Conventions Revisited

Document Sample
Conventions Revisited Powered By Docstoc
					                       Conventions Revisited: A Reply to Critics
                                        Andrei Marmor

Tim Williamson once told me that the most difficult task in philosophy is to
say something that is, actually, true. We always try, of course, yet often we
probably fail. Progress is achieved, however, in small steps. Every criticism of
an argument you encounter is a step forward; even if you think that the critique
is unwarranted, you come to realize that you should have made your argument
clearer. So I am very grateful for the critiques of my book presented in this
symposium (and elsewhere), and I hope that by responding to some of them,
we can make a little bit more progress. I will begin with a few preliminary
remarks on constitutive conventions, and then I will consider some of the
critiques in greater detail.

1. Constitutive Conventions.
David Lewis‟s theory of conventions was a remarkable achievement and, in
chapter one, I tried to explain in some detail what Lewis achieved and why it is
so important. The idea that conventional norms evolve as solutions to large
scale recurrent coordination problems is certainly true in may instances, and
explains a great deal about the essential features of conventional norms. But if
the only tool you have is a hammer, then everything starts to look to you like a
nail. The coordination rationale of conventions has become this hammer over
the years. Every conventional practice can be made to look like a solution to
some coordination problem or other, if you only stretch the idea of social
coordination enough; and then quite a stretch it becomes. In a nutshell, this has
always been my motivation for looking for some alternatives, without
eschewing Lewis‟s valuable insight that conventional norms are, essentially,
arbitrary in the sense he defined. The element of arbitrariness is, indeed, where
we should start. But my definition of arbitrariness differs from Lewis‟s in one
respect, which may look minor, but is significant: Lewis‟s definition is given in
terms of subjective preferences of the relevant agents.1 Instead, the definition I
offered is given in terms of reasons. Conventionality, I argued, is always a
matter of relativity to reasons: a norm is conventional iff its content is
underdetermined by the reasons for having the norm in the first place. If the

    D Lewis, Convention, A Philosophical Study (Blackwell, Oxford 1969), 58.
content of a norm is fully determined by its underlying reasons, then the norm
is not arbitrary and thus, not conventional.
Regarding conventionality in terms of relativity to reasons should help us to see
how unlikely it is that all conventions are responses to problems of
coordination. Social norms evolve in response to a great variety of human
needs and social interactions; those social circumstances and the reasons they
constitute do not always determine a uniquely correct normative solution. In
countless cases, conventions evolve that respond to those reasons. Now, we
can say that conventions evolve as solutions to some identifiable needs or
problems, if we like, but it would be quite incredible to assume all the social
problems conventions respond to are of a coordinative nature. Surely, our
social lives are more complex than that.
For one thing, the coordination rationale underestimates the resourcefulness
and inventiveness of human nature. Our social lives are not limited to solving
problems we encounter. We invent activities and practices that we find valuable
to engage in. People have developed over the centuries numerous structured
activities that enable new forms of social interactions to evolve and new values
to emerge. A case in point is the nature of social activities we engage in by
following a set of interlocking rules, where the rules constitute the activity in
question, such as playing structured games or sports, creating works of art
(within specific genres), forming various institutional practices, etc. Many of
these activities are very complex and quite elaborate, enabling us to realize
certain values that only make sense as instances of engaging in the kind of
activities those rules constitute. To be sure, I have never claimed that all
constitutive rules of this nature are, necessarily, conventional2. But when the
constitutive rules of such activities are arbitrary, in the defined sense, and hence
conventional, then we have a set of constitutive conventions.
The distinction I suggested between coordinative and constitutive conventions
turned out to be more controversial than I had expected. Federico Jose Arena
doubts my distinction on the basis of a quotation from Searle‟s book, alleging
that “constitutive rules are not conventions”.3 Now of course, if we assume
that conventions are solutions to coordination problems, as Searle probably did

  On the contrary: in chapter 4 I argued that constitutive rules determining the literal
meaning of words in a natural language are not, mostly, conventional. Unless otherwise
indicated, all references to my book relate to Social Conventions: from language to law (Princeton,
  Arena, at […], from J Searle, The Construction of Social Reality (Free Press, NY, 1995) 49. Del
Mar makes a similar point in “Marmor‟s Social Conventions: The Limits of Practical
Reason”, Philosophy of the Social Sciences,

at the time, then he is right to doubt that constitutive rules are of this nature;
constitutive rules are not, typically, solutions to coordination problems. Searle
also seems to doubt that constitutive rules can be arbitrary; but his argument
for this claim rests on a very questionable assumption. The assumption is that
any change in a rule constituting a certain act-type would result in a different
act-type; hence the relation between a rule and the act it constitutes is, Searle
claims, necessary, not contingent. I have responded to this line of thought in
the book. First, I argued that we would be wrong to assume that the
constitutive relation here is between a rule and an act-type. Rules constitute
certain activities, not particular acts.4 Second, it is a mistake to think that every
alternative to a constitutive convention, no matter how minor or insignificant it
is, would necessarily result in a different type of practice or activity. To stick
with my warn-out example: it is hardly the case that if the rules of chess would
be slightly different from what they are, it would not be chess anymore. Not
every change in the rules would result in a different game. Playing chess with a
rule that allows castling and playing it without this rule is not all that different.
People can play chess somewhat differently, and it would still be playing chess.
Some of the rules or conventions that constitute an activity are, typically, more
essential than others to what the activity is, and to the kind of values inherent
in it. (If you allow chess players to use physical intimidation in the game, for
example, it might become a very different game indeed.) But this certainly does
not show that there is a necessary connection between constitutive rules and
the practice that is constituted by them. It only shows that some of the rules
constituting a practice are more central or essential than others to what the
practice is, and the kind of things we value about it.5
Having said this, I do not deny that we are bound to face some difficulties with
the application of the arbitrariness condition. It is sometimes very difficult to
determine whether different normative practices are just different responses to
the same reasons, or instantiate somewhat different reasons. In some cases it
may not be evident that different norms amount to alternative ways of doing
roughly the same thing. But the crucial point to realize is that rough identity is
all we need here. Reasons are sometimes a bit fuzzy and imprecise. My reasons
for reading a novel this evening, for example, might not be precisely identical
to my reasons for having started reading it a week ago. But they are, roughly,
the same reasons. Rough identity is bound to be vague, and borderline cases
cannot be ruled out.

    Pp. 31-35.
    Pp. 42-44.

Michael Guidice suggests that my distinction between two types of conventions
rests on a confusion between reasons for following particular norms and the
rationale of the practice in which we find the norms . His argument is worth
quoting in some detail:
          “If the comparison is made at the right level, the difference might
          disappear: one's rationale for playing the game of chess would not
          include a need to coordinate behavior, but likewise one's rationale for
          accepting a system of traffic laws, or any system of laws regulating
          important conduct, would be misunderstood if characterized simply as
          the need to coordinate behaviour with fellow citizens (since one's
          rationale, above and beyond reasons to follow particular rules or laws, would be
          something such as the common good or avoidance of harm). At the
          level of particular rules, reasons for following the particular rules of
          chess -- for example, how one must move one's rook -- look quite
          similar to the reasons for following particular rules of traffic, for
          example, driving on the right. Marmor's stark contrast between chess
          and traffic rules, which he employs to support the distinction between
          constitutive conventions and coordination conventions, only seems
          successful by sliding between two very different levels of comparison.”6
Let us take a closer look at these examples. The rationale of traffic regulations
can certainly be described at various levels of generality; we can say that they
are there to avoid harm, or to avoid accidents, or to avoid collisions, etc. But
now consider the rationale, or the reason, for having a particular norm: say,
driving on the right side of the road. There is a reason for this rule, which is to
avoid collisions or accidents. This reason determines that we should all drive on
one side of the road, but it does not determine on which side. Hence a
convention emerges (let us assume) to drive on the right. It solves the recurrent
coordination problem. Now Guidice would have us believe that if we, similarly,
focus on particular rule of chess – as opposed to the rationale of the game as
whole -- say how to move the rook, it “look[s] quite similar to the reasons for
following particular rules of traffic, for example, driving on the right.” But this
suggestion I find puzzling. Any particular rule of chess only makes sense as part
of a set of interlocking rules that constitute the game. The reasons for
following a given rule of chess cannot be given in isolation from the fact that it
is part of the game. Whereas the reason to follow the driving on the right rule
can be fully stated in terms of the coordination problem it solves, regardless of
how we associate the rule with the general practice it forms part of. (If it does;
coordination conventions do not have to form part of any general practice or
    Notre Dame Philosophical Reviews (June 2010).

type of activity; we find coordination conventions that are quite isolated, so to
speak, emerging as a solution to some particular recurrent coordination
problem.) It is not that first we have a game of chess, and then we face some
coordination problems that are solved by the rules of the game. Without the
rules there is no game of chess, and hence, no coordination problem to solve.
So I don‟t see how the reason to follow the rule of moving the rook in a certain
way is similar to the reason of solving a recurrent coordination problem about
the side of the road to drive on. The distinction here is not between different
levels of generality in describing the relevant practice; it is a distinction between
two very different functions conventions serve, and the kind of reasons
supporting them.7

2. The Rules of Recognition
Dale Smith points out that reasons for having rules or conventions can be
thought of in two ways: we can think of reasons in terms “sound normative
reasons” or in terms of “motivating reasons”.8 And Smith assumes that my
definition of arbitrariness of conventions as relative to reasons takes reasons to
be of the “sound normative” kind. This, allegedly, gets me in trouble when
applied to the rules of recognition in the legal context: it commits me to the
view, Smith claims, that “every rule of recognition is supported by sound
normative reasons”. Now of course I do not want to claim that rules of
recognition, or any convention for that matter, are necessarily good or
normatively sound. Some conventions are silly and others might be plain wrong.
One might think that it is easy to meet the challenge here: I could have simply
clarified that “reasons” in my definition of conventionality refers to motivating
reasons, not normative ones. But Smith is right to assume that this is not what
I claim. So let me try to clarify.
There is, certainly, a distinction between what is a reason for an action and
what a person believes or takes to be a reason for action. Generally, we act on
our beliefs about reasons, hoping that they are true. My feeling of thrist is a
reason to drink a glass of water; so when I pick up a glass of water intending to

  Arena […] complains that my definition of arbitrariness is ambiguous between a functional
explanation of norms and one he calls “practice based”. I am not sure that I understand
what he means by the latter but, in any case, I don‟t see an ambiguity here. Reasons for a
norm can be described in teleological terms, by pointing out the functions or aims of the
norm, and they can be described in various other ways. Nothing in my theory commits me
to any particular way of accounting for reasons.
  “The Role of Conventions in Law” […]

drink it I act on my belief that water will quell my thirst, and therefore that I
have a reason to drink it when I feel thirsty. But the belief is not the reason.
The reason is constituted by the facts, namely, the fact that thirst is a
physiological indication that our body needs water. Therefore, common
nomenclature notwithstanding, motivating reasons understood as beliefs about
reasons are not really reasons. Reasons are the facts that actually count in favor
of doing or refraining from doing something, etc.
Now suppose that you are very thirsty and there is a glass of water that can
quell your thirst, but the water is very polluted. If you drink it you might get
seriously ill. Do you have a reason to drink the water? The plausible answer, I
think, is that yes, you do have a reason to drink it, which is immediately and
unquestionably defeated by the reason not to drink it due to the pollution.9 So
now, when I suggest that there are always reasons for having conventions (or
other social norms, for that matter) I do mean reasons in the ordinary sense,
namely, as facts that count in favor of having the rules. But that leaves open the
possibility that the water is polluted, so to speak; reasons for having a
convention might be defeated, sometimes immediately and unquestionably so,
by reasons not to have it. Unfortunately, the counter-reasons in play are not
always recognized by those whose convention it is. That is why bad
conventions often exist and sometimes persist for a long time. And of course,
the same may be true of various rules of recognition.
Smith‟s main concern with my suggestion that the rules of recognition are
conventional, however, relates to a very different point. It concerns the
question of what is, exactly, that the conventional rules of recognition
determine. He presents me with an uncomfortable dilemma: either we assume
that the rules of recognition determine only what counts as the sources of law,
or else we maintain that they also determine “the content of law”. The latter
option seems implausible, given the great deal of controversy judges and other
experts have about what the content of law is in numerous cases. But then the
idea that rules of recognition determine only what counts as the sources of law
is not very helpful, Smith claims: It tells us nothing about “how to determine
what the law is, since they do not tell us how to determine what effect those
sources have on the content of the law.” “The practice of determining what the
law is”, Smith rightly reminds us “involves much more than simply determining

 If you doubt this, think about a case where one is dying of thirst, literally, and the only
option is to drink the polluted water; one may well decide that it‟s worth taking the risk.
Which means that the person made a decision on the balance of reasons, clearly suggesting
that there is a reason to drink the polluted water, even if it is very risky.

what the sources of law are; it is also centrally concerned with the effect those
sources have on the content of law.”10
All this leads Smith to conclude that I have to concede “that rules of
recognition do not fully regulate the practice of determining what the law is, in
the way that the rules of chess fully regulate the practice of chess.” I do,
actually, concede this last point, but I fail to see how it undermines my thesis.
Let me try to explain. First, a bit more on the similarity between chess and law.
In both cases, I argued, the rules that constitute the practice are surface
conventions of deeper conventions: in the case of chess, deep conventions that
constitute our practices of playing games, and in the case of law, deep
conventions that constitute ways of organizing a legal system. Thus, in both
cases, the surface conventions are already guided, as it were, by being
instantiations of certain values and shared understandings about the nature of
the practice and its underlying conventions. These deep conventions and the
values we associate with them are bound to have an effect on the ways in which
participants interpret the various surface conventions and ways in which they
practice them.
But now to the main point: what is it that the conventional rules of recognition
determine? One lesson we should have learned from Kelsen and Hart is that
the idea of legal validity requires a normative framework.11 In every legal system
in place, certain actions and events in the world are taken to have a determinate
legal significance, while countless others have none. The rules of recognition
that prevail in the relevant society constitute this normative framework. They
constitute grounds for ascribing legal significance to the actions and events that
are deemed legally valid in some specified ways. They tell us how law is created,
modified, or abolished in the relevant legal system, who gets to have these legal
effects and how, etc. So in this sense, we can say that the rules of recognition
identify the sources of law. But they do a bit more than that. In fact, Smith is
quite right to argue that practices of determining what the law requires in
particular cases takes more than identifying the sources of law. A central case in
point concerns the practices of statutory and constitutional interpretation. We
know that in the US, Congress enacts federal law; but how acts of Congress are
interpreted surely bears on what the law in the US legal system is.
Are these practices of interpretation, and similar norms that determine how
different sources of law bear on law‟s content, conventional? Do they form
part of the rules of recognition? The answer to the first question is that some

     Pp. […..]
     See my Philosophy of Law (Princeton, 2011), chapters 1 & 2.

of them are, while others are not, and that it partly depends on the actual legal
culture that prevails in the relevant legal system. For example, some norms of
statutory interpretation might become conventional in a certain legal system,
while remaining unsettled or excluded in other systems. In the US federal
courts, for example, there is a conventional norm requiring judges to pay
attention to the intentions of the legislature in their practices of statutory
interpretation. Though this norm is contested and not universally applied, it is,
by and large, a convention in US federal courts. Needless to say, this
convention might dwindle and cease to be practiced, and in many other legal
system, no such convention prevails. Is this convention of statutory
interpretation part of the rules of recognition of the US legal system? I am not
sure that a determinate answer is possible here. It is, certainly, part of the
interpretative culture of federal courts in the US, and as such, it affects the
ways in which the legislature behaves and ways in which its enactment
procedures are conducted and recorded. So in this respect, at least, the
conventions of statutory interpretation would certainly have an effect on
modes of law creation and on what counts as making law in the US federal
system. Similar conventional norms we find about ways in which, for example
judicial decisions are rendered, published, etc. However, we should certainly
not assume that all mattes of statutory and constitutional interpretation are
necessarily settled by some norms or conventions that prevail in a legal system;
many of these issues remain controversial and normatively unsettled.
None of this, by the way, is unique to law. As I mentioned in my discussion of
deep conventions, surface conventions often come at different levels of depths
and specificity. The general conventions of religious symbolism in European
art, for example, where at times further instantiated by more specific
conventions, such as those of color symbolism, conventions about modes of
depiction of biblical stories, etc. And of course, these conventions varied
between different genres and changed over time.12
Having said this, we should also realize that not all the norms that bear on
determining the content of law in a given legal culture are, necessarily,
conventional. Here‟s one example: acts of creating law are communicative acts,
they are, basically, speech acts. Understanding the law is first and foremost a
case of grasping what lawmakers (legislatures, agencies, judges, etc) say. (Many
doubt that this is all that there is to it, but this is not something we need to
settle here.) Now, under normal conditions, understanding the content

  Nigel Simmonds, in his review of my book (International Journal of Law in Context, vol 6,
(2010), 411), raised some doubts about these examples, though not, I think, against the idea
they are meant to exemplify.

conveyed by a speech act is partly determined by some normative assumptions
that Grice famously labeled maxims of conversation. I am not suggesting that
Grice‟s maxims or norms of conversation apply to legal speech as they apply
to ordinary conversations. On the contrary, I have argued elsewhere that we
have reasons to doubt this. But some of them do, and Gricean maxims of
conversation are not conventional. So there are, certainly, some norms that
bear on how we grasp the content of the law that are not conventional.13
I hope you can see where I am heading with all of this: the dilemma Smith
forces on my view is a spurious dichotomy. The law, I argued, has conventional
foundations. Conventions of recognition determine, in each and every legal
system, what kind of actions and events have the legal significance that we
ascribe to them. We can call this function “identifying the sources of law”, if
we want. These norms, and others further instantiating them, also regulate, to
some extent, ways in which questions about the content of particular legal
requirements are resolved or interpreted in the relevant legal culture. However,
these matters vary from one legal culture to another, some of them
conventionally or otherwise normatively determined, while others are not.
Surely, Smith is right to claim that law is much more complex than chess. Law
is, after all, probably the most complex and elaborate institution that exists in
human societies. Chess, with all due respect, is only a game. So it should
surprise no one that the analogies between the conventional foundations of a
game, like chess, and legal institutions, is very limited and does not explain
every aspect of our legal practices. Arguing, as I do, that law has some
conventional foundations does not commit me to the view that just about
anything that bears on ways in which we determine what the law requires is
conventionally settled.

3. Performatives and Promises.

There is a long tradition in philosophy, probably originating with Hume,
maintaining that promises are essentially conventional. According to this line of
thought, called the practice theory of promising, “promises belong to an
interesting class of acts that cannot be performed other than by invoking some

  See chapter 4, and “Can The Law Imply More than it Says? On Some Pragmatic Aspects
of Strategic Speech”, in Marmor & Soames (eds.), The Philosophical Foundations of Language in
Law, (Oxford 2011), 83.

convention”, as Sheinman puts it.14 And Sheinman sets out in his critique to
defend this view of promising against some skepticism I have expressed about
the practice theory in my book. In fact, Sheinman‟s interesting suggestions
about the nature of promises do not go against my general views about
conventions; on the contrary, his theory of promises employs my distinction
between deep and surface conventions, suggesting that promises can only be
understood on the background of surface conventions that are domain-specific,
as he calls them. The problem with the practice theory of promising, on his
view, stems from failure to realize that the universal convention of promising is
a deep convention, instantiated in particular cases by the domain-specific
conventions of particular promising practices. Although I find Shienman‟s
theory very interesting, and certainly in line with my general theory of
conventions, I am still not convinced that we need any kind of a practice theory
to account for the nature of promises.
A philosophical theory of promises has never been part of my project in this
book or elsewhere. Promises come up in my discussion in two different
contexts, none of which are meant to form part of anything one can call a
theory of promises. I discuss promises in the context of performative speech
acts, arguing that the making of a promise is not the kind of performative
speech act that requires a conventional background, and in the context of the
morality of conventions, as a possible example of cases in which there might be
a moral obligation to comply with conventions, doubting that promising is
such a case. Sheinman acknowledges all of this, but he argues that I am actually
wrong on both accounts. So let me reiterate some of the issues involved here,
beginning with some remarks on the nature of performative speech acts.
Consider the following speech acts:
          1. (Uttered by the chair-person of an official meeting --)
          (a) “The meeting is adjourned”
          (b). “Let‟s wrap it now and continue tomorrow”.
          2. (a). “I promise to be there at 7:00”
          (b) “Sure, I‟ll be there at 7:00”
          3. “Hi there”.

     “The Conventionality of Promising: A Defense”, at […]

The performative speech acts exemplified by (1) – (3) are of three different
types. Performatives of type (1) I called institutional performatives, because the
performative aspect of such speech acts requires some institutional setting, that
may or may not be conventional, determining that by saying X one has
performed Y. Normally, such performatives are also role-specific; the
institutional rules that determine that by saying X one has performed Y, also
tend to determine the institutional role of the relevant speaker whose utterance
of X, under certain circumstances, counts as Y. Notice, however, that the
relevant explicit performative formula is not necessary in such cases, though
this may partly depend on the specific rules of the institution in question. One
can adjourn a meeting by expressing 1(a), or by using some equivalent,
inexplicit performative such as 1(b).
Performatives of type (2), that I called general performatives, are best explained by
the statement theory15: In such cases, including, I argued, in the case of making
a promise, the speaker makes a statement about his or her attitudes or
commitments, a statement that under normal conditions is rendered true by its
utterance alone. To say that “I promise to ” is to make a statement that I
undertake a commitment to , and the expression of this commitment is what
makes the statement true. It is a commitment because I have just stated it.
More about this in the sequel.
Finally, I argued that there is a third class of performatives, exemplified by (3),
in which the speaker does not express a statement of any kind because such
expressions are not truth-apt. In uttering a greeting, like “hi”, or “good
morning”, etc, one makes a move in a conventional social game, as it were, and
it is this conventional practice that confers on the expression the particular
performative significance that it has. Usually, such conventional performatives, as I
labeled this type, require the use of some explicit performative formula, and
mostly because the social-performative meaning of the expression is
determined by the conventions constituting the relevant social practice.
Sheinman objects to my distinction between type (2) and type (3) of
performatives. As far as I can tell, on two grounds. First, he argues that these
two accounts are not mutually exclusive; a given speech act may both be a
statement, along the lines suggested by the Bach & Harnish theory that I rely
upon, and require a convention that is necessarily invoked, along the lines

  So called by Bach & Harnish, ( see their “How Performatives Really Work”, 15 Linguistics
and Philosophy, (1992), 93) first suggested by Peter Strawson in “Intention and Convention in
Speech Acts”, in Logico-Linguistic Papers (Methuen London, 1971), 170.

suggested by Searle.16 Second, Sheinman also doubts that performatives of type
(2) are typically statements, at least when we consider cases where the explicit
performative formula is not used. I think that he is wrong about both of these
Let me briefly answer the first point, which I find a bit perplexing. Searle claims
that performatives, quite generally, require some conventional setting
constituting the fact that saying X in circumstances C counts as doing Y. Bach
& Harnish claim, and I agree with respect to performatives of type (2), that this
condition that Searle claims to be necessary is not, in fact, a condition that is
needed. We can fully account for the performative aspect of speech acts of type
(2), without invoking any background conventions. Suggesting that these two
accounts are not mutually exclusive misses the main point that is at stake here.
The question is whether the conventional background that Searle claims is
necessary to account for the performative aspect of certain expressions is,
indeed, necessary or not. If Searle is right, and we cannot make sense of how
saying that X counts as doing Y without some conventions at the background,
then Bach & Harnish are wrong. And if the latter are right to claim that no
such conventions are needed, then Searle is wrong. As simple as that.
Before I get to Sheinman‟s second, and presumably main argument, let me
make a general comment. I have largely assumed, and still think it quite
important, that no theory of general performatives would be very convincing if
it makes too much of the distinction between explicit and inexplicit
performatives. Whether a speaker uses the explicit performative verb or not,
cannot make a significant difference to what makes a speaker‟s expression have
the kind of performative aspect that it has. Promising would be a good example.
There are countless ways in which a person can express the fact that she
undertakes a promise without using the word “promise”. I can tell you that “I
promise to be at your house at 7PM tonight”, but I can also convey the same
idea in countless other ways, such as “sure, I‟ll be there at 7”, or “see you at 7,
then”, and so on and so forth. Any theory that makes too much of the explicit
use of “I promise” should be very suspect for that reason alone.
One might think that promising is a special case. The argument would rely on
examples where one can express a commitment of some sort without actually
promising to do it. I can tell you, for example, “I‟ll be there at 7, but I do not
promise”, without expressing something contradictory. But these examples are
very misleading. They would seem to suggest that inexplicit performatives are,
at best, implicatures of promising, and as with any regular implicature, the

     Pp. […]

implicated content can be cancelled explicitly by the speaker. So the idea is that
by saying something like “I‟ll be there at 7”, the speaker would normally
implicate that she undertook a promise, but this is only a conversational
implicature that the speaker can cancel by clarifying that she does not, actually,
promise anything.17 I find this analysis, however, wrong-headed. In order to
explain what is going on here, we need some basic distinctions.
There are at least three levels of content that can be conveyed by an occasion
of speech: communicative content may consist of semantic content, assertive
content, and some further content that is implicated by the speech in the
particular context of its utterance. The semantic content of an utterance is
determined by the literal or lexical meaning of the words used and the syntax of
the sentence. This kind of content, however, is often not sufficient to
determine what the speaker actually said or asserted. The assertive content of
an expression is normally the truth-evaluable proposition that is conveyed by
the speaker in the particular context of the utterance. It has long been
recognized by philosophers of language that assertive content often goes
beyond the semantic content expressed by a speaker in a given set of
circumstances; pragmatic features of the conversation often partly determine
what the speaker said or asserted (some examples will be discussed below).
Finally, in addition to content that is asserted by a speaker, there might also be
some content implicated by it, such as cases of conversational implicatures
articulated by Grice.
Now, consider cases where the speaker says “I‟ll be there at 7”; is this a
promise or not? The answer is that it depends on some pragmatic features of
the conversation. In some cases it might be the expression of a promise, while
in others it is not. This is a question about the content that the speaker asserts in
the circumstances, not necessarily or even typically a question of whether the
speaker implicates some additional content beyond what she said. In other
words, there is no denying that pragmatic aspects of communication are often
needed to determine whether an expression is the assertion of a promise or not.
But there is certainly nothing special about that. Pragmatic features often partly
determine assertive content. And sometimes, of course, we want to make sure
that what we assert is, or is not, a promise. But such assurances are not always
needed. And vice versa, one can use the explicit performative “I promise”
without actually making/asserting a promise. (Suppose you ask me what is it
that I am doing while signing a document, and I reply to you that I promise to
buy this car John is selling me; I do not express any promise in saying this.)
   At one point (p. …), Sheinman seems to attribute such a view to me, but maybe just as a
friendly (or not so friendly) suggestion.

To sum up: my point is that the question of what performative speech act one
performs is a question about the assertive content conveyed in a given set of
circumstances, and the use of explicit performative formula is just one tool,
among many, that a speaker can employ in conveying such content. Therefore,
to reiterate, I would find very suspect any theory of general performatives that
puts too much weight on the distinction between explicit and inexplicit
This should help us to see why Sheinman‟s objection to my thesis that general
performatives are statements that are rendered true by their expression alone is
misguided. His objection rests on the claim that my thesis does not work in
cases of inexplicit performatives, because the content that is rendered true by
the expression of such inexplicit performatives is not the promise. As he puts
it: “The semantic content of „I promise‟ is that Hanoch promises. This is also
the proposition rendered true by the utterance. But the semantic content of „I
will be there‟ is clearly not that Hanoch promises to be there; it‟s that Hanoch
will be there. To render this proposition true, my utterance would have to carry
me somewhere.” 18
Here‟s the reply: truth and falsehood is ascribed to assertive content, not to
semantic content. As I already mentioned, the semantic content of an
expression, determined by the literal meaning of the words used and the syntax
of the expression, is only a vehicle for conveying some truth-evaluable or,
generally, assertive content. And quite often, some pragmatic and contextual
features of the speech situation are required to grasp what the speaker asserted
under the specific circumstances of the speech act. There is really nothing
special about promises or performatives here. When I enter a large classroom
to give a lecture and ask “Can everybody hear me?‟, nobody would assume that
“everybody” refers to the entire world, or even to the entire campus. What I
ask is whether everybody in the room can hear me. And this is not something I
implicate beyond what I say; it is, under the circumstances, what I say (or,
rather, what I ask, in this case). Similarly, if you ask me how many children do I
have, and I reply that I have two, you would be quite right to assume that I
asserted that I have two and only two children. But if you ask me if I have a
couple of beers in my fridge and I answer affirmatively, you would not think

  Sheinman makes the same point about Strawson‟s famous example of warning by saying
“the ice over there is very thin”.

that I have mislead you when you open the fridge and find a whole stack of
beers there.19
In short, various pragmatic features of conversational situations are often
essential to grasping what a speaker has said or asserted. And of course, the
same holds true about general inexplicit performatives: whether the utterance
of the words “I will be there” amounts to the expression viz, assertion, of a
promise or not, depends on the pragmatics of the conversation. Suppose, to
give a simple example, that Hanoch and I consider going to a restaurant tonight,
and after discussing some details about where and when to meet, he asks me:
“Shall we meet at the restaurant at 7, then?”, and I reply “I‟ll be there”. Surely,
my reply is a promise, if anything is. In this case, I see no problem whatsoever
with the suggestion that by saying that “I‟ll be there” I have made a statement
with the content of undertaking a promise to be at the restaurant at 7, which is
rendered true by expressing the utterance. Of course, in other contexts, the
same expression might assert something else. If Hanoch asks me, out of sheer
curiosity, if I plan on attending a certain lecture later this afternoon, and I reply
by saying “Yes, I‟ll be there”, I have not asserted any promise, only an
intention to do something later.
None of this settles the question of whether there are conventions of
promising, nor does it settle the question that mostly interests moral
philosophers, whether we can account for what makes promises binding
without invoking some conventional practice that promising necessarily relies
on. Since it was not part of my project in the book to offer a theory of
promises, I will certainly not attempt one here. Sheinman‟s own views about
these issues are very interesting, and I agree with him that if his suggestions
about domain-specific (viz, surface) conventions of promising are correct, his
version of the practice theory would not be vulnerable to the critique I
expressed in my book. I doubt, however, that Sheinman has presented us with
a convincing case for his views, and in two respects.
First, Sheinman claims that my argument, or rather, as he sees it, the lack of it,
for deep conventions of games is more tenuous then the case one can make for
the prevalence of deep conventions of promising. The problem is that I
actually gave a fairly detailed account of what those deep conventions of games
are, how they can be violated, and why they need to be instantiated by surface
conventions.20 Nothing of the sort is suggested by Sheinman about the deep or

   The examples are taken from S. Soames, Philosophical Essays, Volume 1, (Princeton 2009),
chapter 11.
   See pp. 59-61, 73-76, & 100-102.

universal convention of promising. What is the convention here, and what is
the alternative convention we, humans, that is, could have followed instead?
Sheinman does not tell us what is the content of the universal convention of
promising. Instead, he asks us to imagine a world of “the nonpromissories” [sic]
which is characterized by two features: creatures in this world “have no
promises, acts of committing oneself to doing things in the future.” Instead,
they “do not state an intention to do something unless they do it.” So
Sheinman seems to suggest that if we have a reliable practice of stating an
intention, namely, a world in which any expression of an intention to act is a
very reliable indicator that one will in fact act according to the intention, this
would be an adequate substitute or alternative to our practices of promising.
But the simple answer is that it is not; promises are the kind expressions of
intentions to act that are aimed at inducing the hearer to rely on that intention
in some morally significant way. The veracity of the intention to act is not the
problem promises are there to solve, as it were. Many expressions of an
intention to do something in the future, even if addressed to a particular hearer,
are not such they amount to a promise, or to something that the speaker would
be morally bound to comply with. Telling you, for example, that I intend to
begin a serious diet tomorrow, is not a promise (to you, that is); and if I fail to
start that diet, I may fail myself, but not you. So I just don‟t see how this world
of “nonpromisories” is one in which we have a conventional alternative to our
practices of promising.
Sheinman‟s suggestion of domain-specific promising practices I find more
plausible and quite interesting. But I have to admit that I find his specific
example of that a bit unfortunate. The example he discusses at some length is
the practice of promising to surrender in military conflict by raising a white flag.
I certainly agree that this would be a case in which there is a conventional
practice about certain ways of expressing a promise; as I mentioned in the book, I
do not deny that there are such conventions. But Sheinman‟s claim is different.
He wants to show by this example that only by invoking a conventional,
domain-specific, practice we can make sense of the surrendering promise. And
this I doubt. Imagine a world in which the convention is not to raise a white
flag, but to raise a big white board on which the following words are inscribed:
“We Surrender Unconditionally Here and Now”. In fact, each army unit is
equipped with those boards, just in case…. How would this world differ, then,
from ours, where a group of miserable soldiers under attack have no white
banner to raise, but they manage to display a big board saying “We Surrender
Unconditionally Here and Now”. Would they fail to convey their message of
promising to surrender in a way that soldiers in the imaginary society do not?

As long as the enemy can read English, I am not sure that I would put more
faith in the white flag compared with the written words.
Let me reiterate that my doubts here pertain to the specific example Sheinman
relies upon. I don‟t see how it proves his point. But perhaps better examples
can be thought of, I cannot speculate. Generally speaking, my suspicion about
any practice theory of promising stems from the same intuition that motivates
Scanlon‟s alternative account, namely, the realization that promising is just one
member of a family of cases in which people induce others to have certain
expectations about their future conduct in ways which would be morally wrong
to frustrate. But again, nothing in my book was aimed at substantiating this


Shared By: