You are on page 1of 20

The theory of conversational implicature is a good example.

It is often summarised as one of


the earliest and most successful attempts to explain the fact, familiar to common sense, that
what people literally say and what they actually mean can often be quite different matters. In
particular, the theory is concerned with some apparent discrepancies between classical logic
and natural language. More generally, it addresses the question of whether the meaning of
everyday language is best explained in terms of formal linguistic rules, or in terms of the
vagaries and variables of human communication. Grice’s theory developed against the
background of a sharp distinction of approaches to these issues. Philosophers such as Bertrand
Russell, who saw logic as the appropriate apparatus for explaining meaning, dismissed the
differences displayed by natural language as examples of its inherent imperfection. Everyday
language was just too messy and imprecise to form an appropriate topic for philosophical
inquiry. The opposing view is perhaps best summed up by the slogan from Wittgenstein’s later
work that ‘meaning is use’. 3 Philosophers from this school of thought argued that if natural
language diverges from logical meaning, this is simply because logic is not the appropriate
philosophical tool for explaining language. Meaning in language is an important area of study
in its own right, but can be considered only in connection with the variety of ways in which
language is used by speakers. Grice’s approach to this debate was to argue that both views were
wrong-headed. He proposed to think outside the standard disjunction of positions. Logic cannot
give an adequate account of natural language, but nor is language simply a multifarious
collection of uses, unamenable to logical analysis. Rather, logic can explain much, but not all,
of the meaning of certain natural language expressions. More generally, formal semantic rules
play a vital part in explaining meaning in everyday language, but they do not do the whole job.
Other factors of a different but no less important type are also necessary for a full account of
meaning. Most significantly, Grice argued that these other, non-semantic factors are not a
random collection entirely dependent on individual speakers and contexts, but can be
systematised and explained in terms of general principles and rules. The principles that explain
how natural language differs from logic can also explain a wide range of other features of
human communication.

Meaning

‘Meaning’ offers a deceptively straightforward-looking account of meaning, apparently based


on a common-sense belief about the use of language. In effect, Grice argues that what speakers
mean depends on what they intend to communicate. However, his project is much more
ambitious, and its implications more far-reaching, and also more problematic, than this
suggests. First, he introduces hearers, as well as speakers, into the account of meaning. The
intention to communicate is not sufficient; this intention must be recognised by some audience
for the communication to succeed. Hence the speaker must have an additional intention that
the intention to communicate is recognised. More significantly, Grice proposes to broaden out
his account to offer a theory of linguistic meaning itself. Linguistic meaning depends on
speaker meaning, itself explained by the psychological phenomenon of intention; conventional
meaning is to be defined in terms of psychology. In the opening paragraphs of ‘Meaning’, Grice
does something Austin seems never to have attempted. He focuses the methodology of ordinary
language philosophy on the concept of meaning itself, considering and comparing some
different ways in which the word ‘mean’ is used. His conclusion is that there are two different
uses of ‘mean’, reflecting two different ways in which we generally conceive of meaning. The
first use is exemplified by sentences such as ‘those spots mean measles’ and ‘the recent budget
means that we shall have a hard year’. The second includes examples such as ‘those three rings
on the bell (of the bus) mean that the bus is full’ and ‘that remark, “Smith couldn’t get on
without his trouble and strife”, meant that Smith found his wife indispensable’. Grice draws up
a list of five defining differences between these sets of examples, differences themselves
expressed in terms of linguistic relations. For instance, the first, but not the second set of
examples entail the truth of what is meant. It is not possible to add ‘but he hasn’t got measles’,
but it is possible is to add ‘but the bus isn’t in fact full – the conductor has made a mistake’.
Further, the examples in the first set do not convey the idea that anyone meant anything by the
spots or by the budget. It is, however, possible to say that someone meant something by the
rings on the bell (the conductor) and by the remark (the original speaker). And only in the
second set of examples can the verb ‘mean’ be followed by a phrase in quotation marks. So
‘those three rings on the bell mean “the bus is full”’ is possible, but not ‘those spots mean
“measles”’. Grice characterises the difference between his two sets of uses for ‘mean’ as a
difference between natural and non-natural meaning. The latter, which includes but is not
exhausted by linguistic meaning, he abbreviates to ‘meaning NN ’. In pursuing the question of
what makes ‘meaning NN ’ distinctive, he considers but rejects a ‘causal’ answer. He offers
the following as a summary of this type of answer, paraphrasing and partly quoting C. L.
Stevenson:
Logic and Conversation

As Grice’s enthusiasm for ordinary language philosophy became increasingly qualified during
the 1950s, his interest was growing in the rather different styles of philosophy of language then
current in America. Recent improvements in communications had made possible an exchange
of ideas across the Atlantic that would have been unthinkable before the war. W. V. O. Quine
had made a considerable impression at Oxford during his time as Eastman Professor. Grice was
interested in Quine’s logical approach to language, although he differed from him over certain
specific questions, such as the viability of the distinction between analytic and synthetic
statements. Quine, who was visiting England for a whole year, and who brought with him
clothes, books and even provisions in the knowledge that rationing was still in force, travelled
by ship. 1 However, during the same decade the rapid proliferation of passenger air travel
enabled movement of academics between Britain and America for even short stays and lecture
tours. Grice himself made a number of such visits, and was impressed by the formal and theory-
driven philosophy he encountered. Most of all he was impressed by the work of Noam
Chomsky. It may seem surprising that the middle-aged British philosopher interested in the
role of individual speakers in creating meaning should cite as an influence the young American
linguist notorious for dismissing issues of use in his pursuit of a universal theory of language.
But Grice was inspired by Chomsky’s demonstration in his work on syntax of how ‘a region
for long found theoretically intractable by scholars (like Jesperson) of the highest intelligence
could, by discovery and application of the right kind of apparatus, be brought under control’.
2 Less formally, he expressed admiration for an approach that did not offer ‘piecemeal
reflections on language’ but rather where ‘one got a pictureof the whole thing’. 3 Chomsky’s
first and highly influential book Syntactic Structures was well known among Grice’s Oxford
contemporaries. The Play Group worked their slow and meticulous way through it during the
autumn of 1959. 4 Austin, in particular, was extremely impressed. Grice characterised and
perhaps parodied him as revering Chomsky for his sheer audacity in taking on a subject even
more sacred than philosophy: the subject of grammar. 5 Grice’s own interest was focused on
theory formation and its philosophical consequences. Chomsky was taking a new approach to
the study of syntax by proposing a general theory where previously there had been only
localised description and analysis. He claimed, for instance, that ideally ‘a formalised theory
may automatically provide solutions for many problems other than those for which it was
explicitly designed’. 6 Grice’s aim, it was becoming clear, was to do something similar for the
study of language use. Meanwhile, ordinary language philosophy itself was in decline. As for
any school of thought, it is difficult to determine an exact endpoint, and some commentators
have suggested a date as late as 1970. However, it is generally accepted that the heyday of
ordinary language philosophy was during the years immediately following the Second World
War. The sense of excitement and adventure that characterised its beginning began to wane
during the 1950s. Despite his professed dislike of discipleship, Austin seems to have become
anxious about what he perceived as the lack of a next generation of like-minded young
philosophers at Oxford. It became an open secret among his colleagues that he was seriously
contemplating a move to the University of California, Berkeley. 7 No final decision was ever
made. Austin died early in 1960 at the age of 48, having succumbed quickly to cancer over the
previous months. Reserved and private to the last, he hid his illness from even his closest
colleagues until he was unable to continue work. His death was certainly a blow to ordinary
language philosophy, but it would be an exaggeration to say that it was the immediate cause of
its demise. Grice, who seems to have been regarded as Austin’s natural deputy, stepped in as
convenor of the Play Group, which met under his leadership for the next seven years.
Individuals such as Strawson, Warnock, Urmson and Grice himself continued to produce work
with recognisably ‘ordinary language’ leanings throughout the 1960s. Grice’s interests at this
time were not driven entirely by philosophical trends in Oxford and America; he was also
turning his attention to some very old logical problems. In particular, he was interested in
questions concerning apparent counterparts to logical constants in natural language. For
instance, in the early 1960s he revisited a theme he had f irst considered before the war, when
he gave a series of lectures on ‘Negation’. In these, he concerns himself with the analysis of
sentences containing ‘not’, and with the extent to which this should coincide with a logical
analysis of negation. Consideration of a variety of example sentences leads him to reject the
simple equation of ‘not’ with the logical operation of switching truth polarity, usually positive
to negative. He argues that ‘it might be said that in explaining the force of “not” in terms of
“contradictory” we have oversimplified the ordinary use of “not”.’ In another lecture from the
series he suggests that the lack of correspondence between ‘not’ and contradiction ‘might be
explained in terms of pragmatic pressures which govern the use of language in general’. 8 Grice
was hoping to find not just an account of the uses of this particular expression, but a general
theory of language use capable of extension to other problems in logic. He would have been
familiar enough with such problems. The discussion of some of them dates back as far as
Aristotle, in whose work he was well read even as an undergraduate. In Categoriae , Aristotle
describes not just categories of lexical meaning, but also the types of relationships holding
between words. To the modern logician, the use of terms in the following passage may be
obscure, but the relationship of logical entailment is easily recognisable. One is prior to two
because if there are two it follows at once that there is one whereas if there is one there are not
necessarily two, so that the implication of the other’s existence does not hold reciprocally from
one. 9 The relationship between ‘two’ and ‘one’, or indeed between any two cardinal numbers
where one is greater than the other, is one of asymmetrical entailment. ‘Two’ entails ‘one’, but
‘one’ does not entail ‘two’. A similar relationship holds between a superordinate and any of its
hyponyms, or between a general and a more specific term. To use Aristotle’s example: ‘if there
is a fish there is an animal, but if there is an animal there is not necessarily a fish.’ 10 The
asymmetrical nature of this relationship means that use of the more general term tells us nothing
at all about the applicability of the more specific. Aristotle also considers the relative
acceptability of general and specific terms, and in doing this he goes beyond a narrowly logical
focus. For if one is to say of the primary substance what it is, it will be more informative and
apt to give the species than the genus. For example, it would be more informative to say of the
individual man that he is a man than that he is an animal (since the one is more distinctive of
the individual man while the other is more general). 11 Applying the term ‘animal’ to an
individual tells us nothing about whether that individual is a man or not. Therefore, if the more
specific term ‘man’ applies it is more ‘apt’, because it gives more information. This same point
arises in a discussion of the applicability of certain descriptions later in Categoriae . Aristotle
suggests that: ‘it is not what has not teeth that we call toothless, or what has not sight blind, but
what has not got them at the time when it is natural for it to have them.’ 12 A term such as
‘toothless’ is only applied, because it is only informative, in those situations when it might be
expected not to apply. Here, again, the discussion of what ‘we call’ things goes beyond purely
logical meaning to take account of how expressions are generally used. Logically speaking a
stone could appropriately be described as toothless or blind; in actual practice it is very unlikely
to be so described. Grice’s self-imposed task in considering the general ‘pragmatic pressures’
on language use was, at least in part, one of extending Aristotle’s sensitivity to the standard
uses of certain expressions, and examining how regularities of use can have distorting effects
on intuitions about logical meaning. He was by no means the first philosopher to consider this.
For instance, John Stuart Mill, in his response to the work of Sir William Hamilton, draws
attention to the distinction between logic and ‘the usage of language’. 13 He reproaches
Hamilton for not paying sufficient attention to this distinction, and suggests that this is enough
to explain some of Hamilton’s mistakes in logic. Mill glosses Hamilton as maintaining that
‘the form “Some A is B”...ought in logical propriety to be used and understood in the sense of
“some and some only ”’. 14 Hamilton is therefore committed to the claim that ‘all’ and ‘some’
are mutually incompatible: that an assertion involving ‘some’ has as part of its meaning ‘not
all’. This is at odds with the observations on quantity in Categoriae and indeed, as Mill
suggests, with ‘the practice of all writers on logic’. Mill explains this mistake as a confusion
of logical meaning with a feature of ‘common conversation in its most unprecise form’. In this,
he is drawing on the extra, non-logical but generally understood ‘meanings’ associated with
particular expressions. In a passage that would not be out of place in a modern discussion of
linguistics, Mill suggests that: If I say to any one, ‘I saw some of your children to-day,’ he
might be justified in inferring that I did not see them all, not because the words mean it, but
because, if I had seen them all, it is most likely that I should have said so. 15 Mill draws a
distinction between what ‘words mean’ and what we generally infer from hearing them used.
In what can be seen as an extension of Aristotle’s discussion of ‘aptness’, he argues that it is a
mistake to confuse these two very different types of significance. A more specific word such
as ‘all’ is more appropriate, if it is applicable, than a more general word such as ‘some’.
Therefore, the use of the more general leads to the inference, although it does not strictly mean,
that the more specific does not apply. ‘Some’ suggests, but does not actually entail ‘not all’.
Besides his interest in logical problems with a venerable pedigree, Grice was also concerned
with issues familiar to him from the work of recent or contemporary philosophers. In both
published work and informal notes he frequently lists these and arranges them in groups. Part
of his achievement in the theory he was developing lay in seeing connections between an
apparently disparate collection of problems and countenancing a single solution for them all.
For instance, in Concept of Mind , Ryle argues that, although the expressions ‘voluntary’ and
‘involuntary’ appear to be simple opposites, they both require a particular condition for
applicability, namely that the action in question is in some way reprehensible. If they were
simple opposites, it should always be the case that one or other would be correct in describing
an action, yet in the absence of the crucial condition, to apply either would be to say something
‘absurd’. Similarly, although if someone has performed some action, that person must in a
sense have tried to perform it, it is often extremely odd to say so. In cases where there was no
diff iculty or doubt over the outcome, it is inappropriate to say that someone tried to do
something: so much so that some philosophers, such as Wittgenstein, have claimed that it is
simply wrong. Another related problem is familiar from Austin’s work; it is the one summed
up in his slogan ‘no modification without aberration’. For the ordinary uses of many verbs, it
does not seem appropriate to apply either a modify ing word or phrase or its opposite. Austin
was therefore offering a generalisation that includes, but is not restricted to, Ryle’s claims about
‘voluntary’ and ‘involuntary’. For many everyday action verbs, the act described must have
taken place in some non-standard way for any modification appropriately to apply. Austin
offers no theory based on this observation, and indeed Grice was unimpressed by it even as a
generalisation; he claimed in an unpublished paper that it was ‘clearly fraudulent’. ‘No
“aberration” is needed for the appearance of the adverbial “in a taxi” within the phrase “he
travelled to the airport in a taxi”; aberrations are needed only for modifications which are
corrective qualifications.’ 16 Grice’s general account of language, conceived with the twin
ambitions of refining his philosophy of meaning and of explaining a diverse range of
philosophical problems, gradually developed into his theory of conversation. Like his project
in ‘Meaning’, this draws on a ‘common-sense’ understanding of language: in this case, that
what people say and what they mean are often very different matters. This observation was far
from original, but Grice’s response to it was in some crucial ways entirely new in philosophy.
Unlike formal philosophers such as Russell or the logical positivists, he argued that the
differences between literal and speaker meaning are not random and diverse, and do not make
the rigorous study of the latter a futile exercise. But he also differed from contemporary
philosophers of ordinary language, in arguing that interest in formal or abstract meaning need
not be abandoned in the face of the particularities of individual usage. Rather, the difference
between the two types of meaning could be seen as systematic and explicable, following from
one very general principle of human behaviour, and a number of specific ways in which this
worked out in practice. In effect, the use of language, like many other aspects of human
behaviour, is an end-driven endeavour. People engage in communication in the expectation of
achieving certain outcomes, and in the pursuit of those outcomes they are prepared to maintain,
and expect others to maintain, certain strategies. This mutual pursuit of goals results in
cooperation between speakers. This manifests itself in terms of four distinct categories of
behaviour, each of which can be summarised by one or more maxims that speakers observe.
The categories and maxims are familiar to every student of pragmatics, although in later
commentaries they are often all subsumed under the title ‘maxims’. Category of Quantity
1.Make your contribution as informative as is required (for the current purposes of the
exchange). 2.Do not make your contribution more informative than is required. Category of
Quality 1.Do not say what you believe to be false. 2.Do not say that for which you lack adequate
evidence. Category of Relation 1.Be relevant. Category of Manner 1.Avoid ambiguity of
expression. 2.Avoid ambiguity. 3.Be brief. 4.Be orderly. 17 Grice uses the simple notion of
cooperation, together with the more elaborate structure of categories, to offer a systematic
account of the many ways in which literal and implied meaning, or ‘what is said’ and ‘what is
implicated’, differ from one another. In effect, the expectation of cooperation both licenses
these differences and explains their usually successful resolution. Speakers rely on the fact that
hearers will be able to reinterpret the literal content of their utterances, or fill in missing
information, so as to achieve a successful contribution to the conversation in hand. The noun
‘implicature’ and verb ‘implicate’ (as used in relation to that noun) are now familiar in the
discussion of pragmatic meaning, but they were coined by Grice, and coined fairly late on in
the development of his theory. In early work on conversation he suggested that a ‘special kind
of implication’ could be used to account for various differences between conventional meaning
and speaker meaning. He ultimately found this formulation inadequate, together with a host of
other words such as ‘suggest’, ‘hint’ and even ‘mean’, precisely because of their complex pre-
existing usage both within and outside philosophy. The difference between saying and meaning
had an established philosophical pedigree, as well as a basis in common sense. A literature
dating back to G. E. Moore, but largely concentrated in the 1950s and 1960s, offers a significant
context to Grice’s work. Little of this is now read and none of it attempts anything as ambitious
as a generalising theory of communication. With typical laxness, Grice does not refer to this
body of work in any of his writings on the topic, although he must have been aware of it; much
of it originated from the tight circle of Oxford philosophy. Writing for his fellow philosophers,
and in the conventions ofhis time, he was able to refer vaguely and generally to the debate in
the expectation that the relevant context would be recognised. G. E. Moore had argued that
when we use an indicative sentence we assert the content of the sentence, but we also imply
personal commitment to the truthfulness of that content; ‘If I say that I went to the pictures last
Tuesday, I imply by saying so that I believe or know that I did, but I do not say that I believe
or know this.’ 18 This distinction was picked up by a number of philosophers in the 1940s and
1950s. In 1946, Yehoshua Bar-Hillel discussed the sense of ‘imply’ identified by Moore,
proposing to describe it as ‘pragmatical’. 19 Bar-Hillel identifies himself as a supporter of
‘logical empiricism’ (he quotes approvingly a comment from Carnap to the effect that natural
languages are too complex and messy to be the focus of rigorous scientific enquiry) and his
article is aimed explicitly at rejecting philosophy of the ‘analytic method’ in general and of
Moore in particular. However, he suggests that by using sentences that are ‘meaningless’ to
logical empiricists, such as the sentences of metaphysics or aesthetics, ‘one may nevertheless
imply sentences which are perfectly meaningful, according to the same criteria, and are perhaps
even true and highly important’. 20 A full account of the pragmatic sense of ‘implies’ might,
he predicts, prove highly important in discussing linguistic behaviour. A few years later D. J.
O’Connor contrasted the familiar ‘logical paradoxes’ of philosophy with less well known
‘pragmatic paradoxes’. O’Connor draws attention to certain example sentences (such as ‘I
believe there are tigers in Mexico but there aren’t any there at all’) which, although not logically
contradictory or necessarily false, are pragmatically unsatisfactory. His purpose is to
distinguish between logic and language and to urge philosophers to attempt to ‘make a little
clearer the ways in which ordinary languagecan limit and mislead us’. 21 The philosophical
significance of implication was also discussed by those more sympathetic to ordinary language,
in particular by members of the Play Group. Writing in Mind in 1952, J. O. Urmson
acknowledged the ‘implied claim to truth’ in the use of indicative statements, but felt the need
to clarify his terms: ‘The word “implies” is being used in such a way that if there is a convention
that X will only be done in circumstances Y, a man implies that situation Y holds if he does
X.’ 22 Urmson suggests the addition of an ‘implied claim to reasonableness’; ‘it is, I think, a
presupposition of communication that people will not make statements, thereby implying their
truth, unless they have some ground, however tenuous, for those statements.’ 23 Read with
hindsight, Urmson’s suggestion looks like a hint towards Grice’s second maxim ofQuality.
Another member of the Play Group, P. H. Nowell-Smith, appears to add something like a
maxim of Relevance, a notion that Grice discusses as early as the original version of ‘Meaning’.
In his book Ethics , Nowell-Smith suggests that there are certain ‘contextual implications’ that
generally accompany the use of words, but that are dependent on particular contexts. Without
reference to Urmson, he phrases these as follows: When a speaker uses a sentence to make a
statement, it is contextually implied that he believes it to be true. ...A speaker contextually
implies that he has what he himself believes to be good reasons for his statement....What a
speaker says may be assumed to be relevant to the interests of his audience. 24 For Nowell-
Smith the contextual rules are primary, and do not operate with reference to logical meaning;
in effect there is no ‘what is said’. In fact, he specifies that logical meaning is a subclass of
contextual implication, because it is meaning we are entitled to infer in any context whatsoever.
Other philosophers concerned with implication, however, did identify different levels of
meaning. In The Logic of Moral Discourse , 1955, Paul Edwards distinguishes between the
referent of a sentence (the fact that makes a sentence true or, in its absence, false: in effect the
truth-conditions) and ‘what the sentence expresses’ (anything it is possible to infer about the
speaker on the basis of the utterance). 25 Some attempted syntheses of thinking about
implication lacked much in the way of generalising or explanatory force. In his 1958 article
‘Pragmatic implication’, C. K. Grant’s account of a wide variety of linguistic phenomena leads
him to the claim that ‘a statement pragmatically implies those propositions whose falsity would
render the making of the statement absurd, that is, pointless.’ 26 What is implied by a statement
of p, he insists, is not just the speaker’s belief in p, but some further proposition. Isobel
Hungerland, writing in 1960, reviews the range of attempts to mediate between asserted and
implied meaning and exclaims ‘What a range of rules!’. 27 Her suggestion of a generalisation
across these rules is that ‘a speaker in making a statement contextually implies whatever one
is entitled to infer on the basis of thepresumption that his act of stating is normal.’ 28 Grice
was working on his own generalisation, which was to link together the range of principles of
language use into one coherent account, and to give this a place in a more ambitious theory of
language. If the maxims of Quality and of Relation were familiar from contemporary work in
Ethics, those of Quantity and Manner drew on Grice’s long-standing concern about stronger
and weaker statements. This was the idea that had interested him in his rough notes on ‘visa’
for his work on perception with Geoffrey Warnock. The earliest published indication appeared
in 1952, in a footnote to Peter Strawson’s Introduction to Logical Theory . In discussing the
relationship between the statement ‘there is not a book in his room which is not by an English
author’ and the assumption ‘there are books in his room’, Strawson draws attention to the need
to distinguish between strictly logically relations and the rules of ‘linguistic conduct’. 29 He
suggests as one such rule: ‘one does not make the (logically) lesser when one could truthfully
(and with equal or greater linguistic economy) make the greater, claim.’ It would be misleading,
although not strictly false, to make the less informative claim about English authors if in a
position to make the much more informative claim that there are no books at all. Strawson
acknowledges that ‘the operation of this “pragmatic rule” was first pointed out to me, in a
different connection, by Mr H. P. Grice.’ It was typical of Grice that he did not publish his own
account of this pragmatic rule until almost ten years after Strawson’s acknowledgement. ‘The
causal theory of perception’ appeared in the Proceedings of the Aristotelian Society , after
Grice delivered it as a symposium paper at the meeting of the Society in Cambridge in July
1961. His main focus is his long-standing interest in the status of our observations of material
objects. Grice offers tentative support for a causal account of perception. Theories of this type
date back at least to John Locke. They maintain that we directly perceive through our faculties
of sense a series of information, given the title ‘sense data’ by later philosophers, and infer that
these are caused by material objects in the world. For Locke, the inference was justifiable, but
his account of perception, and causal theories ever since, had been dogged by accusations of
scepticism. If we have no direct access to material objects, but only to the sense data they
putatively cause, the way is left open to doubt the independent existence of such objects.
Grice’s defence is tentative, even by his own circumspect standards. He proposes to advance a
version of the causal theory ‘which is, if not true, at least not obviously false’. 30 Some of the
defence itself is presented in quotation marks, as a fictional ‘opponent’ argues with a ficional
‘objector’ to the causal theory, implicitly distancing Grice from the arguments on either side.
This tentativeness might in part be explained by the extreme heresy of Grice’s enterprise for a
philosopher of ordinary language. Some of Austin’s most strident remarks about ordinary
language had appeared in his attack in Sense and Sensibilia on philosophical theories of
perception, and their reliance on the obscure and technical notion of ‘sense data’. Austin’s chief
target here had been phenomenalism which, in claiming that sense data are the essence of what
we think of as material objects, is often seen as a direct competitor to the causal theory of
perception. Although defending one of phenomenalism’s rivals, Grice was suggesting reasons
for retaining Austin’s particular bugbear, ‘sense data’, as a term of art. Indeed, he even goes so
far as to hint that his own, speculative version of the causal theory ‘neither obviously entails
nor obviously conflicts with Phenomenalism’. 31 Grice defends sense data because, as he
argues, the sceptic who wants to question the existence of the material world needs to make
use of them and the sceptic’s position at least deserves to be heard. He defends it also because
of his own ideas about language use. Indeed, his essay on perception seems at times in danger
of being hijacked by his interest in language. This is not entirely out of keeping with the subject;
through Locke and Berkeley to Ayer and Austin, the philosophy of perception had focused on
the correct interpretation of sentences in which judgements of perception are expressed.
However, Grice makes no secret of his hope that his interpretation of such sentences will fit
with views of language and use motivated by other considerations. Much of the essay is taken
up with a discussion of general principles governing the use of language. This is inspired by a
standard argument for retaining ‘sense data’ as a technical term; it would appear to underlie
ordinary expressions such as ‘so-and-so looks F [e.g. blue] to me’. In effect, a suitably rigorous
account of the meaning of such expressions would need to include reference to something such
as a sense datum that corresponds to the subjective experience described. A potential objector
might point out that such expressions are not appropriate to all instances of perception; they
would only in fact be used in contexts where a condition of either doubt or denial (D-or-D) as
to the applicability of F holds. ‘There would be something at least prima facie odd about my
saying “That looks red to me” (not as a joke) when I am confronted by a British pillar box in
normal daylight at a range of a few feet.’ 32 Grice considers two possible ways of explaining
this oddity. It could be seen as ‘a feature of the use, perhaps of the meaning’ of such expressions
that they carry the implication that the D-or-D condition is fulf illed. Use of such expressions
in the absence of the condition would therefore constitute a misuse, and would fail to be either
true or false. Alternatively, such statements could be seen as true whenever the property F
applies but, in the absence of the D-or-D condition, severely misleading. Use in the absence of
the condition would therefore constitute a more general error, because of ‘a general feature or
principle of the use of language’. 33 Grice describes how until recently he had been undecided
as to which of these two explanations to favour. He had, however, lent towards the second
because it ‘was more in line with the kind of thing I was inclined to say about other linguistic
phenomena which are in some degree comparable.’ 34 Towards the end of his discussion, Grice
hesitantly suggests a first attempt at formulating the general principle of language use: ‘One
should not make a weaker statement rather than a stronger one unless there is a good reason
for so doing.’ 35 This is, of course, remarkably similar to the observation attributed to him by
Strawson a decade earlier. It seems that during the intervening years Grice had become
increasingly impressed by how many different types of meaning this rule could explain, in
other words by how general and simple an account of language use it suggested. In two
interesting respects the formulation of the rule had changed. First, it had changed from a
statement about what ‘one does’ in language use to what ‘one should do’. This brought it nearer
to the first maxim of Quantity it was eventually to become. Second, the new formulation of the
rule introduces possible exceptions, by alluding to ‘good reasons’ for breaking it. Again, Grice
does not elaborate on this qualification, but the consideration of reasons for breaching the
maxims, together with a discussion of the consequences of doing so, was to be vital to the
subsequent development of his work. It was what was to give it explanatory and generalising
abilities beyond those of a simple list of ‘rules’ of linguistic behaviour. Grice’s claims about
the use of language offer support to the defender of sense data. They allow the defender to
describe statements of the ‘so-and-so looks F to me’ type as strictly true regardless of whether
theD-or-D condition holds. If such statements call for the use of sense data as a technical term
to explain their meaning, then sense data are applicable to any description of perception. It is
simply that in most non-controversial contexts such statements will be avoided because,
although perfectly true, they will introduce misleading suggestions. Here, perhaps, was Grice’s
greatest heresy. Austin’s rejection of sense data relied on an appeal to what people ordinarily
say, together with an assumption that this is the best guide to meaning and truth. Grice’s
tentative support for it relies on a distinction between what people can truthfully say, and what
they actually, realistically do say. Despite itsemphasis on linguistic use, ‘The causal theory of
perception’ was well received as a contribution to the philosophy of perception. It was
anthologised in 1965 in Robert Swartz’s Perceiving, Sensing and Knowing . In 1967, Grice’s
sometime collaborator Geoffrey Warnock included the essay in his volume on The Philosophy
of Perception , singling it out for praise as an ‘exceedingly ingenious and resourceful
contribution’. 36 As was the case throughout his working life, Grice made no firm distinction
between his teaching and the development of his own philosophical ideas. A series of lecture
notes from the 1960s on topics relating to speaker meaning and context therefore provide an
insight into the lines along which his thinking was developing, as do some rough notes from
the same period. What his students gained from the fresh ideas in Grice’s lectures, however,
they paid for in his cavalier attitude to the topic and aims of the class. He begins one lecture,
which is entitled simply ‘Saying: Week 1’: Although the official title of this class is ‘Saying’,
let me say at once that we are unlikely to reach any direct discussion of the notion of saying
for several weeks, and in the likely event of our failing to make any substantial inroads on the
title topic this term, my present intention is to continue the class into next term. 37 Grice’s
interest in these lectures was in finding out what can be learnt about speaker meaning,
specifically about what sets it apart from linguistic meaning, from close attention to its
characteristics and circumstances. The opening of another of the lectures, entitled ‘The general
theory of context’, tells rather more of his purpose and method than is made explicit in much
of the later, published work. Philosophers often say that context is very important. Let us take
this remark seriously. Surely, if we do, we shall want to consider this remark not merely in its
relation to this or that problem, i.e. in context, but also in itself, i.e. out of context. If we are to
take this seriously, we must be systematic, that is thorough and orderly. If we are to be orderly
we must start with what is relatively simple. Here , though not of course everywhere, to be
simple is to be as abstract as possible; by this I mean merely that we want, to begin with, to
have as few cards on the table as we can. Orderliness will then consist in seeing first what we
can do with the cards we have; and when we think that we have exhausted this investigation,
we put another card on the table, and see what that enables us to do. It is not hard to discern
Austin’s influence here, in the insistence that a particular philosophical commonplace, if it is
to be accepted as a useful explanation, must first be subject to a rigorous process of analysis.
The call for system and order, however, is Grice’s own. He had reacted against precisely the
tendency towards unordered, open-ended lists in Austin’s work. He argues in these lectures
that thinking seriously about context means thinking about conversation; this is the setting for
most examples of speaker meaning. He proposes, therefore, to compile an account of some of
the basic properties common to conversations generally. His method of limiting his hand was
to result in certain highly artificial simplifications, but he made these simplifications
deliberately and knowingly. For instance, the relevant context was to be assumed to be limited
to what he calls the ‘linguistic environment’: to the content of the conversation itself.
Conversation was assumed to take place between two people who alternate as speaker and
hearer, and to be concerned simply with the business of transferring information between them.
A number of the lectures include discussion of the types of behaviour people in general exhibit,
and therefore the types of expectations they might bring to a venture such as a conversation.
Grice suggests that people in general both exhibit and expect a certain degree of helpfulness
from others, usually on the understanding that such helpfulness does not get in the way of
particular goals, and does not involve undue effort. If two people, even complete strangers, are
going through a gate, the expectation is that the first one through will hold the gate open, or at
least leave it open, for the second. The expectation is such that to do otherwise without
particular reason would be interpreted as deliberately rude. The type of helpfulness exhibited
and expected in conversation is more specific because of a particular, although not a unique,
feature of conversation; it is a collaborative venture between the participants. At least in the
simplified version of conversation discussed in these lectures, there is a shared aim or purpose.
However, an account of the particular type of helpfulness expected in conversation must be
capable of extension to any collaborative activity. In his early notes on the subject, Grice
considers ‘cooperation’ as a label for the features he was seeking to describe. Does ‘helpfulness
in something we are doing together’, he wonders in a note, equate to ‘cooperation’? He seems
to have decided that it does; by thelater lectures in the series ‘the principle of conversational
helpfulness’ has been rebranded the expectation of ‘cooperation’. During the Oxford lectures
Grice develops his account of the precise nature of this cooperation. It can be seen as governed
by certain regularities, or principles, detailing expected behaviour. The term ‘maxim’ to
describe these regularities appears relatively late in the lectures. Grice’s initial choices of term
are ‘objectives’, or ‘desiderata’; he was interested in detailing the desirable forms of behaviour
for the purpose of achieving the joint goal of the conversation. Initially, Grice posits two such
desiderata: those relating to candour on the one hand, and clarity on the other. The desideratum
of candour contains his general principle of making the strongest possible statement and, as a
limiting factor on this, the suggestion that speakers should try not to mislead. The desideratum
of clarity concerns the manner of expression for any conversational contribution. It includes
the importance of expectations of relevance to understanding and also insists that the main
import of an utterance be clear and explicit. These two factors are constantly to be weighed
against two fundamental and sometimes competing demands. Contributions to a conversation
are aimed towards the agreed current purposes by the principle of Conversational Benevolence.
The principle of Conversational Self-Love ensures the assumption on the part of both
participants that neither will go to unnecessary trouble in framing their contribution. Grice
suggests that many philosophers are guilty of inexactness in their use of expressions such as
‘saying’, ‘meaning’ and ‘use’, applying them as if they were interchangeable, and in effect
confusing different ways in which a single utterance can convey information. For instance,
Grice refers back to the discussion at a previous class he had given jointly with David Pears,
when the exact meaning of the verb ‘to try’ was discussed. This, of course, was one of the
specific philosophical problems he was interested in accounting for by means of general
principles of use. Stuart Hampshire had apparently claimed that if someone, X, did something,
it is always possible to say that X tried to do it. This was challenged; in situations when there
is no obvious difficulty or risk of failure involved it is inappropriate to talk of someone’s trying
to do something. Grice’s answer had been that, while it is always true to say that X tried to do
something, this may sometimes be a misleading way of speaking. If X succeeded in performing
the act, it would be more informative and therefore more cooperative to say so. Therefore, an
utterance of ‘X tried to do it’ will imply, but not actually say, that X did not succeed. In his
consideration of the desiderata of conversational participation, Grice initially compiles a loose
assemblage of features. By the later lectures these appear in more or less their final form under
the categories Quantity, Quality, Relation and Manner (or, sometimes, Mode). Inarranging the
desiderata in this way, Grice was presumably seeking to impose a formal arrangement on a
diverse set of principles. But it seems that he had other motives: semi-seriously to echo the use
of categories in such orthodox philosophies as those of Aristotle and Kant, and more
importantly to draw on their ideas of natural, universal divisions of experience. The regularities
of conversational behaviour were intended to include aspects of human behaviour and
cognition beyond the purely linguistic. Grice’s collaborative work with Strawson had been
concerned with Aristotle’s division of experience into ‘categories’ of substances. Aristotle’s
original formulation of the list of such properties allows that they can take the form of ‘either
substance or quantity or qualification or a relative or where or when or being-in-a-position or
having or doing or being-affected’. 38 He concentrates mainly on the first four, and these
received most attention in subsequent developments of his work. They were the starting-point
for Kant’s use of categories to describe types of human experience, and his argument that these
form the basis of all possible human knowledge. In the Critique of Pure Reason , Kant proposes
to divide the pure concepts of understanding into four main divisions: ‘Following Aristotle we
will call these concepts categories, for our aim is basically identical with his although very
distinct from it in execution.’ 39 These are categories ‘Of Quantity’, ‘Of Quality’, ‘Of Relation’
and ‘Of Modality’, with various subdivisions ascribed to each. Kant’s claims for both the
fundamental and the exhaustive nature of these categories are explicit: This division is
systematically generated from a common principle, namely the faculty for judging (which is
the same as the faculty for thinking), and has not arisen rhapsodically from a haphazard search
for pure concepts, of the completeness of which one could never be certain. 40 Kant goes so
far as to suggest that his table of categories, containing all the basic concepts of understanding,
could provide the basis for any philosophical theory. These, therefore, offered Grice divisions
of experience with a sound pedigree and an established claim to be universals of human
cognition. Early in 1967, Grice travelled to Harvard to deliver that year’s William James
lectures, the prestigious philosophical series in which Austin had put forward his theory of
speech acts 12 years earlier. Grice’s entitled his lectures ‘Logic and conversation’. He was
presenting his current thinking about meaning to an audience beyond that of his students and
immediate colleagues and was clearly aware of the different assumptions and prejudices he
could expect in an American, as opposed to an Oxford, audience. ‘Some of you may regard
some of the examples of the manoeuvre which I am about to mention as being representative
of an out-dated style of philosophy’, he suggests in the introductory lecture, ‘I do not think that
one should be too quick to write off such a style.’ 41 Addressing philosophical concerns by
means of an attention to everyday language was still a highly respectable, even an orthodox
approach in Oxford. In America it was seen by at least some as belonging to an unsuccessful,
and now rather passé, school of thought. In pleading its cause, Grice argues that it still has
much to offer: in this case, the possibility of developing a theory to discriminate between
utterances that are inappropriate because false, and those that are inappropriate for some other
reason. Despite the difficulties inherent in such an ambitious scheme, and the well-known
problems with the school of thought in question, he does not give up hope altogether of
‘systematizing the linguistic phenomena of natural discourse’. Grice’s ultimate aim in the
lectures is ambitious and uncompromising; his interest ‘will lie in the generation of an outline
of a philosophical theory of language’. 42 He argues for a complex understanding of the
significance of any utterance in a particular context; its meaning is not a unitary phenomenon.
Conventional meaning has a necessary, but by no means a sufficient role to play. Indeed
conventional meaning is itself not a unitary phenomenon. Some aspects of it involve the
speaker in a commitment to the truth of a certain proposition; this is ‘what is said’ on any
particular occasion. Other aspects may be associated by convention with the words used, but
not be part of what the speaker is understood literally to have said. The examples ‘She was
poor but honest’ and ‘He is an Englishman; he is, therefore, brave’ convey more than just the
truth of the two conjuncts, more than would be conveyed by ‘She was poor and honest’ or ‘He
is an Englishman and he is brave’. An idea of contrast is introduced in the first example and
one of consequence in the second. These ideas are attached to the use of the individual words
‘but’ and ‘therefore’, but do not contribute to the truth-conditions of the sentences. We would
not want to say that the sentences were actually false if both conjuncts were true, but we did
not agree with the idea of contrast or of consequence. We might, rather, want to say that the
speaker was presenting true facts in a misleading way. These examples demonstrate implicated
elements associated with the conventional meaning of the words used, elements Grice labels
‘conventional implicatures’. There is another level at which speaker meaning can differ from
what is said, dependent on context or, for Grice, on conversation. In ‘conversational
implicatures’ meaning is conveyed not so much by what is said, but by the fact that it is said.
This is where the categories of conversational cooperation, and their various maxims, play their
part. The onus on participants in a conversation to cooperate towards their common goal, and
more particularly the expectation each participant has of cooperation from the other, ensures
that the understanding of an utterance often goes beyond what is said. Faced with an apparently
uncooperative utterance, or one apparently in breach of some maxim, a conversationalist will
if possible ‘rescue’ that utterance by interpreting it as an appropriate contribution. In this way,
Grice offers a more detailed account of the idea he explored in ‘Meaning’, and in his notes
from that time: that there are three ‘levels’ of meaning, or three different degrees to which a
speaker may be committed to a proposition. His model now includes, ‘what is said’,
‘conventional meaning’ (including conventional implicatures) and ‘what is conversationally
implicated’. The presentation of the norms of conversational behaviour in the William James
lectures is rather different from Grice’s handling of them in his earlier work. The maxims, or
the categories they fall into, are no longer presented as the primary forces at work. Instead, all
are assumed under a general ‘Principle of Cooperation’. The principle appeared late in the
development of Grice’s theory. It enjoins speakers to: ‘Make your conversational contribution
such as is required, at the stage at which it occurs, by the accepted purpose or direction of the
talk exchange in which you are engaged.’ 43 The name ‘Cooperative Principle’ was even later;
it was added using an omission mark in a manuscript copy of the second William James lecture.
Grice may well have been attempting to give a name, and an exact formulation, to his
previously rather nebulous idea of cooperation or ‘helpfulness’. However, the effect was to
change what was presented as a series of ‘desiderata’, features of conversational behaviour
participants might expect in their exchanges, to something looking like a powerful and general
injunction to correct social behaviour. In the development of his theory of conversation, Grice
was much exercised by the status of the categories as psychological concepts. He questioned
whether the maxims were the result of entering into a quasi-contract by engaging in
conversation, simply inductive generalisations over what people do in fact do in conversation,
or, as he suggested in one rough note, just ‘special cases of what a decent chap should do’. He
remained undecided on this matter throughout the development of the theory, content to
concentrate on the effects on meaning of the maxims, whatever their status. By the time of the
William James lectures, however, he seems to be closer to an answer. He is ‘enough of a
rationalist’ to want to find an explanation beyond mere empirical generalisation. 44 The
following suggestion results from this impetus: So I would like to be able to show that
observation of the Cooperative Principle and maxims is reasonable (rational) along the
following lines: that anyone who cares about the goals that are central to
conversation/communication (such as giving and receiving information, influencing and being
influenced by others) must be expected to have an interest, given suitable circumstances, in
participation in talk exchanges that will be profitable only on the assumption that they are
conducted in general accordance with the Cooperative Principle and the maxims. 45 This is a
wordy explanation, and also a troublesome one. It seems to create a loop linking the aim of
explaining cooperation to an account of conversation as dependent on cooperation, a loop from
which it does not successfully escape. The link between reasonableness and cooperation is far
from explicit. Nevertheless, this passage offers Grice’s account of his own preferences in
seeking an answer to the question over the status, and hence the motivation, for the Cooperative
Principle. His preference, particularly his reference to ‘rational’ behaviour, was to prove
important in the subsequent development of his work. However derived, the maxims operate
to produce conversational implicatures in a number of different ways. In many cases, they
simply ‘fill in’ the extra information needed to make a contribution fully cooperative. A says
‘Smith doesn’t seem to have a girlfriend these days’ and B replies, ‘He has been paying a lot
of visits to New York lately’. B’s remark does not, as it stands, appear relevant to the preceding
remark. But it is easy enough to supply the missing belief B must hold for the remark to be
relevant. B conversationally implicates that Smith has, or may have a girlfriend in New York.
46 In other cases the speaker seems to be far less cooperative, at least at the level of ‘what is
said’. In order to be rescued as cooperative contributions to the conversation, such examples
need to be not so much filled out as re-analysed. Because of the strength of the conviction that
the speaker will, other things being equal, provide cooperative contributions, the other
participant will put in the work necessary to reach such an interpretation. In perhaps his most
famous example of conversational implicature, Grice suggests the case of a letter of reference
for a candidate for a philosophy job that runs as follows: ‘Dear Sir, Mr X’s command of English
is excellent, and his attendance at tutorials has been regular. Yours, etc.’ The information given
is grossly inadequate; the writer appears to be seriously in breach of the first maxim of
Quantity, enjoining the utterer to give as much information as is appropriate. However, the
receiver of the letter is able to deduce that the writer, as the candidate’s tutor, must know more
than this about the candidate. There must be some reason why the writer is reluctant to offer
the extra information that would be helpful. The most obvious reason is that the writer does not
want explicitly to comment on Mr X’s philosophical ability, because it is not possible to do so
without writing something socially unpleasant. The writer is therefore taken conversationally
to implicate that Mr X is no good at philosophy; the letter is cooperative not at the level of what
is literally said, but at the level of what is implicated. In examples such as this a maxim is
deliberately and ostentatiously flouted in order to give rise to a conversational implicature;
such examples involve exploitation. These examples, and others Grice discusses in the second
William James lecture, are all specific to, and entirely dependent on, the individual contexts in
which they occur. Grice labels all such example ‘particularised conversational implicatures’.
There are other types of conversational implicature in which the context is less significant, or
at least can operate only as a ‘veto’ to implicatures that arise by default unless prevented. These
are implicatures associated with the use of particular words. Unlike conventional implicatures,
they can be cancelled: that is explicitly denied without contradiction. These ‘generalised
conversational implicatures’ account for many of the differences between the logical constants
and the behaviour of their natural language counterparts. In effect, Grice claims that there
simply is no difference between, say ‘ ~ ’, ‘ Ÿ ’, ‘ ⁄ ’ and ‘not’, ‘and’, ‘or’ at the level of what
is said. The well-known differences are generalised conversational implicatures often
associated with the use of these expressions, implicatures determined by the categories and
maxims he has established. Part of Grice’s motivation for this proposal was the desire for a
simplification of semantics. The alternative to such an account was to posit a semantic
ambiguity for a wide range of linguistic expressions. Grice argues against this, proposing a
principle he labels ‘Modified Occam’s Razor’, which would rule against it in decisions of a
theoretical nature. The principle states that ‘senses are not to be multiplied beyond necessity’.
47 Grice’s reference was to William of Occam, or Ockham, the fourteenth-century philosopher
credited with the dictum ‘entities are not to be multiplied beyond necessity’. This is known as
‘Occam’s razor’, although it is (Chapman, S. (2005). Paul Grice, Philosopher, and Linguist.
Palgrave Macmillan, New York.)