normative notions. Therefore,
Brandom’s defence of the irreducibility of attitudes to dispositions seems to entail the
following modal claim:
(1) Necessarily, normative attitudes, at least when internal sanctions are at work, are
not epistemically reducible to clusters of behavioural dispositions.4
From this follows that, if there are reasons to reject (1), Brandom’s defence is unsuccessful.
The simplest way to refute (1) consists in providing some counterexample. Since (1) is a
necessary statement it suffices to exhibit a putative case in which (1) fails. To secure this
outcome we must understand how is it possible to reduce a normative attitude concerning
an internal sanction to a disposition articulated in naturalistic terms. The major obstacle
seems to be that of expressing a change of normative status without reference to normative
notions. In fact, for what concerns the argument (A), it is possible to notice that, since
individuals who share the same conception of benefit and harm may disagree in their
attitudes towards a given performance, and conversely, subjects having different
conceptions of good and evil may express the same attitudes, the choice of the sanctions is
not relevant to define the status of a performance as correct or incorrect.5 If this were not
precluded by the variability of sanctions, we could express normative attitudes—as
conceptually separate from the specification of the behavioural outputs in which they
result—as dispositions, in purely non-normative terms. For example: we can consistently
imagine a community in which people are disposed to adopt a single type of positive
sanction and a single type of negative sanction. In such a case a dispositionalist account of
normative attitudes entirely couched in naturalistic terms would be available.
But the difficulty posed by the argument (B) is not so easy to cope with. To overcome
the problem linked to the distinction between external and internal sanctions we could first
note that once an internal sanction is applied it is reasonable to expect that the community
members begin to adjust their behaviour in accordance with the change of normative status
following the sanction: for example, by declaring what additional duties the transgressor is
now required to accomplish. Basically, a behavioural disposition is a way of associating a
certain behavioural output with a given circumstance without appealing to inner states.
Therefore, it is still possible to conceive of normative attitudes pertaining to internal
sanctions in dispositional terms, not as dispositions to sanction, but as sets of dispositions to
calibrate one’s behaviour as a consequence of changes of normative status.
Consider again the previous example. Instead of a community in which all individuals
are disposed to sanction in the same way, we can imagine a community in which every
4
Where a is epistemically reducible to b =df everything that can be said about a and its properties,
can be said by speaking only of b and its properties. This formulation aims to capture Brandom’s idea
of reduction as the activity of eliminating, “in favour of non-normative or naturalistic vocabulary, the
normative vocabulary employed in specifying the practices that are the use of language” (MIE, p.
xiii). There is no attempt to reduce ontologically attitudes to dispositions: this issue is beyond the
scope of this essay.
5
It is useful to compare this case with that of communal assessment theories about rule-
following. Brandom maintains that communal assessment theorists make illicit reference to a
normative concept when they define correctness in terms of another normative concept, the concept
of ‘expert’ (MIE, p. 39). This objection seems reasonable, because the reference to the normative
notion of ‘expert’ is necessary in order to determine the extension of the set of correct
performances—since this set is by hypothesis composed of just those performances that are so
evaluated by experts. So in this case—but not in the case of the reduction of attitudes to dispositions
to sanction—the reference to a normative concept is necessary in order to determine the content of
attitudes towards performances.
member must perform some task from a given set A. When someone fails to carry out her
duty, her normative status changes and she is committed to performing some task from
another set, say B. If she fails again, she is assigned a task from a third set, C. Eventually, if
inaccuracy persists, the community member is charged with an external sanction S. If we
should describe the normative attitudes of these folks in evaluating other community
members’ performances, we could avail ourselves of the dispositionalist jargon as follows:
everyone has a set of first-order and second-order dispositions. First-order dispositions are
instructions that specify how to accomplish certain tasks and how to behave towards other
community members; second-order dispositions are instructions that specify how to react to
others’ practitioners performances. If we assume that each individual has some default
dispositions and define a second-order disposition as a disposition to alter one’s own pre-
existing dispositions, we can interpret normative attitudes as second-order dispositions, that
is dispositions to keep track of others’ performances and modify one’s own default
dispositions in accord with new evidence concerning how a certain performance has been
carried out. In this way the original behaviourist motivation behind the dispositionalist
interpretation of attitudes can be preserved without reintroducing normative notions. In
fact, once it is defined a convenient set of first-order dispositions and of rules that explain
in which way an agent alters her first-order dispositions as a consequence of getting new
information, it becomes possible to display a purely naturalistic account of the activity of
assessing performances in which changes of normative status are expressed in terms of
second-order dispositions. Moreover, in this account the difference between internal and
external sanctions is made harmless, because it responds to a difference in the sets of
second-order dispositions.
It is not difficult to adapt this dispositional model to the case of linguistic practice. In
order to accomplish this task it is necessary to take into account inferential as well
representational dimensions of linguistic activity. 6 This can be done, following the trace of
Brandom’s scorekeeping semantic, as follows:
(2) For a given atomic sentence φ, the meaning of φ, Mφ, consists in an ordered
quintuple, Mφ: , where Cφ is the set of commitment-preserving
inferences in which φ serves as conclusion, Pφ is the set of commitment-
preserving inferences in which (along with other auxiliary hypothesis) φ plays an
essential role as premise, Eφ is the set of entitlement-preserving inferences in
which (along with other auxiliary hypothesis) φ plays an essential role, Iφ is the set
of the sentences that are incompatible with φ and Aφ is the set of the states of
affairs of which φ constitutes an appropriate observation report.7
Each speaker in her use of words follows what she believes to be the conditions of correct
use. So, she must have some ideas about what assertions commit to a given sentence φ, to
what further claims she is committed as a consequence of her assertion of φ, to what
assertions she is entitled as a consequence of entitlement to φ, what sentences are
incompatible with the assertion of φ, what states of affairs can be described by means of φ.
However, two or more speakers can differ in their judgments about what inferences are
licensed by a given utterance or about what states of affairs can be correctly described by
means of a given sentence. This complication is acknowledged by Brandom himself when
he recognizes the perspectival character of conceptual content, that is, the fact that what is
assumed to constitute the content of a sentence varies from speaker to speaker, according to
6
Actually, we should consider also ‘practical commitments’, for example commitments to act
following from the utterance of a given sentence (see MIE, ch. 4). However, I will ignore this
problem in what follows.
7
See MIE, pp. 188-189.
the linguistic customs each speaker has acquired.8 Noticing that content is perspectival
imposes a slight reformulation of our previous definition of meaning in terms of sets of
sentences and states of affairs. Since each speaker can attribute a different meaning to a
given assertion, it appears reasonable to index meaning in relation to speakers. As result,
we have that the meaning of a given sentence φ is expressible as an indexed family of sets,
Miφ: , in which each Mjφ represents the meaning of φ in the perspective of a
single speaker j.
Once we have defined what performances are comprised in the activity of classifying
correct and incorrect uses of sentences, it is possible to state more accurately the content of
a pragmatist approach to semantics. In the general case the presence of rules was explained
through the recourse to attitudes instituting normative statuses of performances. If we want
to adjust this style of reasoning to semantic norms we have to identify the corresponding
normative statuses and normative attitudes. The preceding discussion has suggested that
normative statuses pertaining to semantic norms are connected with the activity of
individuating, for a given sentence φ, and for a given speaker j, what inferences form the
sets Ciφ, Piφ, Eiφ, what sentences form the set Iiφ, and what states of affairs are collected in the
set Aiφ. The upshot is that we can easily transpose the abstract analysis of rules to the case of
semantic norms if we take into account the five components of linguistic meaning, , and the corresponding deontic attitudes of treating someone as committed or
entitled to or interdicted from further assertions as a consequence of the utterance of a given
sentence.
Consider, for example, a speaker j and a set of n sentences S. Since meaning is
perspectival, for each sentence φ∈S there is a quintuple Mjφthat supplies with the conditions
of correct use for that sentence in that speaker’s perspective. Mjφ indicates what inferences
follow from φ and what states of affairs can be reliably reported by means of it in j’s
perspective and so explains how j is prepared to use φ in her linguistic exchanges with other
speakers. We can call MjS = {Mjφ| φ∈S} the set consisting of all the quintuples Mjφ for all the
sentences included in S. Now, we can extend this rough model in a simple way. We can
imagine that j, when enters into contact with another speaker k, is prepared to behave
linguistically according to the conditions stated by MjS and also to evaluate k’s utterances
along the same lines. This last condition implies that j will keep track of k’s entitlements
and commitments and of hers and will form corresponding deontic attitudes that specify to
what further performances k is committed or entitled, thus regulating her further linguistic
behaviour on these bases. Moreover, it requires that j will note the divergences between k’s
further utterances and her expectations and revise his future expectations9 and, eventually,
if there is any reason to recognize to k some special linguistic authority, modify her own
attitudes to linguistic behaviour and evaluation.
What does Brandom’s treatment of these issues in Between Saying & Doing (henceforth
BSD) add to our discussion? In his later book Brandom is primarily concerned with the
relationship between languages and practices in a more abstract setting than that deployed
in MIE. Nothing is said about attitudes and dispositions, even if Brandom repeatedly quotes
Huw Price’s thesis that normative vocabulary is irreducible to naturalistic one but that one
can specify in naturalistic terms what a given subject must do in order to deploy some non-
naturalistic vocabularies.10 As for what regards the problem I am considering the most
8
See also MIE, p. 185.
9
This means that j will keep two separate registers, concerning respectively the objective
meaning of k’s assertions—objective in j’s perspective, obviously—and the meaning that k herself
attributes to her assertions.
10
Cf. Price 2004. I think that thi thesis is in tension with the pragmatic strategy adopted in MIE,
but I cannot pursue this issue in this essay.
important reflections are contained in the third lecture. Here Brandom criticizes the tenets
of AI functionalism. More particularly, Brandom critiques the idea that
there is a set of practices-or-abilities meeting two conditions:
1. It can be algorithmically elaborated into (the ability to engage in) an autonomous
discursive practice (ADP).
2. Every element in the set of primitive practices-or-abilities can intelligibly be
understood to be engaged in, possessed, exercised, or exhibited by something that
does not engage in any ADP. (BSD, p. 75, italic in text)
If Brandom’s worries are well founded, it can be noted, the project of semantic naturalism
is not doomed to failure, since one could equally maintain that the primitive abilities that
precede ADPs bear some other relation to ADPs (for example, one could hold that ADPs
emerges from non-semantic abilities). However, it is clear that, if Brandom is right in his
contention, the naturalistic project should be rethought. Surely, in fact, the identification
between attitudes and dispositions cannot be preserved if ADPs are not algorithmically
decomposable into primitive abilities.
Against this idea, Brandom recognizes to have no “knock-down arguments”. However
from the text it is possible to elicit the following qualms:
(A’) The practice of adjusting one’s other beliefs in response to a change of belief is
intrinsically holistic; this raises the problem of revising and updating one’s commitments
and entitlements in the right way, that is in a way that be sensitive to one’s other collateral
commitments and entitlements. Since “any change in any property of one changes some of
the relational properties of all the rest” (BSD, pp. 80-81, italics in text), “it is not plausible
… that this ability can be algorithmically decomposed into abilities exhibitable by non-
linguistic creatures” (BSD, p. 81, italics in text), because each attempt to deal with this
difficulty should face the problem of finding a rule to determine what factors are to be
ignored. This trouble is assuaged in linguistic creatures, for the latter have semantic,
cognitive, or practical access to the complex relational properties they would have to
distinguish to assess the goodness of many material inferences.11
(B’) Brandom contrasts algorithmic decomposition into primitive abilities with training
by an expert. A course of training can be thought of “as having as its basic unit a stimulus
(perhaps provided by the trainer), a response on the part of the trainee, a response by the
trainer to that response, and a response to that response by the trainee that involves altering
his dispositions to respond to future stimuli” (BSD, p. 87, italic in text). Moreover, the
abilities interested by this process “vary wildly from case to case, and depend heavily on
parochial biological, sociological, historical, psychological and biographical contingencies”
(BSD, p. 85). Finally, the question of what algorithmic elaboration is sufficient for a
particular creature, in a particular context cannot be settled empirically.
To these worries it is possible to reply as follows. For what concerns the issues grouped
under the heading (B’). The way in which Brandom treats the phenomenon of training
suggests that there is nothing magical or mysterious in the manner trainees are instructed by
their trainers. After all, Brandom himself proposes an abstract model of what a course of
training should consist in:12 a series of responses from the part of the trainee to which the
trainer reacts with appropriate corrections. It seems to me that in this succession of events
11
Cf. BSD, p. 83.
12
Brandom’s description of training, it can be noted in passing, reminds what Donald Davidson
called, in his later writings, triangulation. See, for example, Davidson 2001.
there is nothing that cannot be algorithmically decomposed: it suffices to set a sequence of
stimuli arranged in a proper way, so that each stimulus be related to the preceding
responses of the trainee. (Obviously the practical implementation of this model can pose
almost insurmountable difficulties; but this problem is connected with the troubles that
come out from the argument A’.) Moreover, it is true that the abilities involved vary
dramatically from case to case and that the success of a particular course of training cannot
be predicted in advance. But these empirical limitations do not affect the algorithmic
decomposability in principle of the process of training. They are rather to be viewed as
contingent features of the training process: since a single course of training can be
implemented by activating different sets of abilities, the variability of the abilities involved
in each particular case is connected with the multiple realizability of the overall process, in
accord with the functionalist thesis.
If this is correct, the point that emerges from the objection (B’) is strictly related to the
worries posed by the argument (A’). It is the practical intractability of the holistic character
of our activity of revising our beliefs that motivate Brandom’s discomfort with the
possibility of an algorithmic decomposition of the ability to engage in an ADP. But,
whereas it is possible to agree with him that a proper treatment of semantic holism poses a
formidable obstacle to our efforts of creating a computational system capable of engaging
in linguistic practice, no evidence is offered for the stronger claim that such a treatment is
in principle impossible. From the fact that current researchers are not able to simulate ADP,
does not follow that this task cannot be accomplished.13
Moreover, the idea that only those creatures which have something like a semantic
access to the complex relational properties they would have to distinguish to assess the
goodness of many material inferences are able to engage in ADPs, is in tension with the
semantic project pursued by Brandom in MIE. In fact, from that work can be elicited a
conception of meaning which exploits, as its basic ingredients, the performances of the
speakers that are caught in the game of adjusting their beliefs in response to the assertions
of other speakers. For these reasons it seems to me that the scepticism that Brandom now
shows towards algorithmic decomposability of the performance of revising one’s beliefs
contrasts with the role that in MIE is assigned to deontic scorekeeping: if an access to
semantic properties is necessary in order to exhibit those abilities that are required to
engage in an ADP the semantic project pursued in MIE becomes irredeemably circular,
because one cannot see what utility possesses a semantic explanation that employ as its
basic ingredients notions that are just semantic in advance.
My conclusion is that Brandom’s critiques to AI functionalism are less than decisive.
But if there is no conclusive argument against the algorithmic decomposability of the
ability to engage in an ADP, and the analogue arguments offered in MIE against the
identification of normative attitudes with dispositions equally fail, there is a naturalistic
reading of Brandom’s pragmaticist semantic that starts with the interpretation of the basic
attitudes as dispositions and ends with a reduction of norms to naturalistic items, as I am
going to show.
2. Supervenience of Norms
I have argued that Brandom’s arguments against the naturalizability of normative attitudes
are far from being conclusive. But if we cannot dismiss the idea that normative attitudes
are, at least in principle, reducible to behavioural dispositions—that can be accounted for in
13
The issue is the object of a large body of work, from connectionist approaches to fuzzy logic,
that far exceeds the scope of this paper. At present, I would like to mention only the recent and
promising attempt by Andy Clark of challenging Fodor’s frame problem. See Clark 2002.
non-normative vocabulary—we must also admit, via the thesis of supervenience of
normative statuses on normative attitudes, that normative statuses can be taken to
supervene on dispositions to regulate one’s own behaviour as a consequence of normative
assessments. And this conclusion poses some problems, because it seems to entail a new
variety of naturalism about norms. But to verify whether this suspicion is a sensible one we
should first try to understand what the supervenience thesis exactly implies.
To see how this happens, it is useful to reformulate the thesis of the supervenience of
normative statuses in a slightly more technical fashion. Saying that normative statuses
supervene on normative attitudes means, in Brandom’s words, that “settling all the facts
concerning normative attitudes settles all the facts concerning normative statuses” (MIE, p.
47, italics in original).14 This description suggests a global supervenience of statuses on
attitudes. 15 However, since normative statuses are instituted by attitudes, it seems
reasonable to interpret the kind of dependence Brandom has in mind as asserting also that if
two individuals entertain the same normative attitudes they institute the same normative
statuses.16 This kind of dependence can be expressed as a weak supervenience of statuses
on attitudes.17 Indeed, global supervenience formally does not entail weak supervenience.
However, it has been shown that it does if we consider only intrinsic properties, and the
property of entertaining a given attitude towards a performance so and so is certainly
intrinsic. In fact, global supervenience entails strong supervenience either—where strong
supervenience is the thesis that if two individuals, whether in the same or different possible
worlds, entertain the same normative attitudes they institute the same normative statuses—
if we limit our attention to intrinsic properties, so in what follows I will assume that
Brandom is committed to the thesis of the strong supervenience of normative statuses on
normative attitudes.18 Moreover, since attitudes are—at least in principle—reducible to
behavioural dispositions, this thesis entails that normative statuses strongly supervene on
dispositions.
This way of stating the matter, however, does not seem completely correct. In fact, in
the last chapter of MIE, Brandom advances a different explanation of the relationship
between normative statuses and attitudes. Brandom stresses the fact that normative statuses
are not instituted by actual attitudes but only by correct attitudes. The institution of statuses
should be understood
in terms of the implicit practical proprieties governing such scorekeeping—not how the
score is actually kept but how, according to the implicitly normative scorekeeping
practices it ought to be kept, how scorekeepers are obliged or committed to adopt and
14
It should be noticed, however, that Brandom’s thesis that the conceptual proprieties implicit in
discursive practices incorporate empirical dimensions (see MIE, pp. 119-120 and 331-332) could
revoke into doubt this global supervenience thesis, since earthlings and twin-earthlings count as
instituting different conceptual contents even though entertaining the same attitudes.
15
See Rosen 1997, pp. 164-165.
16
While it is commonly agreed that Brandom’s theory of supervenience of rules can be expressed
as a weak supervenience thesis, it is a more disputed issue whether it entails a strong supervenience
thesis either—that is a relation of covariance that holds necessarily, for all possible worlds—too. See
Loeffler 2005, p. 58.
17
More precisely, saying that A-properties weakly supervene on B-properties implies that
necessarily (that is, in every possible world), if two objects possess the same B-properties they share
also the same A-properties. For further details, see Kim 2003, p. 559.
18
A-properties strongly supervene on B-properties iff if two objects, whether in the same or
different possible worlds, possess the same B-properties they share also the same A-properties. See
again Kim 2003, p. 560.
alter their deontic attitudes, rather than how they actually do. (MIE, p. 628, italics in
original)
What distinguishes the attitudes that are capable of instituting statuses is their correctness.
The institution of statuses is a consequence of keeping the deontic score in the right way.19
So, not every attitude institutes a corresponding status but only those attitudes that are
adopted according to a proper scorekeeping activity. In this sense normative statuses are a
product of an idealization of the actual scorekeeping practices.20
Therefore, we have to take into account not actual normative attitudes but only attitudes
that are correct in the sense explained above, that is consistent normative attitudes and
deontic attitudes that are correctly taken as a consequence of the encounter with other
community members’ performances. To translate this definition in the dispositionalist
jargon, we should be able to express the idea that normative statuses strongly supervene on
sets of dispositions correctly displayed in patterns of social interaction. Recalling the
previous discussion about the dispositionalist interpretation of attitudes, it can be suggested
that we may attempt to formulate this point in terms of second-order dispositions. For
example, if we interpret the family of sets MjS as determining j’s default first-order
dispositions to linguistic behaviour—that is, the way in which, at the beginning of the
conversation, j is prepared to behave and to evaluate another speaker’s assertions—we can
conceive of deontic attitudes as second-order dispositions that specify in which way a given
speaker j, who keeps track of others’ commitments and entitlements, will modify her
previous dispositions in accordance with new evidence concerning how a certain linguistic
performance has been carried out—for instance, adjusting the deontic status of her
interlocutor (intended as a cluster of first-order dispositions). 21 Therefore, the
supervenience thesis can be expressed as the thesis according to which normative statuses
strongly supervene on correct sets of second-order dispositions—where ‘correct’ can be
obviously defined in non-normative terms through reference to the way in which second-
order dispositions keep track of previous performances—and indirectly on sets of first-
order dispositions. (In fact, a second-order disposition must refer to the criteria set up by
first-order dispositions to individuate the performances that require to be sanctioned and the
first-order dispositions that are to be modified.)
19
In the same vein, Ronald Loeffler has recently maintained that the right way to intend the
supervenience thesis is as asserting that “not de facto normative attitudes, but only attitudes that
should be adopted or that are properly adopted, determine semantic norms” (Loeffler 2005, p. 62,
italics in original).
20
Now, however, one could ask what are the norms according to which the judgments concerning
the correctness of attitudes are made. But Brandom’s answer to this last question could hardly be
considered satisfying. The only partial response, in fact, comes at the very end of the book and seems
to consist in the rather disappointing admission that identifying the parameters of correctness is
entirely up to the interpreter who attempts to reconstruct the discursive scorekeeping practices.
According to Brandom the norms that determine when it is correct for an agent to attribute a certain
doxastic commitment to someone else are not available in advance as a set of explicit principles “but
are implicit in the particular practices by which we understand one another in ordinary conversation”
(MIE, p. 646). Moreover, since the external interpretation of a linguistic community is not
qualitatively different from ordinary scorekeeping activity, “[t]here is never a final answer as to what
is correct; everything, including our assessments of such correctness is itself a subject for
conversation and further assessment, challenge, defense, and correction” (MIE, p. 647). See also
Laurier 2005, pp. 156-158.
21
It is reasonable to imagine that there should be a set of dispositions that specify in which cases
the speaker has to recognize her interlocutor some kind of linguistic authority, but this is a point I will
not pursue further.
Now, we have to understand whether this way of linking normative statuses to
behavioural dispositions matches Brandom’s anti-reductionist premises. I have recalled that
Brandom is committed to the supervenience of normative statuses, and hence of norms, on
normative attitudes.22 But then, if we accept the reducibility of normative attitudes to
behavioural dispositions and formulate the thesis of supervenience of norms as a strong
supervenience thesis, we obtain the following:
(3) Supervenience of norms on behavioural dispositions: If two individuals possess
the same (correct second-order and consistent first-order) behavioural
dispositions, whether in the same or different possible worlds, they can be said to
institute the same norms.23
To make this idea more precise we can define a set of dispositions as a function D:
S→B from states of affairs to behavioural responses. In other words, a set of dispositions
can be conceived as a set of pairs D ⊂ S×B where S is a set of states of affairs and B is a set
of patterns of behaviour. Given this definition we can say that an individual x possesses a
set of dispositions D only if for each (s, b)∈D, if x is in the state of affairs s she will act in
accordance with the pattern of behaviour b.24 Accordingly, the thesis of the supervenience
of norms becomes as follows:
(4) Supervenience of norms on behavioural dispositions: Necessarily, if x, in a certain
state of affairs s, institutes a norm N, there exists a set of dispositions D: S→B
such that s∈S, and x possesses D, and everyone that is in s and possesses D
institutes N.25
This definition states that two subjects can be said to institute the same norm if they have
an identical set of dispositions—and these dispositions are correct, in the sense explained
above. Each individual that presents a correct set of dispositions of this kind can be said to
institute the related norm. It is also important, however, to make clear what supervenience
does not imply. Supervenience entails that individuals having the same correct dispositions
institute the same norms, but the converse does not hold. On the contrary, it is a central
feature of the concept of supervenience that if a set of properties A supervenes on a set of
properties B, a property a∈A can supervene on several different subsets of B. This means
that, if norms supervene on behavioural dispositions, it is possible that two individuals
institute the same norm even if they possess different dispositions. But how does it happen?
In response to this question, one could maintain that the variability in the dispositional
basis is connected with the fact that different individuals may be disposed to react to other
practitioners’ performances in different manners. So, two individuals can institute the same
norm even if, for example, one of them is disposed to apply external sanctions whereas the
other applies only internal sanctions. (It is obviously possible to imagine far more ingenious
variations in the dispositional basis than these differences in the ways of sanctioning. This,
however, would not change the line of reasoning I am considering.)
22
In fact, if settling all the facts concerning normative attitudes settles all the facts concerning
normative statuses, two worlds that are alike for what concerns normative attitudes cannot differ in
their norms.
23
This definition is adapted from Loeffler 2005, p. 58.
24
That possess of a disposition entails a subjunctive conditional—at least in ideal conditions—is
presumably uncontroversial. I will not enter—nor this is relevant for my argument—into the much
debated issue of reducibility of dispositions to conditionals—for further details, see Mumford 2003;
Fara 2006.
25
For this way of formulating strong supervenience see Kim 2003, p. 561.
But if this is so what follows for the prospect of a dispositionalist naturalism about
norms? Well, it remains possible to take the union of the sets of dispositions that institute a
given norm and assume that that norm is coextensive—indeed, reducible—to such a set.
Thus, we obtain the following:
(5) Reduction of norms to behavioural dispositions: If x1, in a certain state of affairs s,
institutes a norm N and x1 possesses a set of dispositions D1: S→B such that s∈S,
and x2, in a certain state of affairs s, institutes a norm N, and x2 possesses a set of
dispositions D2: S→B such that s∈S, and … then N is coextensive, hence reducible
to Di.26
This shows that Brandom’s account points towards a reduction of norms to dispositions.27
But then we face a complete reduction of normative phenomena to naturalistic facts. In
other words, Brandom’s account of norms—and Brandom’s semantics—prove to constitute
a new kind of naturalism about normativity and meaning: accepting Brandom’s elucidation
of norms means accepting the idea that there may be a story entirely couched in naturalistic
terms that explains how individuals, starting from a small set of dispositions to social
behaviour and to acquire new dispositions, can institute a whole world of norms.
3. Conclusions
In this paper I raised some doubts about Brandom’s pragmatist strategy of explanation of
norms. I argued that if we attempt to explain normative statuses through recourse to
normative attitudes it is impossible to preserve a hiatus between norms and regularities of
behaviour. Since understanding of norms is reducible to possessing the right behavioural
first- and second-order dispositions, and since calibrating one’s own behaviour as a
consequence of normative assessments can be described, at least in principle, in non-
normative vocabulary, the upshot of Brandom’s pragmatism about norms is an account of
normative phenomena—and especially of semantic phenomena—that does not require
reference to normative notions.
This result is clearly at odds with some of Brandom’s basic assumptions. As a
consequence, we are left with two possibilities. Either we abandon the primitiveness of
normative notions and accept the reducibility of norms to dispositions and consequently to
naturalistic facts, or, if we want to safeguard the issue of non-reducibility, we must revise
the pragmatist strategy followed by Brandom, especially for what concerns the aim of
offering an explanation of normative phenomena in terms of social activity.28 The first horn
of the dilemma indicates a new route towards naturalism about norms; on the contrary, the
second proposal seems to point in the direction of a partial separation of intentional
vocabulary from the sphere of naturalistic reports. In both cases, however, it should be clear
26
Cf. Kim 1990.
27
A different problem—which I cannot currently pursue—is whether a given disposition may
supervene on different categorical properties—for more on this issue see again Mumford 2003. Note,
however, that for our present concern, once we grant that normative attitudes are indeed reducible to
dispositions, the underlying ontology of dispositions themselves becomes largely irrelevant.
28
This revision could follow several routes: for example, one could argue more directly against
the behaviouristic interpretation of attitudes in terms of dispositions. Otherwise, one could reject the
transition from global supervenience to strong supervenience, by admitting that facts about norms are
instituted by normative attitudes, and hence by behavioural dispositions, merely in the sense that
settling all the facts concerning normative attitudes settles all the facts concerning normative statuses,
without this condition entailing that if two individuals share the same dispositions they institute the
same statuses.
that the original inspiration of Brandom’s approach—combining an account of the social
institution of norms with the primitiveness of normative notions—cannot be preserved.
References
Brandom, R.B. (1994) Making It Explicit: Reasoning, Representing, and Discursive
Commitment, Cambridge, MA, Harvard University Press.
Brandom, R.B. (2008) Between Saying & Doing: Toward an Analytic Pragmatism, Oxford,
Oxford University Press.
Clark, A. (2002) “Local Associations and Global Reason: Fodor’s Frame Problem and
Second-Order Search”, Cognitive Science Quarterly, 2, pp. 115-40.
Davidson, D. (2001) Subjective, Intersubjective, Objective, Oxford, Oxford University
Press.
Fara, M. (2006) “Dispositions”, in E.N. Zalta (ed.), The Stanford Encyclopedia of
Philosophy, Fall 2006 edn, URL =
.
Grönert, P. (2005) “Brandom’s Solution to the Objectivity Problem”, Pragmatics and
Cognition, 13, pp. 161-175.
Hattiangadi, A. (2003) “Making It Implicit: Brandom on Rule-Following”, Philosophy and
Phenomenological Research, 66, pp. 419-431.
Kim., J. (1990) “Supervenience as a Philosophical Concept” Metaphilosophy, 21, pp. 1-27.
Kim., J. (2003) “Supervenience, Emergence, Realization, Reduction” in M.J. Loux, D.W.
Zimmerman (eds.), The Oxford Handbook of Metaphysics, Oxford, Oxford
University Press.
Laurier, D. (2005) “Pragmatics, Pittsburgh Style”, Pragmatics and Cognition, 13, pp. 141-
160.
Loeffler, R. (2005) “Normative Phenomenalism: On Robert Brandom's Practice-Based
Explanation of Meaning”, European Journal of Philosophy, 13, pp. 32-69.
Mumford, S. (2003) Dispositions, pbk edn., Oxford, Oxford University Press.
Price, H. (2004) “Naturalism without Representationalism”, in M. De Caro, D. Macarthur
(eds.), Naturalism in Question, Cambridge, MA, Harvard University Press.
Rosen, G. (1997) “Who Makes the Rules Around Here?”, Philosophy and
Phenomenological Research, 57, pp. 163-171.