Toward a Rational Model of Discourse Comprehension 
Jerry L. Morgan 
Center for the Study of Reading 
University of Illinois 
at Urbana-Champaign 
I. Introduction 
I begin my tale with the moral: a quotation 
from the greatest English grammarian, Otto 
Jespersen (1965). 
The essence of language is human activity-- 
activity on the part of one individual to 
make himself understood by another, and 
activity on the part of that other to 
understand what was in the mind of the first. 
These two individuals, the producer and the 
recipient of language, or as we may more con- 
veniently call them, the speaker and the hear- 
er, and their relations to one another, should 
never be lost sight of if we want to under- 
stand the nature of language and of that part 
of language which is dealt with in grammar. 
But in former times this was often overlooked, 
and words and forms'were often treated as if 
they were things or natural objects with an 
existence of their own--a conception which 
may have been to a great extent fostered 
through a too exclusive preoccupation with 
written or printed words, but which is funda- 
mentally false, as will easily be seen with 
a little reflexion. (p. 17) 
But the temptation to think of language as pure 
form is great; Jespersen himself slips into this 
metaphor a few pages later: 
• • . we always find that there is one word 
of supreme importance to which the others are 
joined as subordinates. This chief word is 
defined (qualified, modified) by another word, 
which in its turn may be defined (qualified, 
modified) by a third word, etc. (p. 96) 
But words do not define, modify, or qualify other 
words. Speakers define, qualify, and modify. 
This confusion is so tempting that it is perva- 
sive in every field that studies language, at any 
level. It is almost universal in linguistics. We 
find it, for example, in the following from 
Halliday and Hasan (1976), who probably know 
better: 
Let us start with a simple and trivial exam- 
ple. Suppose we find the following instruc- 
tions in the cookery book: 
\[1:1\] Wash and core six cooking apples. 
Put them into a fireproof dish. 
It is clear that them in the second sen- 
tence refers back to (is ANAPHORIC to) the 
six cooking apples in the first sentence. 
This ANAPHORIC function of them gives cohe- 
sion to the two sentences, so that we inter- 
pret them as a whole; the two sentences 
together constitute a text. Or rather, they 
form part of the same text; there may be more 
of it to follow. 
The texture is provided by the cohesive RELA- 
TION that exists between them and six cooking 
apples. It is important to make this point, 
because we shall be constantly focusing at- 
tention on the items, such as them, which 
typically refer back to something that has 
gone before; but the cohesion is effected not 
by the presence of the referring item alone 
but by the presence of both the referring 
item and the item it refers to (P. 2). 
There are two serious confusions here. First, 
words do not refer; speakers refer to things by 
using words. The word them does not refer to 
anything at all, obviously so since it can be used 
to refer to any set one wants to refer to. There 
is no particular set of entitles that one can say 
the word them refers to. But one can use it to 
refer to sets of things, when one's intended 
referent will be recoverable in some way by the 
hearer. 
The second confusion is the idea that words 
"refer back" to other words. The muddle here is 
obvious. Whether it is people or words that 
refer, it is things, not (usually) other words, 
that they refer to. Thus in Halliday and Hasan's 
example \[1:1\], it is not the words six cooking 
apples that them is used to refer to; one is not 
being instructed to put three words in a fireproof 
dish. The word them is used to refer to certain 
apples that were previously referred to by use of 
the words six cooking apples. My objection to 
such descriptions is not based merely on a nig- 
gling concern with sloppy language. If it were, 
one might respond that it's clear what Halliday 
and Hasan mean here, so my complaint is beside 
the point. Rather, I think the pervasive confu- 
sion on just this point is a symptom of a serious 
comceptual confusion that renders a lot of the 
related work useless. This is the case with the 
passage from Halllday and Hasan. They say that it 
is some relation between sentences in a text that 
gives it "cohesion", that renders it coherent, 
109 
"so that we interpret them as a whole; the two 
sentences together constitute a text." The re- 
lation that gives rise to this cohesion is that 
them in one sentence "refers back" to the six 
cooking apples in a previous sentence. If we 
interpret this phrase charitably, then the ques- 
tion arises, how do we know what them refers to? 
How do we know that it refers to the apples, and 
not to two of the writer's bachelor Uncles? We 
can't know such a thing. We can only assume that 
the writer is rational, and that the recipe is co- 
herent. If it is coherent, we are justified in 
assuming that it is the apples that are referred 
to by them. But there is a vicious circularity 
here. The recipe has cohesion, is a coherent 
text, just in case them refers to the apples. But 
we are only justified in inferring that them re- 
fers to the apples if we assume that the text is 
coherent. Thus, in spite of Halliday and Nasan's 
claim, it is not the anaphoric facts that give 
rise to cohesion; rather, the assumption that the 
text is coherent gives rise to the inference that 
them refers to the apples. 
This kind of confusion, it seems to me, 
arises from the linguist's habit of looking at 
every aspect of language in terms of linguistic 
forms and relations between them. Thus in this 
case the mistaken characterization of reference 
as a relation between words, and of coherence as 
a property of an abstract linguistic object called 
a text. In the rest of this brief paper I want 
to sketch an opposing view, and to claim that 
notions like "reference," "text structure," 
"relevance" and "coherence" are best treated, at 
least in part, in terms of communicative acts and 
the plans and goals of speakers/writers who per- 
form such acts. 
II. Three Ways of Looking at a Text 
Assume for the moment that we know what a 
text (oral or written) is, and can tell a co- 
herent text from a random transcription of English 
sentences (I will return to what counts as a 
coherent text later). Then there are (at least) 
three kinds of things and relations involved in a 
text. 
I. Sentences. First, conventional wisdom in 
linguistics has it that texts consist of sen- 
tences. I shall accept this for the moment, 
though a bit later I will show cause to modify it. 
But what kind of "thing" is a sentence? 
It is, if anything is, an abstract linguistic 
object, a unit of form. It is not a proposition, 
nor a fact, though it is a means by which such 
things are asserted, denied, questioned, etc. 
Nor is a sentence a speech act, though a speech 
act will usually be performed by means of the 
utterance of a sentence. But a sentence and an 
utterance of a sentence are different kinds of 
things. 
A sentence is not the kind of thing that is 
true or false; "facts," or "propositions," that 
sentences can be used to express, are true or 
false. Or perhaps it would be more appropriate 
to speak of assertions as being true or false. 
At any rate, it is quite clear that it is nonsense 
to speak of sentences as true or false, as 
evidenced by the familiar problem of indexical 
expressions. 
A sentence, then, may be used to assert that 
something is true, or false, or has occurred, 
but the sentence itself is not true or false, 
and does not occur. Thus relations like causa- 
tion, order in time, entailment, and so on, do 
not hold between sentences. It is not clear what 
kind of relation can accurately be said to hold 
between the sentences of a text. 
2. "Facts." The second kind of "thing" involved 
in a text is what I shall call "facts." (Notice 
that I do not say texts consist of or contain 
facts; merely that they somehow involve facts.) 
The term "fact" is a bit misleading--though I 
can think of no better term--in that I wish to 
include as facts events, states, and so forth 
that do not actually hold in the real world; 
"propositions," more or less. 
Relations among the "facts" involved in a 
text, then, consist of two classes: first, the 
same relations that hold between facts in the 
real world--causation, relations of temporal 
order, motivation, and so forth; second, those 
relations that have to do with logic and hypo- 
thetical facts, like entailment and contradic- 
tion. It may be necessary to distinguish facts 
on the one hand and propositions on the other, 
on grounds that relations between facts are of a 
kind different from relations between proposi- 
tions, but I will ignore the problem here. 
3- Speech acts. The third kind of thing involved 
in texts is the "speech act" (by this term I 
mean to include as a sub-case acts of linguistic 
communication by writing). Speech acts are not 
sentences, nor "logical forms," nor propositions, 
in spite of occasional attempts to define them 
in these terms. They are acts, just as the term 
implies. 
Relations between the speech acts involved 
in a text are just those that can hold between 
acts in general. First, since an act is a sub- 
type of event, the relations that can hold be- 
tween events can, in general, hold between acts, 
thus between speech acts: relations of temporal 
order, for example. Second, since a speech act 
is a sub-type of act, relations that can hold 
between acts can, in general, hold between speech 
acts. The most important relation in this regard 
is the relation of purpose: one does such-and- 
such in order that such-and-such; or one performs 
a certain act in order thereby to perform a 
second act. Long chains of these relations can 
hold between acts. I may throw a switch in order 
to turn on a light in order to frighten away a 
burglar in order to save the family jewels. I 
may tell my friend that there is a charging bull 
behind him in order that he realize that he is 
in danger, in order that he get out of the way. 
It is a mistake to ask whether my speech act was 
an assertion or a warning, since this presupposes 
that the two are mutually exclusive. It was both; 
I asserted something and thereby warned somebody, 
just as I threw the switch and thereby turned on 
the lights. I may make a certain mark on a 
piece of paper, thereby marking my ballot for 
Smith, thereby casting a vote for Smith. I may 
ii0 
assert that I will do the dishes, thereby volun- 
teering to do the dishes. And so on. 
It is commonly the case that acts are linked 
by complex relations of purpose and goal, includ- 
ing the case where one act is performed by means 
of performing another act. This is especially 
true of communicative acts. 
There are several subvarieties of speech 
acts, for which several taxonomies have been pro- 
posed; Austin (1962), McCawley (1977), for example. 
One important distinction in kind is the distinc- 
tion between the act of saying a sentence, and 
the act one thereby performs. In performing an 
act of saying the English sentence "Your hair is 
in my yogurt" I may, in the right circumstances, 
thereby inform someone that their hair is in my 
yogurt. The first kind of act, the act of saying, 
includes making sounds in a way that conforms to 
the conventions for what counts as a saying of a 
sentence, or making visible marks in a way that 
counts as a saying of a sentence. Texts, then, 
do not really consist of sentences, but of sayings 
("uses") or sentences; or in the case of written 
texts, of a permanent kind of record of uses of 
sentences. 
III. The Interpretation of Texts 
A. Speech acts. The interpretation of a 
text, then, consists of the interpretation of this 
record of sayings of sentences. Each saying is 
interpreted in terms of some speech act(s) per- 
formed by saying a given sentence. (Henceforth 
by "speech act" I will mean the communicative act 
one performs by saying a sentence, as opposed to 
the act of saying itself.) There are three 
aspects to the interpretation of speech acts: the 
interpretation of what speech acts are performed-- 
assertion, promising, denial, questioning, 
warning, etc.--by the saying of the sentence; the 
interpretation of what "facts" are asserted, 
denied, etc.; and the interpretation of the 
speaker's purpose and goal in performing the 
speech act. 
As an aside I should mention the special 
instance where nothing is directly asserted, 
denied, etc.: the case of speech acts of refer- 
ence. An act of asserting, etc. (for brevity I 
will henceforth use assertion as representative 
of all speech acts types), will usually include 
an act of referring as a subpart; a reference to 
the entity of which something is asserted. But 
acts of referring can occur independently. For 
example, l might say "The door~" to someone under 
a number of circumstances, to get them to open 
it, close it, shoot the bad guy standing in it, 
or merely observe what beautiful hardwood it is 
made of. It would be a mistake to say that 
"The door~" means any of these things, or that 
I have performed (directly) any kind of speech 
act beyond merely referring. I have only re- 
ferred to the door, thereby to call my hearer's 
attention to it, with the expectation that when 
he turns his attention to the door he will realize 
what it is I want him to do about it. 
The typical immediate goal associated with 
speech acts of all kinds is the same: that the 
hearer modify his model of a certain "world" 
(in the sense of "possible worlds ~') in a way 
that involves the "facts" that are asserted, etc. 
in the speech act. The world involved may be 
the real world, or, in the case of story-telling, 
for example, some imaginary world. The modifica- 
tion may include the construction ex nihilo of 
some hypothetical or imaginary world. The rela- 
tion between the "f~cts" of the speech act and 
the intended m0difi~ation vary with the nature 
of the speech act; but in all cases some modifi- 
cation is involved. The simplest case is that of 
assertion; normally the immediate goal of an 
assertion is that the hearer modify the world 
under discussion in a fashion that makes the 
asserted fact true in that world. In the case of 
yes-no questions, the goal is that the hearer 
modify his model of the world such that in that 
world the speaker wants the hearer to tell him 
whether the fact questioned is true. In the case 
of imperatives, the goal is that the hearer 
modify his model such that in that world the 
speaker wants the hearer to bring about the truth 
of the ordered fact, and that certain social 
consequences will follow from non-compliance. 
The raw datum of comprehension, then, is not 
the sentence or the proposition, but the fact that 
a certain speech act has occurred. In comprehen- 
sion, people do not process sentences as abstract 
formulae; they observe that someone has said 
something to them, and attempt to interpret that 
act and its consequences, which may include 
modification of their model of the world. The 
process of modifying the model according to what 
is said is not direct, but the result of several 
steps of evaluation. Interpretation of an 
assertion might go roughly like this, from the 
viewpoint of the hearer (where S is the speaker, 
A the addressee; addressee and hearer may be 
identical): 
S has said x to A. Saying x counts as 
asserting p. S knows that saying x 
counts as asserting p. S knows that 
therefore his saying x is likely to be 
interpreted by A as an assertion of p. 
S has done nothing to prevent A from 
making this conclusion. Therefore S's 
intention is that his saying x be taken 
by A as an assertion of p. Then if S 
is sincere S believes that P is true. 
A must conclude that S has asserted p 
because he wants A to take p as true 
and modify his model of the world 
accordingly. 
But the decision to believe p, i.e. modify his 
model of the world to include p, is a matter of 
choice on H's part, not an automatic consequence 
of processing the "sentence." The steps involved 
in making this decision are equally complex, 
involving the ability to construct a hypothetical 
world just like the real one except that p is 
true, to evaluate the consistency and plausi- 
bility of that world, and so on. Some of the 
facts that are asserted will relate to this 
decision-making process. For example, in saying 
(1) my goal is most likely that the hearer come 
to believe that both facts asserted are true. 
(1) John is here. He has a dog with him. 
iii 
But in the case of (2), I am not so much con- 
cerned with the second asserted fact in itself, 
but with the goal that from concluding that it is 
true, the hearer will be more likely to believe 
the first, since I intend that he take the second 
fact as evidence that my source is reliable. 
(2) The world is flat. It says so 
in the Encyclopedia. 
Matters that are sometimes construed as rhetorical 
relations between sentences fall into this cate- 
gory. Some fact is asserted not because it is 
important in itself, but because it bears on H's 
evaluation of some other asserted fact. Thus the 
relation is not one between sentences, but be- 
tween speech acts. One speech act is performed 
in order to influence the interpretation and 
evaluation of another. At any rate, my point here 
is that in comprehending a text in the serious 
sense, comprehension proceeds not from some dis- 
embodied abstract object called a "sentence," nor 
from a "proposition," but from the perceived fact 
that S has said such-and-such, and that so saying 
counts as a speech act of a certain type. 
There i~ another way in which modification 
of the world model is not a direct function of 
the asserted fact: the widely studied problem 
of inference. Given the hearer's acceptance of 
what the speaker has asserted, incorporation of 
the facts into the model of the world may involve 
more than merely adding the asserted facts. 
There is, for example, a general principle of 
ceteris paribus that comes into play in considera- 
tion of alternative worlds. Roughly, when con- 
structing a model of a world alternative to some 
point-of-reference world (usually the real one), 
the hearer will assume, lacking evidence (from 
assertion or inference) to the contrary, that the 
alternative world is consistent with the point-of- 
reference world in all relevant respects. To take 
an extreme example, if someone is telling me about 
life on Arcturus, I will assume that the laws of 
physics are the same there as on earth, unless 
something the speaker says leads me to believe 
otherwise. In the same way, hearers will assume, 
lacking counter-evidence, that what is typical 
in the point-of-reference (e.g. real) world is 
also typical in the alternative world. They will 
also assume that things of a given type have the 
properties typical of things of that type. 
Gricean rules of conversation support these in- 
ferential strategies in the following way: The 
hearer knows that the speaker knows the hearer is 
likely to make inferences according to these and 
other strategies. The speaker has done nothing 
to prevent the hearer from making them. There- 
fore the hearer is justified in inferring that 
the speaker intends for the inference to be made. 
Using these and other strategies, then, the 
hearer modifies his model of one or more worlds, 
based not on detached sentences or propositions 
floating in some abstract semantic space, but on 
his observation that a certain person has per- 
formed a certain speech act. 
B. Relations among speech acts. But there 
is more to the interpretation of a text than just 
the interpretation of individual speech acts. A 
speech act is performed for some purpose, with 
some goal in mind. And complete understanding 
of a text involves the ability to infer such 
goals and purposes at every level, from in- 
ferring the purpose of referring expressions to 
inferring the speaker's overall goal in con- 
structing the text. One can understand every 
sentence in a text, yet come away puzzled at what 
it was the speaker was trying to say, or what 
the parts of the text had to do with each other. 
To understand the purpose of a speech act is to 
understand how it relates to a goal, how it is a 
step toward the achievement of that goal. The 
most appropriate kind of theory for this aspect 
of a text is a theory of plans, in which pur- 
poses, goals, acts, and intentions play a crucial 
role. 
There are a large number of goals a speaker 
can have in constructing a text, including many 
that are irrelevant to comprehension: to derive 
royalties, for example, or to confuse an enemy 
by furnishing misinformation. A proper theory 
of text comprehension must distinguish goals 
like these from those that are central to com- 
munication and comprehension, probably by means 
of conditions like those Grice (1957) proposes 
as criteria for meaning. 
C. What can go wrong. Then we can sketch 
the task of text comprehension as follows: 
l. From the sounds or markings, H must 
recover what sayings are recorded in 
the text, in what order. 
2. From this H must recover what speech 
acts have been performed, in what 
order. 
3. From each speech act H must recover 
what facts are being asserted, denied, 
promised, etc. 
4. From this H must infer what modifica- 
tions he is intended to make in his 
model of the world, and how to make 
them in the most consistent way; 
this is not a direct function of 
the facts, as discussed earlier. 
5. For each speech act H must infer a 
purpose that is consistent with the 
purposes he inferred for earlier 
speech acts; or he must revise 
earlier hypotheses about purposes 
accordingly. Questions H must infer 
answers to are, "Why did the speaker 
perform this particular speech act, 
at this particular point in the text?" 
and "Why does he want me to have 
this particular fact just now?" 
6. From speech acts and their purposes 
taken jointly, he must construct a 
hypothesis of the speaker's goal in 
the text, and of the plan that the 
speaker is following in advancing 
toward that goal. At each step the 
purpose of a given speech act must 
somehow be construed as consistent 
112 
with, and actually advancing that 
plan, or the plan hypothesis must 
be modified so that it can. 
7. From hypotheses about the speaker's 
plans and goals in the text, the 
hearer will form expectations: 
hypotheses about what the speaker 
is likely to do next in advancing 
toward the goal of the text. 
These matters do not proceed in separate compart- 
ments, of course, but feed each other. The plan 
one has constructed so far can influence decisions 
about what speech act is performed in a given 
utterance, for example, and the interpretation of 
pronouns can be influenced by hypotheses about 
the speaker's goals, just as a decision about what 
a referring expression is being used to refer to 
can influence the process of inferring a plan, and 
expectations about what the speaker will do next 
can influence the interpretation of what he 
actually does. 
From this sketch we can derive a picture of 
where things can go wrong in comprehension, giving 
some insight perhaps into notions like "text 
structure," "relevance," and "coherence." 
The hearer can have difficulty in tasks 1 
through 3, of course, but the matter seems 
straightforward, so I will not discuss it. Diffi- 
culties can arise in task 4 in at least two ways. 
First, the world described may be so factually or 
logically bizarre, or so inconsistent with the 
hearer's beliefs (a description of ping pong in a 
black hole, for example), that the hearer is un- 
able to construct a consistent model with any 
degree of detail. The term "incoherent" might be 
applied to such cases, but I think this is not 
what linguists mean by "textual coherence," which 
I will discuss below. 
A second kind of difficulty with task 4 
arises when the facts are consistent, but the 
hearer lacks the knowledge necessary to figure 
out how to construct a consistent model that 
incorporates those facts. For example, if I 
describe in detail a walk through the South Side 
of Chicago, a person who has been there before 
will be able to construct a much more richly de- 
tailed model of my walk than a person who has not. 
Difficulties can arise with task 5, insofar 
as the hearer is able to understand clearly what's 
being asserted, but unable to determine the 
speaker's purpose in asserting it. Here is the 
place to look for an adequate definition of 
relevance. Actually there are two senses of 
the word in ordinary usage. One can speak of 
relevance as a relation between facts. One fact 
is relevant to another when the truth of one de- 
pends in some way on the truth of the other. But 
I think more often, linguists who speak of 
"relevance" as a problem of text comprehension 
have in mind a problem that is best treated in 
terms of purposes behind speech acts. Given a 
hypothesis about the goal and plans of a speaker 
in a text, a given "sentence" (i.e. speech act) 
is taken to be irrelevant when the hearer is un- 
able to see how it functions within the plan to 
advance toward the goal. Relevance under this 
interpretation, then, is a relation between an 
act and a goal, not a relation between sentences. 
If in recounting my recipe for Wienerschnitzel 
I describe my new driveway, it's not that the 
sentences are irrelevant; rather, I have done 
something irrelevant. The same passage may 
count as full of irrelevancies, relative to one 
goal, but uniformly relevant, relative to another 
goal. 
Task 6 is probably the most complex and 
difficult, and the one we know least about. But 
I suspect that it is a likely source of progress 
in understanding such important but elusive 
notions as "coherence," "text structure," and 
"topic." In understanding a text, the hearer 
unconsciously searches out a primary goal behind 
the text, and tries to construe every part of the 
text as a purposeful step toward that goal, 
according to some plan. If the hearer is unable 
to reconstruct the goal or plan, or indeed 
decides there is none, the text will be judged 
"incoherent." Coherence is not a formal property 
of texts, nor of "logical structures" of texts, 
but a function of the hearer's ability to relate 
of the text to a pla~ for parts achieving some 
goal. If it should turn o~t that the coherence I 
of texts correlates with the number of pronouns, 
it would be a mistake to conclude that lots of 
pronouns makes a text coherent. Rather, it would 
show that coherent texts tend to be ones where 
the speaker says a lot about one or two topics, 
rather than saying one thing about 32 topics. 
It is the coherence of what the speaker is doing 
in the text that gives rise to the abundance of 
pronouns; the formal property of having a lot of 
pronouns does not give rise to coherence. 
At least some aspects of "text structure" 
can also be treated in these terms. An ideal 
unified paragraph, for example, is a unit of 
function, not of form; the speaker formulates 
a subgoal as a step toward the primary goal of 
the text, and sets about to achieve, that goal in 
a series of speech acts. Insofar as the hearer 
is able to discover this, the series of speech 
acts will be judged to be a unit; but a unit of 
function, not of form, defined not in terms of 
sentences or propositions, but communicative acts 
of some person, who uses those sentences to convey 
those propositions. 
It is likely that an understanding of task 6 
will lead to an understanding of "topic" as 
well. At present, there are nearly as many 
definitions of "topic" as there are linguists, 
and none of the definitions is clear enough to 
be usable. For some linguists the topic is a 
certain NP in a sentence; for others a topic is 
something a sentence has, though the NP may not 
be present in the sentences. For some every 
sentence has a topic; for others, only some 
sentences have topics. But I suspect that all 
of these attempts miss by a wide mark. First, 
it is not NP's that are topics, but the things 
in the world they refer to. Second, I suspect 
that such definitions can never be made sense of 
in that i£ is speakers, not sentences or even 
texts, that have topics. If so, then the proper 
theoretical treatment of "topic" would be framed 
113 
in terms of a theory of complex communicative 
acts, not formal linguistic properties. 
IV. Conclusion 
In this speculative paper I have proposed a 
way of looking at the comprehension of connected 
text that is counter to the linguist's usual way 
of looking at language. My main point is that 
certain notions are more likely to receive ade- 
quate treatment in a theory that incorporates a 
theory of speech acts, a theory of plans and goals, 
and a theory of inference, in place of a theory 
that looks for answers in terms of formal proper- 
ties of texts. It remains, of course, to develop 
such theories to a level where my claims can be 
rigorously tested. The construction of such 
theories should be a prime goal of theoretical 
linguistics. 
References 
Austin, J. How to do things with words. Oxford: 
Oxford University Press, 1962. 
Grice, H. P. Meaning. Philosophical Review, 
1957, 66, 377-388. 
Halliday, M. A. K., & Hasan, R. Cohesion in 
English. London: Longman, 1976. 
Jespersen, O. The philosophy of 9rammar. New 
York: Norton, 1965. 
McCawley, J. Remarks on the lexicography of 
performative verbs. In A. Rogers, R. Wall, 
and J. Murphy (Eds.), Proceedings of the 
Austin Conferenee on Performatives, Prer 
suppositions, and Implicatures. Arlington, 
Va.: Center for Applied Linguistics, 1977. 
Footnote 
This research was supported by the National 
Institute of Education under Contract No. 
US-NIE-C-4OO-76-OII6. 
114 
