The use of discourse analysis in
neurolinguistics: some findings from the
narratives of hemidecorticate adolescents
Jean E. Newman, PhD
Assistant Professor
Departments of Linguistics
and Psychology
University of New Mexico
Albuquerque, New Mexico
Maureen W. Lovett, PhD
Associate Professor of Pediatrics
The Hospital for Sick Children
and University of Toronto
Maureen Dennis, PhD
Associate Professor
Research Institute
The Hospital for Sick Children
Associate Professor
Department of Behavioral Science
University of Toronto
Toronto, Canada
BASIC FACT of language research is
that language use in isolation often
bears little resemblance to language use in
context. The researcher who hopes to
learn about the capabilities of the larger
language system by studying its compo-
nents may well find, on assembling the
pieces, that the Gestalt psychologists were
right—the whole is greater than the sum
of its parts. Yet researchers often rely on a
componential approach, either for ease of
study or because, in the case of neurolin-
guistics, the data appear to demand a
reductionist approach. The general ques-
tion to be pursued here is whether the
application of a more global, discourse-
oriented approach to a neurolinguistic
problem can provide insights into the
mechanics of the linguistic systems of the
affected individuals. Of specific interest
are the linguistic contributions of the left
and right hemispheres, which will be
examined by studying the use of pragmat-
ics in the narratives of three hemidecorti-
cate adolescents.
31382
‘TOPICS IN LANGUAGE DISORDERS / DECEMBER 1986,
OVERVIEW OF APPROACHES TO
NEUROLINGUISTICS
The study of those individuals who,
through brain injuries or birth defects,
have developed aberrant linguistic sys-
tems has presented researchers with useful
natural experiments. For years researchers
have concentrated on describing the dis-*
tinctive properties of these subsystems,
including anomias (failure to name when
presented with an object; e.g., Geschwind,
1967), agrammatism in Broca’s aphasics
(eg, Zurif, Caramazza, & Myerson,
1972), and comprehension deficits in
Wernicke’s aphasics (e.g., Grober, Perec-
man, Kellar, & Brown, 1980).
Although there is evidence from a vari-
ety of sources in favor of hemispheric
language specialization, the nature of
unique left and right hemisphere lan-
guage abilities is still being explored
(Blumstein & Cooper, 1974; Tucker, Wat-
son & Heilman, 1977; Weintraub, Mesu-
lam & Kramer, 1981). Recent research has
addressed the extent to which right hemi-
sphere language is imaginal and affective,
rather than propositional (e.g, Millar &
Whitaker, 1983).
‘The language abilities of brain-injured
individuals are usually studied for one of
the following reasons: to establish (on the
basis of spontaneous speech samples or of
speech and language tests) how language
is used and understood in persons or
groups with neuropathology; or, alterna-
tively, to see how samples of brain-injured
language can be used to formulate or
refine psycholinguistic theories about the
organization of the language system (e.g.,
Caramazza & Zurif, 1976; Goodglass &
Baker, 1976). Language comprehension
and use in context, however, has not been
ignored by researchers. Recently a less
extensive but interesting strand of
research on the discourse capacities of
brain-damaged individuals has been de-
veloped, consisting of a set of studies
designed to clarify the ability of various
pathological groups to use linguistic
devices that promote cohesion in dis-
courses and texts (eg. Engel-Ortlieb,
1981; Ulatowska, Doyel, Stern, & Haynes,
1988).
THE ROLE OF PRAGMATICS IN
DISCOURSE PROCESSING
The last several years have seen consid-
erable interest in an area that was for-
merly considered to be outside the realm
of linguistic inquiry. Spurred by develop-
ments in the philosophy of language (Aus-
tin, 1962; Grice, 1975; Searle, 1969), text
linguistics (see de Beaugrande & Dressler,
1981, for a review) and the influence of
functionalism in linguistics (e.g., Bolinger,
1977; Halliday, 1985; and the Prague
School, e.g., Dane3, 1960), researchers
have become increasingly aware of the
importance of studying language in con-
text. Speech act theory (Austin and Searle)
and Grice’s description of conversational
“logic” have drawn attention to the rela-
tion between the speaker's intention, the
words chosen to express this intention, and
their combined effect on listeners. Each
utterance can be classified as an instance
of “speech act,” as the purpose of uttering
is to “do things with words” (Austin’s
[1962] title). An act can be direct, in which
case the intent is clear from the words
themselves, or indirect, in which case the
listener must rely on knowledge or context(including perhaps knowledge of the
speaker's past behavior).
The development of text and story
grammars, which rely on propositional
accounts of meaning and rewrite rules
similar to those found in sentence gram-
‘mars, has given rise to formal descriptions
of extended discourse of varying degrees
of complexity. Psychologists have found
these formalisms to be particularly useful
because the derived representation of a
text (based on underlying meaning rather
than form) can be used to predict patterns
of comprehension and recall (e.
Mandler & Johnson, 1977; McKoon &
Ratcliff, 1980).
Another approach to the study of
extended discourse in context has focused
on descriptive analyses of the wording (or
surface features) of texts that contribute
to their “textuality” and resulting compre-
hensibility. Linguists such as Halliday and
Hasan (1976) and Chafe (1976; 1982)
focus on the function of a particular utter-
ance in its context. Halliday, for example,
has written extensively on the roles of
given and new information in discourse
(see Halliday, 1970, for a summary).
Given information is presupposed, situa-
tionally salient, or shared (Prince, 1981),
usually receives reduced stress, and is typi-
cally utterance-initial. New information
usually receives the greatest pitch move-
ment and is either utterance-final or the
focus of a marked construction. The sen-
tence It was Rover who ate the roast
illustrates the use of a cleft construction to
mark Rover as new information. Although
the propositional content is the same, the
roast, not Rover, is more likely to be
interpreted as new information in the
active version, Rover ate the roast. The
DISCOURSE ANALYSIS IN NEUROLINGUISTICS
existence of markers to differentiate new
from given information allows readers
and listeners to allocate processing
resources efficiently during comprehen-
sion (see Cutler, 1976). Clark and Havi-
land (1977) showed that on encountering
new input in a discourse, readers attempt
to determine whether it is given by search-
ing memory. Failure to find a prior men-
tion or related information forces readers
to spend additional time integrating the
new information,
he burgeoning field of discourse anal-
ysis has been fed by three lines of inqui-
ry—speech act theory, text grammars,
and surface feature analyses—each con-
cerned with pragmatics, or the study of
(purposeful) language use in context.
Although they have been presented as
separate influences, they clearly overlap.
Nonetheless, the present focus is on sur-
face feature analyses and their application
to neurolinguistics.
Studies of surface features, such as
cohesion analysis, provide a convenient
starting point for the assessment of dis-
course knowledge because the surface
features of an utterance are more immedi-
ately available than the underlying struc-
ture, This immediacy directly affects pro-
cessing demands, and as a result, the ease
of comprehension. Furthermore, as was
previously discussed, there are theoretical
reasons for supposing that a speaker's use
of surface features is indicative of his or
her linguistic competence.
Cohesion analysis
Cohesion analysis, originally proposed
by Halliday and Hasan (1976), describes
the aspects of the surface structure that
are responsible for connecting linguistic
3384
TOPICS IN LANGUAGE DISORDERS / DECEMBER 1986
elements to form a text. The prerequisite
for any text is that it be semantically
coherent (i.e, make sense), but this fea-
ture does not necessarily lead to its being
cohesive. Sometimes a speaker may
attempt to present a narrative but fail to
provide the listener with the relevant con-
nections between one semantic unit and
the next, in which case the text is said to
lack cohesion. If, for example, a speaker
uses pronouns from the beginning of a
story, She did it to him. They saw it. John
blamed him, the listener may understand
the sequence of events but be unable to
understand the topic. Who are she, him
and they? Is John the recipient of what-
ever she did? Simply stated, the speaker
has failed to provide referents for the
pronouns and pro-verb (did it), and has
provided a referent without clearly mark-
ing its relation to the preceding discourse.
Without these mental “anchors,” the lis-
Without these mental “anchors,” the
listener will be unable to build a
coherent representation of the
developing text and may give up
attempts to process it as a lost cause.
tener will be unable to build a coherent
representation of the developing text and
may give up attempts to process it as a lost,
cause.
Discourse that is cohesive can in unu-
sual circumstances (such as examples
invented by linguists) be incoherent, as the
following example demonstrates (as cited
by Brown & Yule, 1983, p. 197):
I bought a Ford. A car in which President
Wilson rode down the Champs Elysees was
black. Black English has been widely dis-
‘cussed. The discussions between the presidents
ended last week.
This example appears to be cohesive
because it uses surface ties such as lexical
substitution (a car for a Ford) and lexical
repetition (black followed by Black
English) between adjacent utterances.
However, it is obviously semantically
incoherent.
These examples readily differentiate
cohesion and coherence. Listeners and
readers, however, normally assume that
speakers and authors abide by some ver-
sion of Grice’s Conversational Maxim to
be relevant (Grice, 1975), and in doing so,
convey coherence through the use of cohe-
sion. Individuals who give the appearance
of local cohesion without coherence are
usually judged to be thought-disordered or
impaired in some way, such as Wernicke’s
(posterior) aphasies or schizophrenics (see
Rochester & Martin, 1979). The posterior
aphasic in the following example demon-
strates cohesion by the use of a pronoun
(they) for a previously mentioned noun
(barbers), but the sentence as a whole
makes no sense (from Gardner, 1976, p.
68).
Oh, I'm taking the word the wrong way to say,
all of the barbers here whenever they stop you
it's going round and round...
Examples of coherence, with varying
degrees of successful attempts at cohesion,
can be found in the speech of Broca’s
(anterior) aphasies for whom the produe-
tion of elements marking discourse rela-
tions is particularly difficult, as in the next
example (from Gardner, 1976, p. 61):
Clinician: And have you been going home on
weekends?
Patient: Why, yes... Thursday, er, er, er, no,
er, Friday... Bar-ba-ra. .. wife... .and, oh,car... drive. .. purnpike (sic).
.-restand...TV
you know
Although cohesion may seem to be a
subtle aspect of language use in normal
individuals, it has an important effect on
comprehension. The more consistently a
speaker uses cohesive devices to provide
cues to the discourse topic and the rela-
tions between utterances, the more likely
it is that the listener will correctly, and
easily, arrive at the speaker's intent. Stud-
ies of one important form of cohesion,
anaphoric reference, have shown that sub-
jects take longer to comprehend sentences
in which a pronoun is potentially co-
referential with two previously mentioned
proper nouns (Caramazza, Grober, Gar-
vey, & Yates, 1977), as in John hit Bill
because he was angry. Listeners clearly
expect speakers to make unique assign-
ments of pronouns to referents,
Anaphoric reference
The study of the referential system in
general, and anaphora in particular, has
been particularly important in linguistics,
psycholinguistics, and artificial intelli-
gence. Anaphora (in which one linguistic
entity refers backward to another having
the same identity) represents an economi-
cal means of coding meaning relations in a
text without having to resort to high levels
of redundancy. Psycholinguists have stud-
ied three classes of variables that affect the
listener's assignment of a referent for a
pronoun: (a) information processing; (b)
thematic; and (c) semantic.
One information processing constraint
that affects the degree of activation of a
particular entity is the capacity of short
term memory. For example, Clark and
Sengul (1979) showed that it is easier to
DISCOURSE ANALYSIS IN NEUROLINGUISTICS
identify the referent for a pronoun when
it is mentioned in the preceding sentence
(or clause). With intervening sentences, it
becomes more difficult to determine the
referent as the following sequence shows
(Clark & Sengul, 1979, p. 85): Yesterday I
‘met a woman who had written a book on
viruses. She had studied them for years
and years. It was selling very well.
Because the referent for it is mentioned
two sentences previously, more time must
be devoted to searching short-term mem-
ory for the referent than if the second and
third sentences were reversed.
Surface structure changes that alter the
information structure of an utterance are
termed thematic and include passiviza-
tion (Caramazza & Gupta, 1979) and
emphatic stress (Newman, 1981). These
variables also affect referent choices. For
example, Bill is more likely to be selected
as the referent in the sentence Bill was hit
by John because he was angry than in the
original, active, sentence. Semantic prop-
erties of the surrounding text, such as the
choice of verb, are the third influence on
referent choice. For example, the use of
avoided in the preceding sentence favors
Bill as the referent, whereas phoned
favors John (Garvey & Caramazza, 1974).
Two other semantic variables, the current
discourse context and the listener's knowl-
edge of the world (see Hirst & Brill, 1980),
also have powerful effects on referent
choice.
The processes governing the speaker's
decision to use a pronoun to refer to a
particular entity are less clear. However,
by assuming that speakers behave in
accord with some notion of a conversa-
tional Cooperative Principle (Clark &
Haviland, 1977; Grice, 1975), we may
conjecture that speakers use pronouns
38586
‘TOPICS IN LANGUAGE DISORDERS / DECEMBER 1986
judiciously and are aware of the three
classes of constraints faced by listeners.
For example, speakers are expected to
introduce referents before using pronouns
to refer to them. Failure to do so would
indicate that the speaker had violated
what Clark and Haviland have termed the
Given-New Contract (i.e., in this example,
to provide sufficient background informa-
tion to allow the listener to find an ante-
cedent for the pronoun).
Speakers might also be expected to fol-
low some principle of cognitive economy
as they are undoubtedly bound by infor-
mation-processing constraints of their
own, Such constraints would cause them to
use pronouns when feasible rather than
unnecessarily reintroducing the referents.
Indeed, using a proper name where a
pronoun will do might lead the listener to
assume erroneously that the speaker has
introduced a new person with the same
name. Appropriate use of a referential
system provides cohesion and indicates
topical links across utterances. Texts with-
cout these cohesive devices appear to be
disjointed and lacking in fluency, and are
more difficult to understand.
The role of pragmatics in language pro-
cessing has only recently been addressed
in the neurolinguistic literature. Foldi,
Cicone, and Gardner (1988) point out a
number of pragmatic aspects of language
that are affected by brain damage. While
it provides a number of interesting find-
ings and suggestions for future research,
their review is far from conclusive. Inton-
ation, for example, appears to be a pri-
marily right hemisphere function when its
affective and perceptual aspects are con-
sidered as shown by findings of flat inton-
ation contours in right hemisphere-dam-
aged patients (Foldi, Cicone, & Gardner,
1983) and the results of dichotomic listen-
ing tasks (Blumstein & Cooper, 1974). The
possible contribution of the left hemi-
sphere to propositional uses of intonation,
however, remains unresolved. For exam-
ple, Foldi, Cicone, and Gardner (1983)
cite unpublished evidence that Wer-
nicke’s aphasies produce “hypermelodic”
contours that are clearly divorced from
semantic content. The contours of the
effortful, agrammatic utterances of Bro-
ca’s aphasics, in contrast, show evidence of
planning and are appropriately adjusted
to accommodate utterance length.
Foldi, Cicone, and Gardner (1988)
argue that aphasies generally retain some
aspects of pragmatics (such as the distinc-
tion between direct and indirect speech
acts) even if the semantic content of their
utterances is unclear. Furthermore, right-
hemisphere damage can disrupt prag-
matic knowledge even though (proposi-
tional) linguistic knowledge may be intact.
Thus the disruption of left-hemisphere
processing in aphasics may be partially
compensated for by the remaining prag-
matic abilities of the right hemisphere.
For example, Foldi, Cicone, and Gardner
(1983) report that Broca’s aphasics may
rely on the environment, gesture, and
knowledge of conversational rules in order
to clarify the referent in an utterance.
Broca’s aphasics were also reported to be
sensitive to the distinction between given
and new information, Wernicke’s apha-
sics, in contrast, were not, nor were they
capable of using pragmatic information to
signal reference.
Tt should be apparent that the roles of
the right and left hemispheres in the pro-
duction and comprehension of connected
discourse are far from understood.
Although the hypothesis of right-hemi-sphere dominance for processing prag-
matics is attractive, it seems far too simple.
The pragmatic components of language
appear to be additionally affected by left-
hemisphere damage; otherwise the differ-
ences reported to exist between Broca’s
and Wernicke’s aphasics (both left-hemi-
sphere damage) for reference and given-
new information could not have been
obtained. Clearly, further exploration of
the role of pragmatics in discourse is
important for a full understanding of lan-
guage use in brain-injured populations.
DISCOURSE PROCESSING IN
HEMIDECORTICATE
ADOLESCENTS
The cohesive use of pronouns
One way to study the pragmatic func-
tions of discourse production and compre-
hension in brain-injured individuals is to
explore how pronouns are manipulated in
their oral narratives to achieve referential
cohesion. Establishing pronominal refer-
ence requires the simultaneous apprecia-
tion of information about the semantic
content and the grammatical context in
which a pronoun occurs. In addition, a
study of pronoun cohesiveness in brain-
damaged subjects, for whom other lan-
guage test performance is available, not
only might be expected to reveal the level
of narrative discourse skill, but also the
relationship between conventional mea-
sures of language test performance and
newer methods of evaluating how lan-
guage is used and understood in context.
The language of young
hemidecorticates
A number of published studies discuss-
ing the cognitive abilities of three young
DISCOURSE ANALYSIS IN NEUROLINGUISTICS
subjects after a hemispherectomy have
been reported (e.g., Dennis, 1980; Dennis,
1983; Dennis & Kohn, 1975; Dennis, Lov-
ett, & Wiegel-Crump, 1981; Dennis &
Whitaker, 1976). These individuals have a
medical history that is as dramatic as it is
unique. All were born with a brain malfor-
mation that caused early and intractable
epileptic seizures that were treated by
removal of half of their brain mass in the
first weeks of life (Hoffman, Hendrick,
Dennis, & Armstrong, 1979). Since the
removal of their malformed brain tissue
antedated speech development, this
meant that the subjects had to acquire
language with one cerebral hemisphere
rather than two. The speech and language
abilities of these subjects have been stud-
ied in detail because such data bear on the
important question: Which skills can the
left and the right cerebral hemisphere
master when each brain half develops in
isolation?
‘The answer to this question is complex.
Language under conditions of early and
atypical lateralization was not uniformly
good or consistently poor. Rather, there
has emerged a characteristic pattern of
language development. Either brain half
could acquire and maintain normal mech-
anisms of speech production and articula-
tion, a well-functioning semantic system,
and a fair degree of verbal fluency. How-
ever, when the derivation of meaning is
dependent upon responsiveness to lan-
guage form or syntactic knowledge, the
young subject possessing the left-brain
half has been reliably superior to the two
with language development mediated by
a right hemisphere.
The various studies of early hemidecor-
ticate language have revealed hints that
the two hemispheres might be differently
38738
‘TOPICS IN LANGUAGE DISORDERS / DECEMBER 1986
However, when the derivation of
meaning is dependent upon
responsiveness to language form or
syntactic knowledge, the young
subject possessing the left-brain
half has been reliably superior to
the two with language development
mediated by the right hemisphere.
skilled in the use of pronouns to achieve
text cohesion. For instance, the ability to
integrate both semantic (the features of
person, number, and gender) and syntac-
tic (the grammatical category features of
pronoun and case assignment) pronoun
information was more poorly developed in
children with early left hemidecortication
than in the right hemidecorticate. The
isolated right hemisphere was less adept at
manipulating the reference relationships
of pronouns in short written texts (Dennis
& Whitaker, 1976). In order to analyze in
more detail and with more formal tools
how each hemisphere used pronouns in
longer textual narratives, a recent study
was conducted (Lovett, Dennis, & New-
man, 1986) on the narrative discourse
skills of the hemidecorticate subjects who
were of normal intelligence (IQ 90-99]
and aged 12 to 14 at the time of the
study,
A STUDY OF PRONOUN COHESION
IN NARRATIVE DISCOURSE
The study used four texts (Little Red
Riding Hood, The Frog Prince, The Prac-
tical Princess, and Goldilocks) of folk or
fairy tales. The hemidecorticate subjects
listened to each story read by the exam-
iner and then were required to retell it
using story puppets if they chose.
Analyses of the story protocols indicated
that the complexity of episode structures
differed among the children. The right
hemidecorticate (MW) produced stories
that were different from those of the left
hemidecorticates (SM and CA) who were,
in addition, different from each other. For
instance, MW produced stories that devel-
oped more logically from setting through
conflicts to resolution, while left hemide-
corticate CA began many episodes that
she failed to resolve. The overall impres-
sion of listening to a story retold by MW is
one both of clear comprehension and
sophisticated language use, The left hemi-
decorticate subjects, in contrast, told sto-
ries that were less complex (though still
recognizable as the target stories) and sub-
jectively seemed to be more difficult to
listen to and comprehend.
The goal of the cohesion analyses was to
devise a systematic means of making
objective these perceived differences in
the comprehensibility of the story retell-
ings. In particular, an attempt was made
to discover how some of the differences in
story organization related to differences in
the nature of the relations between text
units in the stories,
The following excerpts will illustrate
the kind of textual issues that emerge from
the story retellings that were the focus of
the data analyses. The story is the classic
fairy tale, The Frog Prince. In this epi-
sode, the princess heroine reluctantly fol-
lows through on her promise to feed and
house a frog who had earlier retrieved her
golden ball from the bottom of a pond:
MW: And he started eating off the plate with
her. And after dinner, he said, “Wi you-willyou please carry me up to your bed?” So she,
she did. She laid him on the bed, hoping that
by the morning he would be gone.
SM: But she kept her promise and she told the
frog to come in and eat dinner with her and
sleep in her bed, And that night the princess
hoped that the frog wouldn't be in her bed
the next morning.
CA: So the-then the princess remembered the
promise that what she made to him. So they
brought into the house and they started to
eat, And-and then after he was tired. So the
princess took the frog up to her room and
then-um they went to sleep. And then the
next day and-then she thought the frog
would be gone the next day.
The analysis of the data followed two
distinct paths. The first path studied pro-
noun use in relation to such issues as: the
types of pronouns used and their distribu-
tional features; how pronouns and explic-
itly stated referents co-occur; the function
of pronouns in conveying new, old, and
presupposed information; and the role
that pronouns play in assigning textual
speech roles. This path originated with
Halliday and Hasan’s (1976) work on
cohesion, although their scoring system
was modified extensively in the course of
applying it to these texts. In the second
path, the dynamic interplay between pro-
nouns and referents throughout the story
was represented in picture rebus form.
THE USE OF PRONOUNS IN STORY
RETELLING
All of the hemidecorticate subjects
made use of text cohesion in their story
retelling. They produced a broad range of
pronouns, and their pronouns were gener-
ally correct. Each was able to use cohesion
to some degree to indicate the textual
DISCOURSE ANALYSIS IN NEUROLINGUISTICS
properties of stories. At some level, each
succeeded in producing a narrative dis-
course that would properly be termed
cohesive. But there were differences, as
well as similarities, between the hemide-
corticate subjects. They did not achieve
the same type or level of pronominal
cohesion in their texts, and both individual
and hemispheric differences were appar-
ent in the patterns of pronoun use and in
general narrative strategy.
The pronouns produced by the three
subjects were almost exclusively anaphoric
(referring backward), and were primarily
restricted to endophoric (within-text)
instances. The perceived differences in
comprehensibility did not appear to be
due to simple quantitative differences in
the use of particular types of pronouns
(e.g, nominative, possessive, relative).
The proportional distribution of pronoun
types across story retellings closely paral-
Jeled the distribution found in the stimulus
stories for all three subjects (Lovett, Den-
nis, & Newman, 1986).
The first cohesion analysis demon-
strated that although the three children
referred to virtually identical numbers of
characters in the stories, their means for
doing so differed considerably. MW (the
right hemidecorticate) pronominalized
referents more frequently than the left
hemidecorticate children, as is evident in
the sample from his retelling, in which he
used pronouns exclusively. In contrast, SM
frequently reintroduced referents rather
than using the appropriate pronouns. This
difference is verified by the relative fre-
quencies with which each child selected
pronouns over explicit restatement of a
referent. MW had the highest ratio of
Pronoun use to referent mention (1.85
38940
‘TOPICS IN LANGUAGE DISORDERS / DECEMBER 1986
pronouns per referent mention), followed
by CA (1.65); and SM (1.18) used pro-
nouns and referents in almost a one-to-one
ratio. If the stimulus texts’ average of 1.65
can be considered a rough estimate of the
number of pronouns per referent in narra-
tive texts, MW used slightly more pro-
nouns than he heard, CA matched the
texts, and SM was far below baseline.
Not only did SM use fewer pronouns
per referent than both the other children
and the texts, but he also used the simplest
form of anaphoric tie, most frequently
linking pronouns directly to their refer-
ents, In contrast MW, and toa lesser extent
CA, sustained a number of indirect, or
‘mediated links between pronouns and
their referents. In instances of mediation,
the listener must find the referent for the
pronoun by noting that it is co-referential
with another pronoun that is either
directly linked to the referent or is itself
mediated. MW’s overwhelming prefer-
ence for mediated rather than direct links
(by more than a two-to-one ratio) indi-
cates that he was able to keep a particular
referent “on stage” (and presumably in
the listener's short-term memory) through
a chain of anaphoric reference to it. By
comparison, SM (who had slightly more
direct links than mediated links) was con-
siderably less proficient at introducing
and maintaining a discourse element by
using the referential system. As a result,
his stories have a disjointed and poorly
connected referential flow.
Measurement of the physical dimen-
sions of the pronoun-referent relationship
further illuminated the differential abili-
ties of the children to sustain cohesive ties
in their discourses. Lovett, Dennis, &
‘Newman (1986) counted the number of
Measurement of the physical
dimensions of the pronoun-referent
relationship further illuminated the
differential abilities of the children
to sustain cohesive ties in their
discourses.
propositions (independent clausal seg-
ments) intervening between a pronoun
and its referent (a measure of span) and
the number of individual elements in a
mediated chain (depth). Span indicates
how much of the text intervenes between
the pronoun and its referent, while depth
indicates the complexity of the link. In the
sample from CA's retellings, the use of her
in her room is directly linked to its refer-
ent, princess, The link between she in she
thought and the same referent is therefore
mediated through the pronoun her, since
both refer to the same character.
As might be expected, MW had both
the longest median span between pro-
nouns and referents (2.14 propositions)
and the greatest median depth to his
mediated chains (8.04 mediations be-
tween a referent and the last pronominal
reference to it). Both of these values were
somewhat higher than the corresponding
values for the stimulus texts. In contrast,
the median depths for both SM (2.28) and
CA (2.88) were less than that of MW, and
that of the stimulus texts (2.58). CA’s span
measure (1.88) was close to the stimulus
text value (which was 2.00), while SM’s
median span (1,28) was considerably less.
The results of these analyses indicate
why the narratives of MW sounded more
comprehensible and closer to an “ideal”discourse. Conceivably, the resulting dis-
course makes fewer processing demands
on the listener, thereby leaving him or her
more resources for dealing with the con-
tent. Paradoxically, MW’s use of mediated
links, although they require the listener to
reinstate the referent, may require less
processing than direct links because the
topic is clearly marked as given through
anaphoric referential ties. SM’s greater
reliance on direct links and lower rate of
pronominal use may force the listener to
set up separate representations for each
mention of a referent. As a result, his
stories would be more difficult to process.
In contrast, CA tended to use pronouns
more frequently, although she sometimes
failed to tie them clearly to explicit refer-
ents. Her stories, as a result, contained a
degree of pronoun clutter (Lovett, Dennis,
& Newman, 1986). Combined with her
greater tendency to produce aborted epi-
sodes, CA’s clutter may overtax the listen-
er, who must repeatedly discover referents
for superfluous pronouns only to realize
that the characters mentioned are often
subsequently (and confusingly) dropped.
Finally, MW was able to maintain
simultaneous story lines in his narratives
with several pronouns and referential rela-
tionships cohesively embedded. The use of
cohesive referential chains clearly furthers
this goal as it allows the listener to keep
track of separate story elements. MW also
made better use of anaphoric reference in
retelling dialogue from the stories: his use
of anaphora was equal in narrative and
dialogue, while the other two children
were more likely to use anaphora in the
narrative portions of the texts, The addi-
tional dialogue component probably con-
tributes to the more varied texture of
DISCOURSE ANALYSIS IN NEUROLINGUISTICS
MW’s stories and to the listener's subjec-
tive impression of their richer content.
Discourse analysis techniques are valu-
able tools for the language researcher and
the language clinician. Because they
objectify and quantify whether and how a
language sample adheres to normal (or
tacitly accepted) conventions of discourse,
the techniques have implications, not only
for clinical diagnostic purposes, but for the
broader study of language. At the same
time, the techniques provide the re-
searcher with an objective, rather than
subjective basis for assessment and a con-
crete point of departure from which to
plan remedial procedures.
The present study prompts two ques-
tions: one is concerned with how the data
should be interpreted in the context of
current views about the constituents of the
language system, the other is involved
with how the data bear on the question of
hemispheric asymmetry for pragmatic
linguistic functions.
Toa considerable extent, anaphoric ref-
erence depends on mastery of the syntac-
tic system. Because previous research has
identified an isolated left-hemisphere
superiority for oral and written syntax
(e.g., Dennis & Kohn, 1975; Dennis, Lov-
ett, & Wiegel-Crump, 1981; Dennis &
Whitaker, 1976), the findings for pronom-
inal reference might be considered as
another instance of the greater capability
of the left hemisphere for this component
of language. However, reference is some-
times treated as a form of pragmatic
knowledge (e.g., Foldi, Cicone, & Gard-
ner, 1983) and the differences observed in
4142,
‘TOPICS IN LANGUAGE DISORDERS / DECEMBER 1986
the present cohesion analyses were not due
to grammatically incorrect pronoun
usage, but rather to pragmatic inappro-
priateness. In order to determine the rela-
tive contributions of pragmatics and syn-
tax to narrative discourse in the two sides
of the brain, it will be necessary to exam-
ine further aspects of pragmatics in this
and in other brain-damaged populations,
perhaps including such functions as the
use of intonation to indicate given and
new information; sensitivity to distinctions
between direct and indirect speech acts;
and the ability to discern topic mainte-
nance and shift through variations in sur-
face and information structure.
A difference in pragmatic competence
between the two sides of the brain has
been postulated on the basis of studies of
left- and right-sided hemisphere brain
injury, usually cerebrovascular accidents.
The proficient performance of the right
hemidecorticate, MW, shows that the left
hemisphere in isolation is quite capable of
dealing with pragmatic information, as
indicated by the use of pronominal refer-
ence. MW’s intact left hemisphere handles
the problem of pronominal reference effi-
ciently and effectively by using minimal
repetition of the referent, relying instead
on previous mention (or the use of given
information) to indicate cohesion. Two
different strategies were shown by the two
children with an isolated right hemisphere
(supporting the previously described view
that the right hemisphere mediates certain
aspects of language less uniformly than
the left): excessive repetition of the refer-
ent (reintroduction of given information
as new) and excessive use of pronouns
without adequate links to referents (pre-
supposing that the information is given
without providing the “anchor”). In each
instance, the right-hemisphere speaker
may be considered to have violated to
some degree the Given-New Contract
between speakers and listeners (Clark &
Haviland, 1977). In contrast the child with
the isolated left hemisphere appeared to
abide by the tacit contract, thereby facili-
tating the listener’s task.
It is difficult to propose a simple inte-
gration of these results on pragmatics in
story retelling with the reports of pri-
marily non-cohesive pragmatic tests in
adult patients with right hemisphere
lesions. First, it is evident that sustaining
damage to a portion of a right hemisphere
structure or function at maturity will
result in a different type of brain organi-
zation from having acquired language
without a right hemisphere almost from
birth. Second, a full spectrum of studies of
pragmatic functions following brain dam-
age has yet to be conducted.
Indeed, the role of pragmatics in dis-
course processing is only beginning to be
investigated, and our understanding of it
is certainly incomplete. Much more work
remains to be done before a complete
model of discourse processing can be for-
mulated. In addition to the kinds of prag-
matic variables discussed throughout this
paper, such questions as genre and modal-
ity (Chafe, 1982) and sociolinguistic con-
cerns such as the relation between speaker
and listener (see Gumperz, 1982) will need
to be addressed. The results of such studies
will open a new and valuable perspective
on how speakers achieve—and sometimes
fail to achieve—their communicative
Goals.REFERENCES
DISCOURSE ANALYSIS IN NEUROLINGUISTICS
Austin, JL: (1962). How todo things with words. Oxford,
UK: Osford University Pres.
Blumstein, $, & Cooper, W-E. (1974). Hemispheric pro-
‘essing of intonation contours, Cortex, 10, 146-158.
Bolinger, DL, (1977), Meaning and form. London: Long-
Brown, G., & Yule, G, (1989). Discourse analysis. Cam-
bridge, U.K: Cambridge University Press.
Caramazza, A, Grober, E., Garvey, C, & Yates, J. (1977)
‘Comprehension of anaphoric pronouns. Journal of Ver-
bal Learning and Verbal Behavior, 16, 601-608.
Caramazza, A. & Gupta, S (1979) The roles of topicalisa
tion, parallel function and verb semantics in the inter
pretation of pronouns. Linguistics, 17, 497-518.
Caramazza, A., & Zutif, EB. (1976), Dissociation of
algorithmic and heuristic processes in language com-
prehension: Evidence from aphasia. Brain and Lan-
‘guage, 8, 572-582.
Chafe, W.L. (1976). Givenness, contrastivenes, definite
ness, subjects topics. and point of view. In C.N. Li (Ed),
Subject and topic. New York: Academic Press
Chafe, W.L. (1982). Integration and involvement in
speaking, writing, and oral literature. In D. Tannen
(64), Spoken and written language: Exploring ality
and literacy. Norwood, NJ: Ablex.
(Clark, HLH, & Haviland, SE. (1977). Comprehension and
the given-new contract. In R.O, Freedle (Ed,), Dis-
course production and comprehension, Norwood, NJ
bles.
Clark, H.HL, & Sengul, CJ. (1878). In search of referents
for nouns and pronouns. Memory & Cognition, 7,
85-41
‘Cutler, A. (1976). Phoneme-monitoring reaction time asa
function of preceding intonation contour. Perception &
Peychophyses, 20, 85-60
de Beaugrande, R., & Dressler, W. (1981). Introduction to
{ext linguistics. London: Longman
Danes, F, (1960) Sentence intonation from a functional
point of view. Word, 1, 34-54,
Dennis, M. (1980). Language acquisition ina single hen
sphere: Semantic organization. In D. Caplan (Ed),
Biological studies of mental processes. Cambridge,
MA: MIT Press.
Dennis, M (1983). Syntax in brain-injured children. In M.
‘Studdert-Kennedy (Ed), Prychobiology of language.
Cambridge, MA: MIT Pres
Dennis, M., & Kol, B. (1975). Comprehension of syntax
in infantile hemiplegies after cerebral hemidecottics-
tion: Left hemisphere superiority. Bratn and Language,
2, 472-488.
Dennis, M., Lovett, M., & WiegelCrump, C.A. (1981)
‘Written language acquisition aftr let or right hemide-
corticaton in infaney. Brain and Language, 12, 54-91.
Dennis, M,, & Whitaker, H.A. (1976). Language aequis-
tion following hemidecortication: Linguistic superiority
of the left over the right hemisphere. Brain and Lan-
‘guage, 3, 404-435,
Engel-Orlieb,D. (1981), Discourse processing in aphases,
Text, 4, 361-383
Foldi, NS, Cicone, M, & Carder, H, (1989). Pragmatic
aspects of communication in brain-damaged patients.
In 8, Segalowitz (Ed), Language functions and lan-
‘guage organization. New York: Academic Pres.
Gardner, H. (1876). The shattered mind. New York:
Vintage Books
Garvey, C, & Caramazza, A. (1974), Impliet causality in
verbs. Lingulstic Inquiry, 5, 459-464,
Geschwind, N. (1967). The varieties of naming errors
Cortex, 3, 97-112.
Goodglass, ., & Baker, E (1976),Semantc eld, naming,
tnd auditory comprehension in aphasia, Brain and
Language, 3, 59-514
Grice, H.P. (1975). Logie and conversation. In P. Cole &
JUL Morgan (Eds), Syntax and semantis 8: Speech
facts, New York: Academie Pres
Grober, E, Pereeman, E, Kellar, L., & Brown, J (1980).
Lexical knowledge in anterior and posterior aphasies
Brain and Language, 10, 318-890.
Gumpera, J. (1982). Discourse strategies, Cambridge,
U.K. Cambridge University Press.
Halliday, M.A.K. (1970). Language structure and lan-
juage function. In J. Lyons (Ed.), New horizons in
Iinguistics. Harmondsworth, U.K.: Penguin Books.
Halliday, M.A.K. (1985). An introduction to functional
‘grammar. London: Edward Arnold.
Malliday, MAK, & Hasan, R. (1976), Cohesion in
English, London: Longman,
Hirst, W., & Brill, G.A. (1980). Contextual aspects of
pronoun assignment. Journal of Verbal Learning and
Verbal Behavior 19, 168-175
Hoffman, H.J., Hendrick, E.B., Dennis, M., & Armstrong,
, (1979), Hemispherectomy for Sturge-Weber syn
drome. Chila's Brain, 5, 253-248,
Lovett, M, Dennis M, & Newman, J.E (in press). Making
referents: The cohesive use of pronouns inthe narrative
discourse of hemidecorticate adolescents. Brain and
Language.
Mandl, JM. & Johnson, NS. (1977). Remembrance of
things parsed: Story structure and recall, Cognitive
Poychology, 8, 111-15.
4344
‘TOPICS IN LANGUAGE DISORDERS / DECEMBER 1986
MeKoon, G,, & Ratelif, R. (1980). Pri
recognition: The organization of propositions in mem-
cory for text. Journal of Verbal Learning and Verbal
Behavior, 19, 369-886.
Millar, J.M., & Whitaker, HLA. (1983). The right hemi
sphefe’s contribution to language: A review of the
evidence for brain-damaged subjects. In S, Segalowitz
(Ed), Language functions and language organtzatton
(pp. 87-118), New York: Academie Press,
Newman, JE. (1981). Explorations of the function of
‘emphasis in connected speech. Unpublished doctoral
dissertation, University of Toronto.
Prince, E-F. (1981). Toward a taxonomy of given-new
information. In P. Cole (Ed), Radical pragmatics. New
York: Academic Pres.
Rochester,S, & Martin, J. (1978). Crazy talk: A study of.
in item
the discourse of schtzophrento speakers. New York:
Plenum Pres.
Searle, .R. (1969). Speech acts. Cambridge, U.K Cam-
bridge University Pres.
Tucker, DM., Watson, RT, & Hellman, KIM, (1977).
Discrimination and evocation of affectively fatoned
speech in patents with right parietal disease, Neurolo
ay, 27, 947-050.
Ulatowske, HK, Doyel, A.W., Stern, RLF, & Haynes,
SM. (1983). Production of ‘procedural discourse in
aphasia, Bran and Language, 18, 315-541.
Weintraub, §, Mesulam, M.M. & Kramer, L. (1981)
Disturbances in prosody: A right-hemisphere conteibu-
tion to language. Archives of Neurology, 98, 742-744
Zurif, EB, Caramazza, A, & Myerson, R. (1672). Gram-
matical jadgmonts of agrammatio aphasies, Neuropsy-
chologia, 10, 405-417