Some Psycholinguistic Constraints on the Construction and 
Interpretation of Definite Descriptions I 
Andrew Ortony 
Center for the Study of Reading 
University of Illinois at Urbana-Champaign 
Introduction 
It is a curious and important fact about 
natural languages that they can be and often are 
used nonliterally. Whereas in artificial lan- 
guages it is normally impossible to distinguish 
between the meaning of an expression on the one 
hand, and the intended meaning the user wishes to 
convey with that expression on the other (the 
two are identical), in natural languages a par- 
allel distinction between sentence meaning and 
speaker meaning (see, for example, Searle, 1975, 
in press) underlies all nonliteral uses. One of 
the central concepts in the analysis of non- 
literal uses of language is that of indirectness. 
It is a notion that has begun to attract the 
attention of linguists, philosophers, psycholo- 
gists, and computer scientists in their various 
efforts to come to a better understanding of 
natural languages and of human linguistic per- 
formance. One of my purposes in this paper is 
to show how central a concept indirectness is 
with respect to the production and comprehension 
of definite descriptions. 
The main problem with which I am concerned 
is a multi-level one. At the most general level 
it concerns the way in which people determine the 
referents of definite descriptions, and how lan- 
guage users choose the definite descriptions they 
do. More specifically, I am concerned with the 
question of the constraints that exist upon how 
a thing can be referred to. What makes this an 
interesting problem is the fact that it seems not 
to be necessary for a referring expression to be 
based on either information that has already been 
made explicit in the preceding discourse, nor 
even on information that is entailed by what has. 
Yet clearly, there are constraints on the ex- 
pressions that can be used if there is to be a 
realistic hope of communicative success. 
The question of what is to count as a rea- 
sonable way of referring to something in part 
depends for its answer on what counts as a rea- 
sonable indirect use of language. When, for 
example, one refers to the 1977/78 Seattle 
basketball team as The Cinderella of the NBA one 
is using a definite description based on a predi- 
cate that is not literally true of the intended 
referent but that is metaphorically applicable 
to it. As one thinks about the processes that 
might be involved in the production or comprehen- 
sion of such an expression they appear to be very 
complex, yet however complex they may be, people 
usually engage in them without any apparent 
difficulty. At present there appears to exist 
no adequate theoretical account of what these 
processes are like, perhaps because a comprehen- 
sive treatment of definite descriptions has as a 
prerequisite a theory of indirectness, and that 
in turn seems to hinge on a more comprehensive 
theory of speech acts than is currently available. 
My own proposals are not intended to fill all 
these gaps, but they are intended to sketch a 
possible direction for doing so. The main goal 
that I have is to suggest a way of imposing 
limits on indirectness, and then, to show how 
those same limits are needed to account for some 
important constraints on successful definite 
descriptions. 
Definite Descriptions and their Textual Relations 
I shall take it as axiomatic that every 
definite description is based upon a predicate 
that is supposed to be uniquely applicable (at 
least within the context of the discourse) to 
some entity relevant to the discourse. Thus, 
the definite description The first man on the moon 
is based on the predicate is/was the first man on 
the moon, and it is assumed to be applicable to 
some entity (e.g. Nell Armstrong) relevant to the 
discourse. It is important to note at the outset 
that coreferential expressions cannot always be 
substituted for one another without a change of 
meaning. For instance, if, on arriving in a 
strange unfamiliar hotel in a foreign land one 
were to utter (la), it hardly makes sense to say 
that it is equivalent to uttering (Ib). 
(la) I feel like the first man on the moon, 
(Ib) I feel like Nell Armstrong. 
However, if the speaker can safely assume that 
his audience knows that Nell Armstrong was the 
first man on the moon, (lb) could be used as an 
indirect way of achieving the communicative 
intent of (la). Notice, it is not necessary to 
know who the first man on the moon was in order 
to fully understand (la), whereas it is neces- 
sary to know that Nell Armstrong was the First 
man on the moon in order to properly understand 
(#b) (although one might not understand why the 
speaker used (Ib) with its unnecessary demands 
'on additional knowledge and inferences in 
preference to (la).) In any event, it seems 
that even the relationship between definite 
73 
descriptions and proper names may sometimes depend 
on a notion of indirectness (see 0rtony & Anderson, 
1977). 
The interpretation of definite descriptions 
often relies heavily on the establishment of in- 
ferential relationships of various kinds in order 
to determine which entity is being referred to. 
Such inferences tend to be forced jointly by the 
desire of the hearer or reader (hereafter referred 
to simply as "the hearer") to make sense of the 
discourse, and the assumption that the speaker or 
writer (hereafter, simply, "the speaker") is com- 
municating in accordance with the cooperative 
principle (see Grice, 1975). This latter assump- 
tion is critically important in cases where the 
predicate underlying the definite description is 
not obviously true of the intended referent--and 
since these cases appear, at least on the surface, 
to constitute the most difficult ones, I shall 
concentrate on their analysis, to some extent at 
the expense of simpler examples. I shall call 
such cases "definite descriptions of inference." 
The overriding logic of the determination of the 
referents of such definite descriptions is that 
if the speaker is communicating in accordance 
with the cooperative principle certain assump- 
tions have to be recognized in order for the 
expression in question to successfully identify 
the intended referent. These assumptions often 
serve to "sneak in" new information about the 
referent (in much the same way as appositive 
relative clauses introduce new information). The 
identification of these assumptions is based on 
inferences of various kinds. 
Definite descriptions of inference can be 
contrasted with definite descriptions based on 
entailment relations. Definite descriptions 
based on entailment are those for which the 
referent can be determined either by transforming 
a predicate that has already appeared in the dis- 
course into a definite description (descriptions 
based on the principle of identity), or by re- 
lating the description to predicates that have 
appeared earlier, on the basis of rules of formal 
logic (e.g. modus ponens) applied to them. 2 The 
important difference between a definite descrip- 
tion of entailment and a definite description of 
inference is that the interpretation of the former 
does not depend on the provision Of suppressed 
premises drawn from the comprehender's general 
world knowledge. In a definite description of 
inference it does. An example of a definite 
description of entailment can be found in (2), 
where the underlined expression is entailed by 
the content. 
(2) A well-dressed man entered the room 
and greeted the hostess. Although 
eyeryone else was drinking sherry, 
he asked the waiter for a scotch. 
The waiter gave him one. The man 
with the scotch walked over to his 
host. 
In this case, if the waiteP gave the man a 
scotch, it entails that the man had the scotch, 3 
and so, within the constraints imposed by the 
context, he can be uniquety identified by the 
definite description the man with the scotch. 
With definite descriptions of inference, as 
with communication in general, success often 
depends on the speaker and the hearer sharing a 
common background of knowledge (see, for example, 
Stalnaker, 1974). Definite descriptions of in- 
ference are more complex. For example, suppose 
that in (2) the sentence The waiter gave him one 
is omitted. Then, the definite noun phrase 
The man with the scotch only succeeds in re- 
ferring to the right man if it is assumed that 
the man who asked for a scotch was given one. 
Unfortunately, only in biblical circles is it 
true that asking for something guarantees being 
given that thing. So, in order for the hearer 
to identify the intended referent he has to 
assume that the man got his scotch. Of course, 
this assumption comes easily for it can be made 
on the basis of a plausible inference requiring 
only the introduction of plausible suppressed 
premises, such as that when a guest asks a 
waiter for a particular kind of drink at a cock- 
tail party, the waiter normally obliges if that 
drink is available. This constitutes a simple 
example of a definite description of inference. 
On encountering a definite description, the 
hearer has to assume that the description does 
indeed refer to some already mentioned person or 
thing, in this case, say, the guest. In doing so, 
he makes inferences that fill in what went 
before--that is he makes inferences about what 
might have been asserted to enable the predicate 
underlying the description to be both applicable 
and relevant. The comprehender might reason as 
follows in the present example: "If this ex- 
pression refers to the guest, then it must be 
the case that the waiter gave him a scotch. This 
is quite plausible since it is customary for 
waiters at cocktail parties to give guests the 
drinks they request if those drinks are available. 
It is plausible that scotch was available, since 
it is a frequently served drink at such occasions. 
So I shall assume that this is what happened and 
that is why the guest was referred to as 'the man 
with the scotch'." Whether or not people normally 
construct such chains of reasoning in order to 
identify the antecedents of definite descriptions 
is not the issue here. What is the issue, as we 
shall see later, is that it be possible to con- 
struct such a chain. Certainly, one has to suppose 
that the kind of general world knowledge required 
to do so is normally available during the compre- 
hension process. The relevant frames, scripts, 
schemata, or whatever other knowledge structures 
are supposed, are presumably activated. 
Definite descriptions of inference involving 
indirectness, like (3) and (4) below, tend to be 
more complex. They are characterized by the fact 
that the applicability of the predicates under- 
lying them often depends on the utilization of 
knowledge that Morgan (1978) calls knowledge 
about the language, as opposed to knowledge of 
the language. These turn out to be cases of in- 
ferences involving knowledge about illocutionary 
forces and perlocutionary effects (see Austin, 
1964). 
(3) The hostess offered the guest some 
cake. He told her that he was on a 
diet. His brother, who was with him, 
told her that he personally was not 
74 
on a diet. The man who had refused 
the cake walked over to his host. 
(4) The hostess asked the man where his 
wife was. He replied "Mind your 
own business, you old bag. ~' The 
hostess was furious that the man 
who had insulted her had been 
invited to her party. 
The interpretation of (3) requires not only 
semantic and general world knowledge in the way 
that (2) does, it also requires the knowledge 
that saying that one is on a diet can count as 
rejecting an offer to eat something. In the 
case of (4) it requires the knowledgethat the 
violation of certain language-use conventions 
can count as offensive behavior. Of course, in 
a sense, this kind of knowledge about the con- 
ventions of language use and the social/ 
communicative consequences of their violation 
is knowledge of the world, just as knowing that 
waiters normally serve the drinks they are asked 
to is. But, insofar as it is knowledge of con- 
ventions about the use of language, and insofar 
as this is an area which has been singled out as 
being of core concern in pragmatics, it is worth 
separating such cases from the other kinds of 
cases, like (2). In fact, I think, the distinc- 
tion is difficult to uphold because the mechanism 
required to deal with indirectness is the same 
kind of inferential mechanism as is required to 
deal with "ordinary" knowledge of the world. 
The question that eventually has to be 
answered concerns the constraints that there are 
on the predicates employed in definite descrip- 
tions. My view is that the answer to this 
question depends on finding an answer to a more 
general question about the pragmatics of lan- 
guage, namely the question: what constraints 
are there on what is relevant (in the sense used 
by Grice, 1975 and others). Staying, for the 
moment, with definite descriptions, compare (5) 
and (6) below: 
(5) The hostess offered the guest some 
cake. He told her that he was on a 
diet. His brother, who was with him, 
told her that he personally was not 
on a diet. The man who thought h e 
ought not to eat fattening things 
walked over to his host. 
(6) The hostess offered the guest some 
cake. He told her that he was on a 
diet. His brother, who was with him, 
told her that he personally was not 
on a diet. The man who was not hungry 
walked over to his host. 
It seems to me that whereas (5) is perfectly 
coherent, (6) is not. It becomes coherent, how- 
ever, if the context is changed so that instead 
of (he) told her that he personally was on a diet 
it reads (he) told her that he personally had 
just eaten, then both (3) and (6) seem perfectly 
acceptable. It seems, then, that the appropri- 
ateness of the definite description depends on 
the appropriateness of its underlying predicates. 
Telling someone that one has just eaten is an 
appropriate, relevant, piece of information for 
permitting the inference that one is not hungry, 
and/or that one does not want the offered food. 
By contrast, telling someone that one is on a 
diet is an appropriate piece of information for 
permitting the inference that one does not want 
what is being offered, but it is not appropriate 
for the inference that one is not hungry. 
Indirect Speech Acts 
A major part of my thesis is that the predi - 
cate underlying a definite description of in- 
ference is constrained by the relevance relation 
in just the same way as that relation constrains 
what counts as an indirect speech act in a dis- 
course. If this is so, then it will help to have 
a working hypothesis about the constraints that 
exist on indirect speech acts. 
Suppose the situation is that described in 
(3), namely, one in which someone is offered 
some cake and in uttering (7) intends to refuse 
the cake. 
(7) I am on a diet. 
The question we have to answer is this. Since it 
does not follow logically from (7) that the in- 
tention was to refuse the cake, on what basis 
does a listener come to the conclusion that indeed 
that was the intention? Furthermore, why does, 
for example, (8) not succeed in communicating 
the refusal? 
(8) My mother is an opera singer. 
Perhaps one should reject (8) on some very general 
grounds. For example, on the grounds that one 
cannot randomly assign a sentence to an intention 
and expect to be understood. But the same old 
question arises about what constitutes a random 
versus a non-random assignment as arises about 
appropriateness and relevance. 
The solution I propose is based on the notion 
of a "plausible chain of reasoning." It is this. 
For an indirect speech act to be understood as 
being relevant, or appropriate, it must be able to 
participate as a premise, or as a sub-conclusion, 
in a quasi-logical, or better, psycho-logical, 
chain of reasoning that plausibly relates the 
event that initiates it to its intended illocu- 
tionary force. To see the full implications of 
this proposal, let us see how it works with the 
example. The man is offered some cake, and this 
offer is the event that initiates his response. 
From the perspective of the man, (9) is true. 
(9) I am being offered some cake. 
Such an event calls for one of two responses, an 
acceptance or a refusal, appropriately modified 
by politeness conventions. Let us assume that 
"Yes, please" and "No, thank you" count as direct, 
literal speech acts for accepting and refusing, 
respectively. They certainly are conventionally 
regarded as direct ways of accepting and refusing. 
Now we can see that in this particular case, the 
proposal is this: for (7) to be understood as a 
refusal, it must be able to participate as a 
premise, or as a subconclusion, in a psycho- 
logical ~chain of reasoning that plausibly relates 
7.5 
the original offer to its acceptance or rejection. 
Such a chain of reasoning might look something 
like (9) - (15). 
(9) I am being offered some cake 
I0) I am on a diet 
ll) People on diets ought not to eat 
fattening things 
12) Cake is fattening 
13) (It follows logically that) 
I ought not to eat any cake 
14) (It follows deontically that) 
I will not eat any cake 
(15) (It follows conventionally that) 
I will refuse the cake 
This chain of reasoning, including the intermediate 
and final conclusions does not constitute a deduc- 
tively valid argument in the usual logical sense. 
The relationships that exist between (13) and 
(14), and between (14) and (15) are not entailment 
relations, but they are characteristic of human 
reasoning. 
A number of important observations have to be 
made about the chain of reasoning--observations 
that amount to constraints on what it normally is. 
First, there are no unnecessary premises in it. 
Every premise is needed for the establishment of 
the first subconclusion, (13), which in its turn 
is needed for establishing the final conclusion. 
Second, although the order of the premises that 
are introduced from the speaker's general know- 
ledge can be manipulated, the most natural order 
is one in which each premise invokes a concept 
that has been foregrounded (in the sense of Chafe, 
1972) by the preceding one. If this were not the 
case, the possibility of introducing irrelevant 
premises would arise--a possibility that could 
serve no useful purpose in the present context. 
In fact, this constraint probably needs to be a 
little more liberal than I have described, but for 
the reasons I have indicated, something close to 
it needs to operate. Third, the conclusion of the 
chain contains the information appropriate for a 
direct response to the initiating event, an event 
that need not itself be a linguistic one (as it 
is in the present example). The initiating event 
might be an observed event to which an appropriate 
response might be a description of it, or of a 
reaction to it. Consequently, in the general 
case, "response" should not be taken to mean 
"reply." 
There are doubtless other constraints that 
a more detailed analysis would reveal, but for the 
moment I want only to suggest that the conjunction 
of these (or some comparable set of) constraints 
constitutes what I mean by "plausibility" in the 
context of my requirement that the chain of 
reasoning be a psycho-logically plausible one. 
We are now in a position to consider what 
happens from the perspective of the hearer. The 
most important thing is that the hearer assumes 
that the speaker is constrained in what he says in 
just the kind of way that I have indicated. The 
hearer, therefore, attributes to the speaker some 
plausible chain of reasoning. However, the 
hearer may not have all the knowledge that is 
available to the speaker -(he may not know that he 
is on a diet, for example). Consequently, he 
may have to make inferences of his own in order 
to reach some of the premises required. This 
would be true if, for example, the response to 
the offer of cake had been (12) rather than (I0). 
Sometimes these inferences are incorrect and one 
understands correctly what was intended, but for 
the wrong reasons, or one misunderstands it al- 
together. As we shall see, this fact, that the 
heater's interpretation is only probabilistically 
determined, has some important consequences for 
the speaker's selection of his utterance. 
The most crucial claim that I wish to make 
about the chain of reasoning is this. Assuming 
that the speaker does not choose to express him- 
self directly (for whatever reason), then within 
the limits of the context, any of the premises 
or subconclusions in the chain from the initiating 
event to the (direct) conclusion can function as 
more or less easily interpretable surrogates for 
the conclusion--any of the steps can constitute 
an indirect speech act appropriate to the direct 
speech act that constitutes the conclusion. Thus, 
any of (10), (ll), (12), (13), and (14) can serve 
as indirect response to the offer. And, if some 
other response is made, it must be able to serve 
as a step in a similar chain of plausible 
reasoning. If it cannot, it is an inappropri- 
ate response. It is precisely these constraints 
that prevent (8) from being a possible indirect 
response to the offer, since there is no basis 
of shared knowledge that will normally permit a 
hearer to reconstruct an argument in which (8) 
figures to be relevant on the chain from initi- 
ating event to conclusion. 
An important question that now needs to be 
answered is why do people use language indirectly 
in the first place, and why, given that they can 
choose from a restricted range of indirect 
communicative acts, do they select the ones they 
do. Why, for example, would a speaker choose 
(10) instead of, say (12)? The answer to the 
first part of the question depends on exactly 
what kind of indirect language act is being used. 
For example, metaphors may be used for purposes 
of communicative economy, communicative vivid- 
ness, or even communicative possibility (see 
Ortony, 1975). With indirect speech acts, the 
answer is very often that the speaker gets 
"two for the price of one." For example, he can, 
with one utterance, not only refuse the offer, but 
also satisfy certain social conventions by pro- 
viding a good reason for his refusal, or at least 
hinting at one. As Searle (1975) points out, in 
an indirect speech act the speaker intends both 
the sentence meaning and the speaker meaning to 
be recognized by the hearer. So, indirectness 
affords economy as well as, often, politeness 
and sensitivity. 
There remains the question of why a speaker 
should select one form over another. The answer 
again lies in the fact that the communication of 
the literal meaning of the indirect language act 
is intended. Some of the knowledge that is 
needed to construct the reasoning chain may be 
more publicly available than other knowledge 
required. Thus, most people know that people 
on diets ought not to eat fattening things 
76 
(depending on the purpose of the diet, of course). 
Consequently it can be assumed that a hearer has 
more ready access to that fact than to the fact 
that the speaker is on a diet (which possibly very 
few people know). Thus, the speaker's selection 
of the particular language act can teke advantage 
of his beliefs about what the hearer is likely to 
know. It can also take advantage of the fact 
that some of the choices seem to have a stronger 
force than others. This is a complex issue. My 
intuitions are that (13) leaves open the possibil- 
ity of ultimately accepting some cake rather more 
readily than does (lO), perhaps because once (lO) 
is used it must be relevant to the chain of rea- 
soning, whereas if (13) is used, it could be used 
to reach a different conclusion. After all, most 
people occasionally do things that they ought not 
to do, and that possibility seems wide open if 
the response to the offer is (13). This is not 
the place to explore these issues further, but it 
is worth noting that many jokes capitalize on 
expectations of plausible reasoning chains of the 
kind I have been discussing--the trick is to make 
them go awry! 
My proposal shares certain characteristics 
with that of Searle (1975) in that it suggests a 
not necessarily conscious chain of reasoning. It 
differs from Searle's account insofar as it makes 
claims about the constraints on what can be said 
and understood. Searle's chain of reasoning con- 
tains many metalinguistic premises about in- 
directness that I have treated as background 
assumptions. My focus, by contrast, is on the 
content of the chain. What I have proposed is a 
possible answer to the question "How indirect can 
an indirect speech act be?" I have suggested 
that it cannot be so indirect that it could not 
participate in a chain of plausible reasoning 
relating a representation of the initiating event 
to an appropriate direct response to that event. 
I have also suggested that the illocutionary 
effect of all steps within such a chain will be 
appropriate for that initiating event. 
Definite Descriptions of Inference 
When speakers and writers produce, and 
hearers and readers comprehend definite descrip- 
tions, they do so against a background of know- 
ledge that includes their tacit knowledge about 
indirectness. This knowledge is often brought 
to bear in dealing with definite descriptions 
of inference--descriptions, that is, in which 
the underlying predicate could appear on the 
reasoning chain and that could constitute a 
direct or indirect speech act. Thus, for example, 
(3) and (4) are cases in which the underlying 
predicate could constitute the conclusion of a 
chain of reasoning--i.e, a direct language act, 
while (5) is a case in which the underlying 
predicate could constitute a premise in a plau- 
sible chain of reasoning--i.e, an indirect lan- 
guage act. 
In a sense, what I have proposed is a partial 
account of relevance in Grice's (1975) sense, or 
perhaps better yet, a partial account of when 
apparent violations of relevance are indeed only 
apparent, and why. It is quite clear that the 
predicates underlying definite descriptions have 
to be relevant to the discourse just as any other 
comparable meaningful components of it must be. 
This is the sense in which I claim that the con- 
straints that govern what definite descriptions 
can be used by a speaker who hopes to be under- 
stood are the same as those that govern what a 
speaker can in general say, if he has those same 
aspirations. There is no doubt that a detailed 
translation of my proposals about indirectness 
into comparable ones about definite descrip- 
tions is no easy matter. One reason is that the 
reasoning process that underlies the determina- 
tion of a referent may be from a conclusion to 
an initiating event, as in (3). Another is that 
not all the steps in the chain can be employed, 
but only those that contain information applicable 
to the referent--not, for example, generalizations 
like (ll). In such cases, if the premise is to be 
incorporated, it has to be embedded as the comple- 
ment of an appropriate verb of propositional 
attitude. Nevertheless, it seems to me that some 
of the notions that I have laid out might prove 
helpful, if only by virtue of the fact that they 
may eventually lead to better proposals by others. 
Earlier, I suggested that perhaps descrip- 
tions of inference involving indirect speech acts 
and those not involving them, really hinge on 
fundamentally the same kind of processes. On 
the surface, the basic difference concerns whether 
or ~ot they involve the addition of pragmatic 
knowledge. It turns out, however, not to be an 
easy matter to decide what is pragmatic knowledge 
and what is merely semantic or factual. For 
example, (16) is a description of inference: 
(i6) The navigator had heard that 
the weather might be unpleasant. 
He had always been concerned for 
the comfort of the passengers. 
He proposed taking a more indirect 
route to avoid the possible storms. 
The captain disagreed. He felt 
that the sooner they arrived at 
their destination the better--he 
wanted a drink and a decent meal. 
After a long argument the cautious 
one got his way. 
Now, for a hearer to determine that the cautious 
one and the navigator are coreferential, it is 
necessary for him to invoke general knowledge 
about what constitutes a cautious act (contrast 
this with an offensive act). This in turn re- 
quires inferences to be made about human actions 
and intentions. To be sure, the actions in 
question are not linguistically performed acts, 
but that appears to be the only difference. 
Furthermore, had the pilot asserted that he wanted 
to take the shortest route because of his frivo- 
lous desires, would he not, thereby have been 
indirectly recommending a (possibly) reckless 
act? Surely, what is pragmatic and what is not 
cannot come down to performative verbs. Yet, if 
it is to be broader than that, what criteria are 
to be used to separate the semantic from the 
pragmatic? The old notion of semantics as en- 
tailment is certainly too restrictive to be useful 
as a model of natural language processing, but the 
new notion of pragmatics seems to amount to little 
more than the notion that language processors are 
rational beings who engage their reasoning pro- 
cesses in language comprehension and production 
77 
just as they do in perception and action. Even 
the notion of a speech act seems to have very 
fuzzy boundaries unless it is trivialized by in- 
voking psychologically uninteresting surface 
structural aspects like the presence or absence 
of performative verbs. 
Distinctions between different classes of 
linguistic phenomena are usually difficult to 
maintain i'n any rigid way, particularly if they 
are supposed to have psychological correlates. 
This is true of the distinction between syntax 
and semantics, of that between semantics and prag- 
matics, of that between literal and nonliteral, 
and of that between descriptions of entailment 
and descriptions of inference. As usual, clear 
cases are easy to recognize, but there is always 
a large grey, undecided area in the middle where 
the classification seems sterile. In the case of 
the distinction between descriptions of entailment 
and descriptions of inference, the problem is 
exactly the same as the classical philosophical 
one that plagues the analytic/synthetic distinc- 
tion. This is hardly an accident since my dis- 
tinction is really no more than the analytic/ 
synthetic distinction in disguise. Maybe all that 
needs to be said is that some inferences (e.g. 
ones based strictly on the rules of logic) are 
generally easier to make than others. If this is 
right then it merely means that some relationships 
between descriptions and their intended referents 
are more transparent than others. Nobody could 
object to that. 
The last question I want to deal with is the 
psychological status of my claims, particularly 
with respect to the inference patterns that I 
have proposed. My position is not that it is a 
necessary condition for the comprehension or 
production of a definite description of inference 
that a person actually construct such a chain of 
reasoning. My claim is only that it should be 
possible to do so--there has to exist some de- 
terminable connection between the predicate 
underlying the definite description and the dis- 
course in which the description occurs. But, 
being determinable and being determined are dif- 
ferent things. As a matter of fact, there are 
often other clues that will permit the hearer to 
make a good guess about the referent's identity, 
discourse topic being one of them. It is almost 
certainly the case that people sometimes do go 
through some such reasoning process as I have out- 
lined, and if and when they do not, they could 
probably be induced to do so by being asked 
suitable questions about what they took the re- 
ferent to be, and why it was reasonable or 
plausible to do so. 
References 
Austin, J. L. How to do things with words. 
Oxford: Oxford University Press, 1962. 
Chafe, W. L. Discourse structure and human 
knowledge. In J. B. Carroll and A. O. 
Freedle (Eds.), Language comprehension and 
the acquisition of knowledge. Washington, 
D.C.: V. H. Winston, 1972. 
Grice, H. P. Logic and conversation. In P. Cole 
and J. L. Morgan (Eds.), Syntax and semantics 
(Vol. 3): Speech acts. New York: Academic 
Press, 1975. 
Morgan, J. L. Two types of convention in 
indirect speech acts. In P. Cole (Ed.), 
Syntax and semantics (Vol. 9): Pragmatics. 
New York: Academic Press, 1978. 
Ortony, A. Why metaphors are necessary and not 
just nice. Educational Theory, 1975, 2_55, 
45-53. 
Ortony, A., & Anderson, R. C. Definite descrip- 
tions and semantic memory. Cognitive 
Science, 1977, i, 74-83. 
Searle, J. R. Indirect speech acts. In P. Cole 
and J. L. Morgan (Eds.), Syntax and semantics 
(Vol. 3): Speech acts. New York: Academic 
Press, 1975. 
Searle, J. R. Indirect speech acts. In A. 
Ortony (Ed.), Metaphor and thought. 
Cambridge: Cambridge University Press, 
in press. 
Stalnaker, R. C. Pragmatics. In D. Davidson 
and G. Harman (Eds.), Semantics of natural 
l@nguage. Boston: D. Reidel Publishing 
Co., 1972. 
Footnotes 
l This research was supported in part by 
the National Institute of Education under 
Contract No. US-NIE-C-4OO-76-Ol16, and by a 
Spencer Fellowship awarded to the author by the 
National Academy of Education. I am grateful to 
my colleagues Glenn Kleiman, Robert Kantor, and 
Jerry Morgan for their helpful comments on 
earlier drafts of this paper. 
21 call them definite descriptions of 
entailment because technically they both are. 
The principle of identity, that pDp, represents 
an admittedly trivial entailment. It is impor- 
tant in the present context because it represents 
the case in which some predicate is literally 
transformed into the body of a definite descrip- 
tion. More complex cases are still based on 
the usual rules of propositional logic such as 
modus ponens, ((p~q).q) ~q). 
3Caution is needed here. Some cases of 
giving do not entail having. One can give 
somebody a pat on the back, or a kick in the 
teeth; the. recipient gets it alright, but he 
doesn't have it! However, if we specify the 
appropriate constraints on the object the en- 
tailment will hold. 
78 
