ON THE DERIVATION OF A CONVERSATIONAL MAXIM 
Th. R. Hofmann 
Shimane National University 
Matsue City, Shimane, Japan 690 
Summary. 
A conversational maxim of relation -- a principle 
of connectedness of text -- is derived from a 
more general & independently motivated principle 
governing the interpretation of sentences in con- 
text. This principle -- PIM, principe d'inter- 
pretation minimale -- that sentences are inter- 
preted in the most redundant way possible, is 
thus seen to be fundamental, & a general argument 
is given that the statement of, & the effective 
semantic procedures lie in the interpretive (to- 
ward semantic structure) & not in the generative 
(away from there) direction. 
Text. 
My purpose here is to show that a conversa- 
tional maxim --- believed to be a maxim for the 
speaker in well-formed conversation -- is in fact 
only a special case of a much more general rule. 
This more general rule, moreover, is a rule for 
the interpretation of sentences. This leads to 
questions about conversational maxims & the 
nature of the relation between syntactic form & 
semantic structure. 
We shall ist argue for a narrow & specific 
conversational maxim; a connectedness principle 
according to which texts & conversations should 
exhibit a quality of connectedness. It appears 
that this covers the ground of Grice's relation 
maxim: "be relevant". 
Grice i formulated several maxims for the 
speaker, which may be assumed of a speaker in 
usual circumstances. Together with the utterance 
of a sentence, they give rise to his 'conversa- 
tional implicatures'. They are grouped under 4 
headings: quantity, quality, relation, & manner. 
We are concerned here mainly with the 3 °, but 
shall have some comments later on the 1 ° 2. The 
4 ° group, "avoid obscurity", "avoid ambiguity", 
"be brief" & "be orderly", like the 3 ° "be rel- 
evant", are extremely vague & unformalizable 
stated as they are, as maxims for the speaker. 
To illustrate their operation with the i st 2 
groups, consider one saying "I think it is rain- 
ing" instead of "It is raining". Clearly this is 
less informative of the state of the world 
(though more informative about the speaker) than 
what could have been said, violating the quantity 
maxims in one direction or the other (be inform- 
ative enough, & only enough), & it also violates 
brevity. Hence the quality maxim must be domin.- 
ant here, "do not say that which you believe to 
be false or for which you lack adequate evidence." 
It follows that the speaker does not know or have 
adequate evidence that it is raining. 
We propose here as a conversational maxim, a 
connectedness principle. A simple statement is 
"stick to the subject" or "don't change the 
subject". Thus, "you were rather drunk last 
night .... Say, it's rather cold today, isn't it?" 
is an obvious violation of this principle. That 
some relation maxim is operating here, is beyond 
question. 
While the above statements of this maxim 
suggest that there is some sort of subject or 
topic that one might alternatively stick to or 
change, these topics or subjects turn out to be 
illusive, However, these injunctions are uttered 
when there is little overlap of semantic content 
with the preceeding context. We shall understand 
this connectedness principle in the following way; 
"there should be a high degree of overlap with 
preceeding context. 
It is true that many cases where these injun- 
ctions are uttered can be characterized by one 
person refusing to continue in a direction (topic) 
which is unpleasant to him. However, it is also 
common experience that a conversation between 2 
or more people may wander exceedingly. Quite 
often, in fact, it takes considerable effort by 
the participants to reconstruct what they discus- 
sed or how they arrived at i topic from a quite 
dissimilar topic 15 minutes earlier. And in such 
a reconstruction, one usually finds 3 or 4 other 
equally dissimilar topic-areas which were passed 
through. 
In such cases, one seldom even perceives a 
"change in topic", even though the topic does 
change in a rather radical way. On the other 
hand, when the topic of a conversation does not 
change, the conversation will often be judged as 
boring. We may conclude that what is avoided is 
not changes in topic, but rapid changes in topic; 
what is excluded by this maxim is a sentence with 
little or no overlap with the preceeding context. 
This lack of overlap will be found in all 
cases of objectionable topic change, but there 
are numerous cases of gradual topic change with- 
out problem. Moreover, with rapid change of 
topic, misunderstanding may arise, as we shall 
see below. 
It is however not true that the human mind, 
or human language, cannot accept rapid changes in 
topic. Rapid changes can & are often made, but 
must be accompanied by a marker of the change. 
There are numerous devices to indicate gross 
changes of topic. In spoken English, we have 
such things as "Says ..." or "Did you know that 
..." to mark a big jump. In more formal English, 
there is "Incidentally, ..." or "By the way, ..." 
which carry an implicit promise to return to the 
subject at hand after a short interruption, which 
is often not fulfilled in conversation. 
Temporary topic shifters like by the way 
both mark that overlap need not be found, 
236 
& also that the following segment will not serve 
as a context later on for overlap .... i.e. it is a 
segment probably slightly overlapping, but which 
can be forgotten later without loss. 
In more formal English, we find conLDlete 
shifters of topic,"to change the topic, ..." or 
"speaking of .., ...". These instruct the addres- 
see to seek no overlap, here or later. 
The existance of these markers shows that 
one can change the subject, but if so, that 
change should be formally marked. Thus we can 
reformulate this principle as "stick to the sub- 
ject unless you indicate otherwise" or "changes 
in subject should be overtly marked". This may 
be seen as a refinement of Grice's relation maxim 
"be relevant", but may also be seen as related 
also to his manner maxim "be orderly", in so far 
as it entails marking changes. 
However, given that conversations can & do 
wander, changing the subject gradually, without 
any annoyance or hinderance, or even awareness, 
to the participants, we must revise our notion of 
relevance to: relevance to the immediately pre- 
ceeding context, & the connectedness principle to 
"rapid changes in the subject must be marked". 
To say that there is a topic or subject 
throughout a conversation is too strong. Never- 
theless, sentences must bear some connectedness 
(semantic overlap) to their preceeding context. 
I propose that this is all that can & should be 
said in the area of Grice's relation maxim t "be 
re levant". 
Elsewhere 2 I have proposed a principle of 
semantic interpretation, PIM. This principe 
d'interpretation minimale disambiguates words, 
constructions, pronouns, &c, based on selecting 
the meaning which adds the least to the prior 
semantic context. Thus "flying planes can be 
dangerous" is not ambiguous in a context "don't 
fly airplanes. " or in a context "watch out for 
airplanes. ". In each case, there is only 
way to interpret this sentence with a minimal 
addition to the preceeding semantic context. 
Similarly, some word disambiguations can be ex- 
plained only by PIM. For example, the word light 
in the sentence "it is light" is not ambiguous in 
contexts like "how heavy is it? " or "is it 
still dark? " because the semantic components 
of only i sense will overlap sufficiently with 
the preceeding context in each case. 
This principle PIM may seem ontologically 
weak, as it is claimed to be true Only in cases 
of good communication -- i.e. cases of communic- 
ation where the idea is transmitted without am- 
biguity. PIM can be easily falsified by construc- 
ting a text where the reader is left in doubt as 
to what is said. This appears circular at first 
glance. It is not, however. 
Good communication (or well-written text) is 
empirically identifiable. In such cases, PIM is 
met. Even more easily identifiable is poorly 
written text, poor because it leaves ambiguities 
in the reader's mind. In most if not all such 
cases, it appears, PIM is violated. Such cases 
can be rendered non-ambiguous by the addition of 
enough words to disallow all but I way of integ- 
rating (overlapping) it into its preceeding con- 
text with a minimum of new information. 
Lastly, as it states a condition necessary 
for good (or even adequate) communication, it 
makes strong predictions about what speakers or 
writers will use (pronoun, noun+adjective+defin- 
ite, or just noun+definite), given a prior con- 
text & an idea to express. These predictions are 
easily tested, & it is thus empirical. 
Assuming then that this principle PIM does 
exist, it is easy to see why large changes in the 
topic must be marked. If a jump of topic is not 
marked, the initial sentence in the new topic 
will be interpreted in such a way as to add the 
least possible to the preceeding. This will gen- 
erally lead to misinterpreting it, if there is an 
overlap possible. Or, if the addressee cannot 
find overlap, he must draw upon scenarios 3 & 
belief-structure~that the speaker might be 
assuming, failing which he perceives a "lost" 
feeling, as when a speaker appears to say some- 
thing completely incoherent, "it is raining. 
moreover, elephants are big." 
The markers of topic change simply indicate 
to the addressee that he should not use PIM in 
interpreting what follows. They are necessary 
whenever the following material might be open to 
a different interpretation when PIM is applied. 
The existance of PIM requires that topic change 
markers exist & be used in this case. 
Moreover, as one often does not know in ad- 
vance exactly what words or constructions he will 
use, topic change markers are a useful heuristic 
device to avoid potential misinterpretations, 
when making a zapid topic change. Indeed, when 
such a marker is approprite, but missing, the 
addressee may (& will) search for some interpre- 
tation of the following sentence which will over- 
lap to at least some degree. In such a case, it 
is a kindness to the addressee (& part of the 
notion of good communication) to indicate that 
such a search is unnecessary. 
The existance & use of topic changing mark- 
ers is thus required by PIM. In fact, the whole 
prohibition of rapid topic change (unless marked) 
is simply a reflection of PIM. With no marker, 
the listener will assume, by PIM, that each sen- 
tence overlaps seriously, in a maximal way, with 
its preceeding context. Hence, a speaker -- who 
presumably talks to be understood by someone -- 
must construct each sentence so as to overlap. 
This is precisely what we have seen to be the 
interpretation of the connectedness principle or 
relation maxim. 
Thus this relation maxim can be derived from 
PIM. However, PIM cannot be derived from this 
maxim. PIM is much more general, disambiguating 
even pronouns based on collocated verbs, for ex- 
ample. In the following example, .she is under~ 
stood as neither the closest female antecedent, 
nor the subject of the preceeding sentence, 
237 
in the absence of other context. "Mary told 
Susan that Betty was coming. Unfortunately, she 
misunderstood it." Rather, she is understood, in 
accordance with PIM, as referring to whatever 
allows the greatest overlap with the preceeding 
context, here to misunderstand overlaps in mean- 
ing with to tell (which is something like 'to 
cause to understand, by speaking'). In such an 
example, a change in the verb in the 2 ° sentence 
causes a change in the interpretation; if it were 
come or arrive, she would refer instead to Betty, 
while if the 2 ° sentence were "she said it soft- 
ly", she could only refer to Mary. 
Thus the relevance maxim can be seen as a 
reflection for the speaker of a principle for the 
interpretation (disambiguation) of sentences. It 
is perhaps for this reason that the relevance 
maxim has been resistant to formalization. 
Happily however, the interpretation principle PIM 
is formalizable. 
Because the connectedness principle can be 
derived from PIM, but PIM cannot be derived from 
it, & because PIM is formalizable, but not appar- 
entl~an~elevance maxim, we may conclude that it 
is so derived, & that it either has no independ- 
ent existance, or possibly has existance only as 
a heuristic for the speaker. 
Moreover, it does not seem possible to state 
PIM for the speaker. In the pronoun case above r 
stated for the speaker, it could be stated only 
as a rule for when a pronoun cQuld be used in 
place of a more extensive description or name. 
Even when so stated, it is still an interpretive 
rule; a pronoun can be used if no ambiguity will 
result, taking ambiguity to be measured relative 
to the verbs or other things collocated with the 
prospective pronoun. A more directed procedure 
is probably the case: try integrating the pro- 
jected sentence (with pronouns) into the precee- 
ding context's semantic structure. If that 
results in an undesired meaning, then find or 
construct a more complete description. When this 
is executed in real time, it may result in the 
post-sentence nominal (extra-position or right 
displacement) which is not uncommon in spoken 
language, "he came yesterday, the tall boy, that 
is". 
Although we have concluded that this i maxim 
is derived from an interpretive principle, not 
all maxims are interpretive rules. One of 
Grice's quantity maxims, "do not make your con- 
tribution more informative than is required" is 
probably better formulated as MCCawley's least- 
effort principle. 5 What is minimized is not the 
transfer of unrequired information, but something 
like the number of lexical items or perhaps the 
number of syllables. Speakers are generally 
lazy. Either formulation, however, is clearly 
describing speaker behaviour, though to call it 
a maxim seems a bit inappropriate if MCCawley's 
version is correct. 
I note however that as a principle describ- 
ing speaker behaviour, this 'maxim' is available 
for use by a listener to discover more infer- 
mation than is overtly encoded. By studying 
what a person could have done (or said) but did 
not, one can make more or less sure conjectures 
as to the cause(s) of avoidance. This maxim 
seems to be no more than that. It is seldom 
used by a speaker to transmit more information 
than is encoded. 
The other quantity maxim, "make your con- 
tribution as informative as is required", is in 
many cases more a maxim for proper behaviour in 
the social interaction, question-answer. When 
one executes the question performance, the 
addressee is obliged to give an answer, by the 
rules of the social pact. "Where are you going? 
... Out" violates this maxim as stated, but is 
adequate to satisfy the social ritual initiated 
by the question. This maxim, then, may be deriv- 
ative from the meanings of performative elements, 
like question, command, &c. They are only lin- 
guistic expressions of socio-linguistic rituals; 
it is these rituals which define the proper be- 
haviour of the participants, in precisely the 
same way that performative expressions like "I 
promise ..." or "I pronounce you man & wife" do. 
Grice's quality maxims also, "de not say 
what you believe to be false" & "do not say that 
for which you lack adequate evidence", can be 
derived from the 'meaning' of the declarative or 
assertion performative. As this is something 
like "I warrentee or sign my name to the fact 
that this sentence (the proposition) is true", & 
the speaker opens himself to accepting public 
criticism if it is wrong, it is clear why most 
speakers (whom we assume would avoid criticism) 
will behave according to these quality maxims. 
Thus it appears that most of Grice's conver- 
sational maxims derive from other things, either 
principles of interpretation or the social ritu- 
als expressed in performatives. The others, 
those of manner, are not only vague, but I sus- 
pect, also derive from other things. But that 
is the subject of an other study. Our purpose 
here is the destruction of I maxim, & the reper- 
cussions that follow from this. 
I believe that I have shown that at least I 
conversational maxim is a special case of a gen- 
eral rule for interpretation. Some other maxims 
are derivable from other sources. This inter- 
pretive rule throws considerable doubt on the 
possibility of a completely generative component 
to connect semantic structure & syntactic form. 
It is, then, a challenge to proponents of gener- 
ative semantics or other non-interpretive theor:- 
ies of semantics. 
Indeed, the proper statement of most if not 
all semantic processes appears to lie in the in- 
terpretive direction. The reverse direction, 
creating syntactic strings out of semantic 
structures, is a highly non-deterministic pro-- 
cess, or rather art, often done by trial & error, 
even by master writers. Its fine execution, 
high rhetoric, is one of the highest human cap- 
acities, & is seldom attained except by 5% of 
the population. These well-known facts alone 
238 
should warn us that formalization of the gener- 
ative direction may be very difficult. The re o- 
verse direction, which all human beings can do 
easily, quickly, & with seldom any hesitation or 
backtracking (within the limits of the person's 
vocabulary), is at least deterministic for sam- 
ples of good communication. It is thus much 
more liable to formalization. 
Thus I hope to have shown that one more as- 
pect of semantic processing allows & demands 
formal representation only in the interpretive 
direction. The game of speaking appears to be 
to compose sentences that will be interpreted in 
the way desired. For this task, heuristics, but 
no fail-safe procedures can be given. A sentence 
may be composed, tested for its meaning in the 
context, & revised until it bears the desired 
additional meaning. The game of understanding 
is much simpler; decode lexical items into their 
meaning components, & the syntactic structures 
as well into meaning relations, & integrate that 
resultant structure into the semantic structure 
of the preceeding context by overlapping it as 
much as possible. Ambiguities met in this pro- 
cess should be resolved by this maximal over- 
lapping process, which is the principe d'inter- 
pretation minimale. There are of course many 
constraints about definite marking, topic & 
focus marking, scenario-searching, &c. 
As a challenge to proponents of semantic 
interpretation without a specific semantic rep- 
resentation, let me note that any & every state- 
ment of this procedure requires the use of a 
notion of "prior semantic context". This is, I 
believe, nothing more or less than a semantic 
representation. 
Lastly, I would propose the following 
speculation based on observation of how we have 
used the notion of topic. In the early part of 
this discussion, we used a notion of topic or 
subject of a conversation as if we knew what it 
was. In my limited experience, however, I have 
not found it easy to identify quite what the 
topic is, for most paragraphs or discussions, 
except in a vague way. Yet it is a notion much 
used in prescriptive studies of rhetoric (re~ 
ducing semantic content to expression). Could 
it be that this notion is no more than: that 
which ought not be changed without warning ? 
That is, I question whether this notion of 
topic -- roughly, what one is talking about -- 
is anything more than a speaker-oriented reflec- 
tion of the need for semantic overlap. If so, 
then it is not much more than the sentential 
topic in a topic-comment syntactic structure, 
which is the portion marked to overlap obliga- 
torily. 2 This notion of topic of discourse need 
have little more existance than the successive 
overlaps found between pairs of successive sen- 
tences. 
Model writing usually does of course have 
an identifiable topic at any given point. And 
to be sure, there are rhetorical figures based 
on a recurrent topic. But such figures are not 
obligatory at all. For ordinary writing or con- 
versation, I find it difficult to say more than, 
"this paragraph is about X, & Y, & Z, ..." 
It is thus that I find the notion of topic or 
subject of a text suspicious, that it is perhaps 
an imaginary concept invented for teaching rhet- 
oric. Indeed, we have found no use here for it, 
except in the prohibition of rapid changes in 
topic, which was really a requirement for some 
overlap between sentences. 

References 

i. Grice, H.P. 'Logic & conversation' (1975) in 
Cole & Morgan (Ed) Syntax & Semantics 3: 
S~eech Acts, Academic Press (N.Y.). & (1967) 
part of William James Lectures. 

2. Hofmann, Th.R. Description s~mantique & dy- 
namique du discours (1978)th~se pour la 
Sorbonne (Paris)j & G.E.T.A (Grenoble). 

3. --- 'What some semantic theories can't do' 
(1979) C.L.S. Parasession on the elements, 
Univ Chicago (Chicago). 

4. --- 'How to ignore pragmatics & get away with 
it, almost' (1980) Canadian Semiotics Assn 
