.I. Neurolmgursrics, Volume 3. Number 2, pp. 161-183, 1988. 091 f-6044/88 $3.00 + .OO Prmted in Great Britain 0 1988 Pergamon Press plc Assigning Linguistic Roles: Sentence Interpretation in Normal and Aphasic Kannada-English Bilinguals Jyotsna Vaid Department of Psychoiogq Texas A & M University Shyamala Chengappa Department of Speech Pathology All India Institute of Speech and Hearing ABSTRACT In interpreting a sentence, listeners rely on a variety of linguistic cues to assign grammatical roles such as agent and patient. In normal sentence comprehension these cues converge to enable sentence interpretation, yet when the cues are placed in competttion they are differentially used by speakers. The present study investigated the relative strength of three cues to agenthood - word order, noun animacy and subject-verb agreement - in normal and aphasic Kannada-English bilinguals and Kannada monolingual controls. The findings are discussed with respect to other cross- linguistic evidence using the sentence interpretation paradigm and with respect to their bearing on theories of bilingual language representation. In interpreting a sentence we rely on a variety of surface linguistic cues ranging from phonological stress, morphological markers and syntactic-semantic cues such as word order and noun animacy. The information conveyed by these cues allows us to identify the topic of a sentence, to distinguish between given and new information and to assign grammatical roles. In acquiring a first or a second language, a language user must learn what kinds of information are conveyed by particular linguistic cues present in the surface form. The mapping between form and function is rarely one-to-one for a single form can map to several functions and several forms can map to the same function. For example, in English, cues such as word order, pronoun case inflection and noun animacy all convey information about grammatical roles. Although more than one
Sentence Interpretation in Normal and Aphasic Kannada-English
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
.I. Neurolmgursrics, Volume 3. Number 2, pp. 161-183, 1988. 091 f-6044/88 $3.00 + .OO Prmted in Great Britain 0 1988 Pergamon Press plc
Assigning Linguistic Roles: Sentence Interpretation in Normal and Aphasic
Department of Psychoiogq Texas A & M University
Department of Speech Pathology All India Institute of Speech
In interpreting a sentence, listeners rely on a variety of linguistic cues to assign grammatical roles such as agent and patient. In normal sentence comprehension these cues converge to enable sentence interpretation, yet when the cues are placed in competttion they are differentially used by speakers. The present study investigated the relative strength of three cues to agenthood - word order, noun animacy and subject-verb agreement - in normal and aphasic Kannada-English bilinguals and Kannada monolingual controls. The findings are discussed with respect to other cross- linguistic evidence using the sentence interpretation paradigm and with respect to their bearing on theories of bilingual language representation.
In interpreting a sentence we rely on a variety of surface linguistic cues ranging
from phonological stress, morphological markers and syntactic-semantic cues
such as word order and noun animacy. The information conveyed by these cues
allows us to identify the topic of a sentence, to distinguish between given and new
information and to assign grammatical roles. In acquiring a first or a second
language, a language user must learn what kinds of information are conveyed by
particular linguistic cues present in the surface form.
The mapping between form and function is rarely one-to-one for a single form
can map to several functions and several forms can map to the same function. For
example, in English, cues such as word order, pronoun case inflection and noun
animacy all convey information about grammatical roles. Although more than one
162 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
cue conveys information relevant to the same comprehension process, the cues vary
in how well they convey this information. The accuracy with which a cue conveys
information is known as its validit),. Cue validity may be viewed as the product ofa
cue’s availability and reliability. Availability is a measure of how often a cue is
present, reliability is a measure of how often a cue points to the correct interpre-
tation when it is present. Cue validity can be calculated from adult language input
to the language learner (cf. McDonald 1984). While overall cue validity predicts
order of acquisition, conflict validity (see McDonald 1987) predicts order of adult
strength of usage.
The concept of cue validity has been incorporated into a model of sentence
comprehension developed by Bates and MacWhinney (1982. 1987). This model,
known as the Competition Model. offers a probabilistic account of how cues
combine during comprehension. ln this model. each cue is accorded a strength or
weighting proportional to its validity. This weighting in turn determines the
amount of activation of a particular interpretation. If two cues agree as to an
interpretation, their strengths are added, leading to a greater activation of that
interpretation compared to when a single cue is present. If they disagree. the
interpretation with the highest activation level is chosen. In this way, cues are
thought to cooperate and compete in the comprehension process.
SENTENCE INTERPRETATION PARADIGM
The notion of cue validity and its psychological counterpart cue strength has
been empirically examined by MacWhinney et al. (1984) in cross-linguistic studies
of sentence processing. The basic paradigm in this research involves a sentence
interpretation task where native speakers of different languages are presented with
simple, transitive sentences containing two concrete nouns and a concrete action
verb and are asked to identify the agent - that is, who performed the action
described in the sentence. While in normal discourse various cues arecorrelated, in
the sentence interpretation paradigm the sentences always represent orthogonal
combinations of lexical-semantic, pragmatic, grammatical and/ or phonological
cues, thereby allowing for an independent examination of the relative contribution
of each of these types of information. The result of selecting sentences in this way is
that a number of semi-grammatical sentences are included along with grammatical
sentences. Nevertheless. their syntactic or semantic oddness does not appear to
affect how they are processed as previous research with this paradigm has shown.
Bilingual Sentence Interpretation 163
For example, Hungarian speakers comprehend ungrammatical sentences in which
the obligatory case markings have been deleted in the same way in which they
process grammatical sentences where the omission of this case marking is allowable
(MacWhinney et al. 1985).
In some of the sentences in this task the various cues converge to a single decision
about who did the action while in other sentences one or all of the cues may conflict.
An example where the cues of word order, subject-verb agreement and noun
animacy conflict is in the sentence “The dog lick the pencils”. Standard subject- - verb ~ object (SVO) word order in English would assign the preverbal noun dog
as actor as would the animacy cue, but noun-verb agreement would favor the plural
noun pencils, which agrees in number with the plural verb form. In such cases
native speakers of English rely most on word order, discardingpencils as the actor
despite the agreement cue. Italian speakers, however, tend to choosepencils as the
agent, for agreement cues are stronger than word order cues in Italian. For
German speakers, animacy and agreement turn out to be stronger cues than word
order and for Hungarian and Serbo-Croation speakers case is a stronger cue than
word order. In short, adult native speakers interpret sentences by making preferen-
tial use of the cue that is the most valid (or informative) in their language. Thus, the
relative ranking of cues may often differ across different languages, even when the
languages on a formal level are typologically similar (e.g. Italian and English).
Developmental studies using the sentence interpretation paradigm to study
children as young as two years of age indicate that cue validity based on adult usage
is a strong predictor of the developmental sequence of acquisition of the cues.
Word order (SVO) is the first cue to emerge in English-speaking children and
nominative/accusative case marking is the first to emerge in Turkish children
(Bates and MacWhinney 1987). Within the Competition Model framework, lan-
guage learning is viewed as a process of incrementing and adjusting the weights of
form-function mappings until there is an optimal fit with the processing environ-
ment (MacWhinney 1987).
STUDIES WITH BRAIN-INJURED INDIVIDUALS
Given that there are differences in how normal listeners monitor different cues in
the input, one might expect differences in cue validity to be reflected in the pattern
of language impairment of brain-damaged individuals. In fact, language-specific
differences in cue strength appear to be preserved in aphasics (see Bates et al. 1987).
In general, the cue that has the highest strength in premorbid language use appears
to be the most resistant to impairment following brain damage. At the same time,
there appears to be a global impairment of certain cues (e.g. subject-verb agree-
ment) regardless of their premorbid status in a particular language, and a selective
164 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
sparing of other cues. e.g. word order (see Bates et al. in press). Research on
aphasics using the sentence interpretation paradigm is still in its infancy.
SENTENCE INTERPRETATION IN BILINGUALS
A recent extension of the sentence interpretation paradigm has involv,ed a
comparison of sentence interpretation strategies used by bilinguals or second
language learners in their two languages. This topic has thus far been addressed in
users of predominantly Indo-European languages including Spanish (Wulfeck 6’1
al. 1986) and German (Bates and MacWhinney 1981: McDonald 1984). but a few
other language families are also being studied. typically with English being one of
the two languages under study. These other languages include Dutch(Kilborn 1986:
Kilborn and Cooreman 1987). Korean (Park 1986), Japanese (Kilborn and Ito
1987) and Hindi (Vaid et (11. 1987).
On a theoretical level. the sentence interpretation task in bilinguals 15 of interest
in that it allows a test of different hypothesized views about bilingual language
representation. Most previous work in this area has concentrated on the lexical
level, typically using lists of unrelated words. The sentence interpretation task
allows for an examination of higher-level language comprehension strategies in
POSSIBLE OUTCOMES IN BILINGLIALS
A priori. one can distinguish between four different possible outcomes on
the task, reflecting different forms of internal representation of bilingual discourse.
Assuming language-specific differences on the task (as reflected in different
strategies used by monolingual speakers of the different languages). one possible
outcome for bilinguals on this task may be a transfer of first language strategies to
the second language. This outcome would be expected for individuals at early
stages of second language exposure and/or proficiency and would support a
“subordinate” form of bilingual language organization. where the first language
cues remain dominant and the second language is processed in terms of cues that
are salient in the first language (see McDonald 1987). A second possible outcome
is one in which the bilingual shows a different pattern in each language, correspon-
ding to the ranking of cues of monolingual speakers of the respective languages
This kind of response would support a”coordinate” form of representation where
distinct sentence comprehension strategies are used for the two languages. A third
possible outcome is where bilinguals show a unified strategy in each 01 their
languages. but this strategy differs from that used by monolingual speakers of the
Bilingual Sentence Interpretation 165
languages and represents an amalgam or merger of cues that are dominant in each
of the languages. Such an outcome would be consistent with a “compound” form of
representation (see Wulfeck et al. 1986). Finally, it is possible that bilinguals may
show a reverse subordinate pattern such that their performance on their second
language is comparable to that of monolingual speakers of that language and their
performance on their native language is influenced by their second language. Such
an outcome might be expected to arise if the bilinguals are more practised in their
second language than in their first. These four hypothesized outcomes are not
mutually exclusive nor are they rigid over time. Rather, at different stages of a
bilingual’s language history, different outcomes may be observed. For example,
McDonald (1984) found that L2 speakers with little exposure to the second
language use Ll strengths; with increasing L2 exposure the L2 speakers shift to
appropriate L2 weights. The outcomes should thus be viewed merely as a useful
means of classification.
In what follows, sentence interpretation data are presented from adult bilingual
speakers of Kannada and English. A language from the Dravidian language family
used primarily in south India, Kannada is spoken by over five million speakers.
Sentence interpretation data are presented from a group of brain-intact Kannada
monolingual adults and Kannada-English bilinguals and from a smaller group of
aphasic counterparts. To date there has been only one published study of sentence
interpretation in bilingual aphasics (Wulfeck ef al. 1986).
RELEVANT CHARACTERISTICS OF KANNADA GRAMMAR
Some remarks are in order at this point about the characteristics of Kannada of
relevance to the sentence interpretation task. First. Kannada is a case-marked
language. The case system of Kannada is similar to that of other Dravidian
languages in that various suffixes are added to the noun stem to indicate different
relationships between the noun and the other sentence constituents. The cases
include the nominative, genitive, dative, locative, instrumental; ablative, vocative
and accusative (Schiffman 1979).
The basic colloquial accusative marker is - anna/ - annu. It is used with both
“rational” (capable of thought) and “non-rational” referents (usually animals,
young children and inanimate objects). It is also possible for the accusative markers
to be omitted entirely except in the case of rational (masculine and feminme) nouns.
Case markers used with plural nouns are usually the same as those used with the
singular, but often in colloquial Kannada the plural marker -galu is not used, 50
the singular can be found even when more than one object is meant. In previous
sentence interpretation studies performed on other case-marked languages (e.g.
Serbo-Croation), case has emerged as a dominant cue. It is likely to play a strong
166 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
role in Kannada sentence interpretation as well. Since our interest was In com-
paring Kannada and English on cues that both languages possess to a comparable
degree, it was decided to delete the accusative case marker from the Kannada
stimuli so as to render the task of interpretation at least as ambiguous as that in
English. The cues chosen for study in the two languages. then. were noun animac!;.
noun-verb agreement and word order.
Word order may perhaps be a weaker cue in Kannada than in English since
Kannada permits more variation in word order, given that linguistic roles are
normally signalled by case markers. Moreover, Kannada uses a subject-object
verb (SOV) word order. VSO or VOS word orders are stylistic variations referred
to as “after thought” word orders used in colloquial speech (Schiffman 1979). The
nature of the word order effect might be expected to differ in Kannada and English.
a SVO language.
The subject in Kannada sentences agrees with the verb in number and gender. We
would thus expect the agreement cue to be at least as strong as that in English and
perhaps stronger since non-verb agreement in English is only on the basl5 of
number. There is no a priori basis for expecting a difference in the strength of the
animacy cue in the two languages. In neither language IS noun ammacy overtI>,
marked; rather it 1s one of the semantic properties associated with a noun.
Sentence Interpretation in Brain-intact Kannada Speakers
In order to understand Kannada-English bilinguals performance on the sentence
interpretation task, it is important to understand the cues used by monolingual
Kannada speakers. Thus both groups were tested.
Subjects included a group of 8 Kannada monolinguals and 21 Kannada-Enghsh
bilinguals. All were right-handed. The monolingual sample included 2 males and 6
females ranging in age from 21 to 60 years with a mean age of 40 years. The bilingual
sample ranged in age from 20 to 62 years with a mean age of 3 1 years and included 6
females and 15 males.
The bilinguals’ mother tongue was Kannada in all cases. They had studied
English in school until their second year of college and had been enrolled in
English-medium schools throughout their education. Many subjects also received
post-graduate training in English. As such, the sample was highly fluent in English.
A brief questionnaire administered to the bilinguals revealed that the majority of
Bilingual Sentence Interpretation 161
subjects reported using English at work and with friends and in reading the
newspaper and writing letters; Kannada was used primarily at home and with
Stimuli For each language, a total of 54 sentences were constructed. The stimuli were all
simple, active, declarative sentences with two concrete nouns and a transitive action
verb. All three word orders (NVN, VNN and NNV) were tested in orthogonal
combinations of noun animacy (both nouns are animate, yielding semantically-
reversible sentences; only the first noun is animate; only the second noun is
animate) and noun-verb agreement (the verb agrees with both nouns in number;
the verb agrees only with the first noun; the verb agrees only with the second noun).
The 3 X 3 X 3 factorial design resulted in a total of 27 sentences. Two examples of
each sentence type were included, yielding a total of 54 sentences. To minimize the
effect of extralinguistic, real world knowledge on sentence comprehension, the
stimuli were prepared using a random assignment of animate and inanimate nouns
to the designated slots in the sentences. While the Kannada and English sentences
were not translation equivalents, they were constructed from the same pool of
words. The set of nouns and verbs used to generate the stimuli in each language is
provided in the Appendix.
Subjects were all tested individually. The sentences were read out loud one at a
time in a neutral intonation by the examiner (SC) who is a native speaker of
Kannada and fluent in English. Upon hearing each sentence, subjects were to indi-
cate which of the two nouns - the first or the second - had performed the action
described in the sentence. Practice trials were given to ensure that subjects under-
stood the task. The order of testing for the bilinguals was alternated across subjects
such that half were tested in Kannada first followed by English while the remainder
were first tested in English. A gap of at least a day separated the two testing sessions.
If the first preverbal noun mentioned was chosen as the actor, subjects received a
score of 1. Choice of the second noun was assigned a score of0. Since there were two
sentences with each possible combination of the three factors, the maximum score
possible was 2, indicating that subjects chose the first noun both times for that
sentence type. The dependent variable was the percentage choice of the first noun as
agent. Chance performance would be designated by 50%; 100% would indicate a
choice of the first noun on every item and 0% a choice of the second noun.
168 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
Analyses of variance were performed on the percentage of lirst preverbal nouns
chosen as the agent for each sentence type. The monolingual data were analyzed
first tn a separate analysis of vartance and were subsequently compared with the
bilinguals’ Kannada data tn a combined analysts of vartancc. The bilinguals’ data
were further analyzed to compare performance tn their two languages.
Mean scores ranging from 0 to 2 on the sentence interpretation test were entered
into a 3 X 3 X 3 analysis of variance as a function of noun animacy, word order and
verb agreement. There were two significant effects: a main effect of animacy (pc
.OOOl) and an animacy X agreement interaction (pc .OOl). The animacy effect
accounted for 84c7, of the variance while the two-way interaction accounted for an
additional 8qc of the vartance. Table I summarizes the ANOVA results for the
Summary of ANOVA of
Kannada Normal Monolinguals (n=8)
Source df F P
Animacy 2, 14 80.63 ,000 1*
Agreement 2, 14 3.12 ,076 ns
Word order 2. 14 0.14 ,672 ns
Animacy X agreement 4, 28 5.97 .001*
Animacy X order 4, 28 0.85 .507 ns
Agreement X order 4. 28 0.94 456 ns
Animacy X agreement X order 8. 56 0.79 ,612 ns
ns, not significant
The main effect of animacy indicated that the percentage choice of the first noun
as agent was much higher when the first noun m the sentence was animate and the
second noun was inanimate (88Yc) than when both nouns were animate (6357’) or
only the second noun was animate (254). The interaction of animacy and agree-
ment revealed that verb agreement was used as a cue only when animacy cues were
neutral or absent. as when both nouns in the sentence are animate. In this
condition, the percentage choice of the first noun was 794% when the first noun
agreed in number with the verb, whereas tt fell to 40 c?c when the second noun agreed
Bilingual Sentence Interpretation 169
with the verb. With animacy cues present, the agreement effect was either greatly
diminished as in the animate-inanimate condition (compare 94% vs. 81% first-
noun choice in the first-noun-agreement and second-noun-agreement conditions,
respectively) or was completely absent, as in the inanimate-animate condition
(where the percentage choice of first noun remained a low 29% for first-noun and
The word order effect was not significant in Kannada: subjects showed a
consistent preference of the first noun (around 6OYo) for all three word orders.
Figure 1 illustrates the performance of the Kannada monolinguals relative to
that of English monolinguals tested in previous studies by Bates and colleagues.
Note that while the agreement effect is weak in both language groups, the animacy
effect is particularly strong in Kannada relative to English while the reverse is true
for the word order effect.
NVN VNN NNV
‘7 I 1
I \ \ d’ 1 \ I I \ 4 I b
I I I
An ~macy Agreement
Figure 1. Kannada vs. English Monolinguals: Main Effects of Word Order,
Animacy and Agreement.
Kannada Monolinguals vs. Kannada-English Bilinguals on Kannada
A 2 X 3 X 3 X 3 factorial analysis of variance was performed on the Kannada data
170 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
of monolinguals and bilinguals. All three within-subject variables (animacy,
agreement and word order) were highly significant, as were two two-way mter-
actions: animacy X agreement (p < .OOOl) and animacyX word order(p .:: .002). It
is noteworthy that there were no significant bilingual/monolingual group dif-
ferences nor were there any significant interactions involving the group variable.
Kannada-English Bilinguals on Kannada vs. English
A four-way analysis of variance (animacy by agreement by word order by
language) was performed on the bilinguals’data to compare their performance on
their two languages. A summary of the ANOVA results is presented in Table 2.
The analysis indicated significant main effects of animacy, agreement and word
order as well as 2 two-way interactions: animacy X agreement (p c: .0009) and
Summary of ANOVA for Kannada-English Normal Bilinguals (n = 21)
Source df F P
Animacy X agreement
Animacy X order
Agreement X order
Animacy X language
Agreement X language
Order X language
Animacy X agreement X language
Animacy X order X language
Agreement X order X language
Animacy X agreement X order
Animacy X agreement X order X language
2.40 168.95 .0001*
2.40 16.65 .0001*
2,40 8.55 .0008*
1,20 1.63 .22 ns
4,80 5.19 .0009*
4,80 6.76 .0001*
4.80 0.31 .87 ns
2,40 0.53 .59 ns
2.40 0.53 .59 ns
2.40 1.04 .36 ns
4,80 4.80 .002*
4,80 1.38 .25 ns
4,80 3.04 .02*
8,160 3.26 .002*
8.160 1.81 .08 ns
ns. not significant
Bilingual Sentence Interpretation 171
animacy X order (p < .OOOl). In addition, there were three higher order interactions:
animacy X agreement X order (p < .002) and language X animacy X agreement
(p < .02). Figure 2 illustrates the three main effects in each language.
The main effect of animacy accounted for the greatest proportion of the variance
(82%), while the main effect of agreement accounted for 8% and the interaction
effects together constituted about 1070 of the variance.
5 fro- e ‘tB, --Cl
‘1 .E E 40-
0 I I I NVN VNN NNL
I I I AA Al IA
I I I
NJ0 Agl Ag2
Normal bilinguab ( n = 2 I)
Figure 2. Kannada-English Bilinguals: Effects of Word Order, Animacy and
Agreement by Language.
Let us examine the two higher-order interaction effects involving language. The
means for the language X animacy X agreement effect are summarized in Table 3.
Note first that in both Kannada and English the presence of an animate first noun
greatly increases the percentage choice of the first noun as agent while the presence
of an inanimate first noun greatly reduces it. When animacy cues are absent (when
both nouns in the sentence are animate) the choice of agent is governed by
agreement cues. These cues appear to be stronger for Kannada than for English.
172 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
Kannada-English Normal Bilinguals (n = 21):
Language X Animacy X Agreement
Interaction Means (in %)
AgO f%l AS AgO Agl AS
AA 72 86 47 59 68 53
Al 88 93 78 80 87 66
IA 8 24 12 16 36 20
The other interaction effect was language X agreement X word order. Table 4
provides a summary of the interaction means. Inspection of the table suggests that
the source of the interaction hes in the NNV condition: when agreement is with the
ftrst noun in this word order condition, the percentage choice of the ftrst noun 1s
much higher in Kannada (7O?e) than it is in English(56qe). The analogous effect for
NVN word order does not appear to be present. that is, the percentage choice of the
first noun is only marginally higher in English than in Kannada (74$$ vs. 68%)) when
agreement is with the first noun in the NVN condition.
Kannada-English Normal Bilinguals (n = 21):
Language X Word Order X Agreement
Interaction Means (in %)
AgO Ad A@ AgO Ad 49
NVN 61 68 55 60 74 52
VNN 53 65 44 48 60 46
NNV 54 70 40 47 56 42
Bilingual Sentence Interpretation 173
It would appear that animacy is the cue with the greatest cue strength in
Kannada at least among the cues tested, since case would presumably have been a
strong cue in Kannada as well. In monolinguals and bilinguals alike, the animacy
main effect accounted for by far the largest proportion of the variance. The cue used
next appears to be verb agreement, although its effect is evident only in the absence
of animacy cues. The weakest cue seems to be word order.
The performance of bilinguals on Kannada was indistinguishable from that of
Kannada monolinguals. While this finding may not be all that surprising given that
Kannada was the first language of the bilinguals, it is nevertheless interesting that
the bilinguals were not in any way influenced by their exposure to English, despite
their apparently high level of proficiency in English. As such, the present study does
not support hypothesized outcome 4 stated at the outset. One can speculate that the
word order cue was not used by Kannada-English bilinguals when processing
English because the English spoken in India has different structural properties than
the English spoken in America. While there may be some basis for this speculation
it should be pointed out that HindiiEnglish bilingual speakers do rely on word
order when processing English (see Vaid et al. 1987). Thus the observed lack of
reliance on word order in Kannada speakers would appear to suggest a language-
specific influence of Kannada, rather than sociolinguistic differences in the nature
of English used in India.
When we consider the bilinguals’ performance on Kannada vs. English, it is
noteworthy that across both languages animacy again accounted for the most
variance (over 80%). Word order was a relatively weak cue, even in English. In the
two instances where language differences were significant, the nature of the
difference was such that the effect (of agreement in one case and of word order in
another) was more pronounced in Kannada than in English. Specifically with
regard to the word order effect, while NVN sentences yielded a slight first-noun
preference in both Kannada and English, NNV sentences yielded a strong first-
noun preference in Kannada as compared to that in English. This difference reflects
the different structural properties of the two languages, inasmuch as Kannada is a
SOV language whereas the typical response of native English speakers to the non- canomcal NNV word order is to choose the second noun as agent (see MacWhinney
et al. 1984).
Thus, it would appear that the response of the bilinguals on this task supports the
first of the four hypothesized outcomes, namely, a transfer of first language
strategies onto the second language. The results do not support a situation where
bilinguals show a particular ranking of cues in one language and another ranking in
the other language. Rather, their performance on Kannada is almost identical to
174 Journal of Neurolinguistics. Volume 3, Number 2 (1988)
that of Kannada monolinguals. while their performance on English is a muted
version of their Kannada-based strategy and in no way resembles the performance
of English-speaking monolinguals. This outcome is somewhat surprising given the
high degree of English language proficiency reported by the bilinguals. The present
results suggest, therefore. that the sentence interpretation strategies observed need
not be related in any obvious way to the individual’s level of proficiency. Alter-
natively, it may be that. at the level of discourse processing, proficiency may mean
Sentence Interpretation in Aphasic Kannada Speakers
In this section of the paper, we summarize the findings from the sentence
interpretation task administered to a group of left-hemisphere damaged native
Two monolingual Kannada aphasics, S.A. and B.B., were tested at the National
Institute of Mental Health and Neurological Sciences (NIMHANS) in Bangalore.
Both were patients at the Speech Pathology unit at NIMHANS where, on the basrs
of a standard aphasia test battery adapted to Kannada. they were diagnosed as
being receptive aphasics. In addition, three Kannada-English aphasic bihnguals
who were also patients at NIMHANS were tested. Two of these patients( K. K. and
J. P.) had been diagnosed as expressive aphasics and the third (N.D.) was a
receptive aphasic. While it was not possible to obtain a detailed premorbid
language use inventory on the bilingual patients, a brief interview with the patients’
immediate family confirmed that the bilingual patients’ language background and
premorbid language use was not all that different from that reported for the
bilingual controls in Study 1.
Criteria used to exclude patients from participation in this study included the
following: (a) a history of multiple strokes; (b) significant hearing and/or visual
deficits; (c) severe gross motor disabilities; (d) severe motor speech involvements
such that speech was all but unintelligible; and (e) evidence of neurological
instability and/ or testing less than 3 months post-onset of insult.
Both monolingual patients S. A. and B. B. spoke Kannada as a mother tongue
and had Kannada as the medium of instruction m school through graduation
S.A. had suffered a post-traumatic cerebral contusion and an acute subdural
Bilingual Sentence Interpretation 175
hematoma in the left temporal lobe following a traffic accident. On the basis of a
Kannada adaptation of the Western Aphasia Battery, S.A. was diagnosed as
having Wernicke’s asphasia. S,A.‘s language functions were tested 15 days post-
insult. At the time of testing, S.A. had a mild hemiparesis and aphasic symptoms.
B.B. was diagnosed on the basis of a CT scan as having had an infarct in the left
temporoparietal region. His language functions were tested four months post-
insult using the Western Aphasia Battery and, like S.A., he was also classified as a
Wernicke’s aphasic. At the time of testing one and a half years post-insult, B.B. had
almost recovered from a right hemiparesis but still had aphasic symptoms.
All three bilingual aphasics. J.P., K.K. and N.D., had studied Kannada and
English in school as their first and second language, respectively. From grade 10, all
three had English as the medium of instruction and used English in their work.
The etiology of J.P.‘s brain injury, as indicated by a CT scan, was a left cervical
internal carotid aneurysm with cerebral embolism. The other two patients were not
Background Information on Kannada Aphasics
Group Subjects’ Aphasia
Sex Age Hand Etiology
K mono B.B. Receptive M
K mono S.A. Receptive M
K-E bi N.D. Receptive M
K-E bi K.K. Expressive M
K-E bi J.P. Expressive M
R Left internal carotid
Right hemiparesis 1.5 years
R Left temporal lobe problems
due to left subdural
R Left internal carotid
R Left internal carotid
R Left internal carotid
176 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
advised to have a CT scan done and were diagnosed simply as having a left carotid
insufficiency, that is, inadequate blood supply to the cortical lobes by the middle
cerebral artery, perhaps as a result of a thrombosis or embolism. All three patients
had hemiplegia with aphasia. N.D. had almost recovered from the hemiplegia at the
time of testing.
Language testing of K.K. was done three months after the cerebral insult, one
month post-insult for J.P. and two years post-insult for N.D. Table 5 provides a
summary of the background characteristics of the patients.
Stimuli and Procedure
The same set of sentences used in Study 1 were administered to the brain-
damaged subjects. Sentences were read aloud to the patients by the same examiner
who had tested the normal subjects. Patients signalled their responses by pointing
to the appropriate noun on the stimulus sheet. Sufficient practice trials were given
to ensure that the patients understood the task. Bilinguals were tested in therr two
languages in separate sessions, with the order of the sessions being varred across
The data were coded as in Study I with the additional feature that a score of 5
was assigned on any trial on which subjects failed to make a response.
Performance of Kannada-English Bilingual Aphasics
(Percentage Choice of First Noun)
Animacy AA 84 44
AI 72 72
IA 61 16
Agreement AgO 61 56
AgJ 84 28
Ag2 72 50
Word order NVN 72 34
VNN 61 44
NNV 84 56
Bilingual Sentence Interpretation 177
Given the small sample size, an analysis of variance was not conducted. Instead,
the data from the two monolingual aphasics are individually summarized in Table
Inspection of the mean percentage choice of first-noun scores suggests that the
animacy cue was used by both subjects and was particularly dominant for B.B.
When the first noun of the sentence was animate. B. B. chose the first noun as agent
72% of the time as compared to 16% of the time when the first noun was inanimate.
Animacy may have been the only cue utilized systematically by B. B. inasmuch as
his performance on the other two factors appears to be essentially random.
S. A., on the other hand, showed some sensitivity to verb agreement, choosing
the first noun as agent 84% of the time when that noun agreed with the verb, as
compared to 61% of the time when agreement was ambiguous. S. A. also appeared
to favor a first-noun preference for sentences with the order NNV (84%) and to a
lesser extent for NVN sentences (72%).
Bilingual Aphasics on Kannada vs. English
Owing to the small sample size it was not advisable to conduct a statistical
Performance of Kannada-English Bilingual Aphasics
(Percentage Choice of First Noun)
N.D. J.P. K.K.
Animacy AA 39 72
AI 83 100
IA 0 6
Agreement AgO 61 56 83
Ag 1 33 67 67
Ag2 39 56 61
Word order NVN 50 67 67
VNN 44 61 67
NNV 39 50 78
178 Journal of Neurolinguistics, Volume 3. Number 2 (1988)
Animacy AA 61
Agreement AgO 39 61 56
Agl 61 50 50
Ag2 50 50 44
Word order NVN 44 56 50 VNN 56 50 44 NNV 61 50 56
Table 7 - Contd.
analysis of the data. The data from the three bilingual aphasics are therefore
summarized in Table 7.
In all three bilingual patients, the animacy cue appears to have been spared, in
both Kannada and English. The percentage choice of the first noun was generally
higher when the first noun was animate (the scores ranged from 83 to 100%) than
when the first noun was inanimate (the scores ranged from 0 to 3996).
By contrast, the agreement cue appears to be largely unavailable in either
language. for none of the subjects showed a clear main effect of agreement. Had
there been more subjects, it would have been possible to examine whether the
agreement cue is in fact present under certain conditions of the other two factors.
Similarly, word order does not appear to have been a salient cue for the bilingual
aphasics. particularly in English, where performance was essentially at chance level
for all subjects. In Kannada, one subject (K. K.) showed a slight effect of word
order, favoring a first-noun preference for NNV sentences (78”~).
Given the small numbers of patients tested, any generalizations drawn from the
aphasia data must be made with caution and should be viewed largely m terms of
their heuristic value. Even with such a small sample size, certain trends were
observed. For one thing, it would appear that semantic information, as reflected in
the animacy variable, is clearly preserved in these aphasics. As such the present data
are consistent with those reported for German aphasics (Bates et ul. 1987) who also
Bilingual Sentence Interpretation 179
showed a sparing of animacy. This sparing of animacy is consistent with the notion
that what should be preserved in aphasia are cues that were premorbidly the
strongest cues in the language. As we have already seen in Study 1, animacy was in
fact the strongest cue for Kannada speakers.
There does not appear to be a sparing of grammatical morphology in the present
sample of aphasics. This vulnerability of morphological cues is consistent with
observations reported on native English-speaking and Italian aphasics (Bates et al.
1987). but does not appear to be restricted to aphasics, for it is found even among
non-aphasic brain-damaged individuals (Bates et al. 1987). The theoretical signifi-
cance of this finding remains to be determined.
Finally, word order appears to have been partially spared in the present sample.
One of the monolinguals (S. A.) and one of the bilinguals (K. K.) showed a first-
noun preference for NNV sentences in Kannada, consistent with the strategy
observed in brain-intact monolingual and bilingual Kannada controls. This prefer-
ence for the first noun as agent in sentences in which two nouns precede the verb
reflects an SOV interpretation of this particular word order. It is noteworthy that,
even in the absence of case markers, both brain-intact and brain-damaged indi-
viduals were sensitive to this statistical and structural property of Kannada.
The normative bilingual data (Study 1) support a view of bilingual represen-
tation in which first language discourse strategies are mapped onto the second
language. The fact that this particular outcome was observed in our subject sample
despite the relative fluency and competence of our subjects in English is particularly
interesting and contrasts with the pattern observed for a group of Hindi-English
bilinguals (Vaid et al. 1987).
In the aphasic bilingual group as well, when there was a sparing of cues, it was in
favor of cues that are used in Kannada or, alternatively, the cue that is most
dominant in Kannada (namely, animacy) was extended to English sentence
interpretation as well, whereas the cue that is normally the strongest cue in English
speakers (namely, word order) was relatively absent in Kannada speakers, aphasics
and normals alike.
180 Journal of Neurolinguistics, Volume 3, Number 2 (1988)
This research was partially supported by a National Institute of Health Research
Grant on cross-linguistic studies of aphasia awarded to Elizabeth Bates of the
University of California, San Diego and by a Council for International Exchange
of Scholars Indo-American Advanced Research Fellowship awarded to Jyotsna
Vaid. We acknowledge with gratitude the Speech Pathology unit of the National
Institute of Mental Health and Neurological Sciences in Bangalore. India, under
whose auspices the data were collected. Tom Battocletti and Miguel Quinones of
Texas A & M University assisted in data analysis. We thank Janet McDonald for
her comments on the manuscript. Requests for reprints may be directed to the first
author c/o Department of Psychology. Texas A & M University. College Station,
TX 778434235, U.S.A.
Bilingual Sentence Interpretation l&II
Bates, E., A. Friederici, and B. Wulfeck
1987 “Comprehension in Aphasia: A Cross-linguistic Study,” Bruin and
Language 32, 19-67.
1988 “On the Preservation of Word Order in Aphasia: Cross-linguistic
Evidence,” Brain and Language 33, 323-364.
“Grammatical Morphology in Aphasia: Evidence from Three Lan-
guages,” Correx (in press).
Bates, E. and B. MacWhinney
1981 “Second Language Acquisition from a Functionalist Perspective:
Pragmatic, Semantic and Perceptual Strategies,” in Annals of the Neti*
York Academy of Sciences Conference on Native and Foreign Lan-
guage Acquisition, H. Winitz (ed.), New York: New York Academy of
Sciences. 1982 “Functionalist Approaches to Grammar,” in Language Acquisition:
The State of The Art, E. Wanner and L. Gleitman (eds.), New York:
Cambridge University Press.
1987 “Language Universals, Individual Variation and the Competition
Model,” in Mechanisms of Language Acquisition, R. MacWhinney
(ed.), Hillsdale, NJ: Lawrence Erlbaum Associates.
Bates, E. and B. Wulfeck
1987 “Cross-linguistic Studies of Aphasia,” Center for Research in Lan-
guage, University of California, San Diego.
1987 “Sentence Processing in a Second Language: Seeking a Performance
Definition of Fluency,” Unpublished Doctoral Dissertation, Univer-
sity of California, San Diego.
Kilborn, K. and A. Cooreman
1987 “Sentence Interpretation. Strategies in Adult Dutch-English Bilin-
guals.” Applied Psycholinguistics, 8, 415-43 1.
Kilborn, K. and T. Ito
1988 “Sentence Processing Strategies in Adult Bilinguals: Mechanisms of
Second Language Acquisition,” in Cross-linguistic Studies of Sentence
Processing. B. MacWhinney and E. Bates (eds.), Cambridge: Cam-
bridge University Press (in press).
1984 “The Mapping of Semantic and Syntactic Processing Cues by First and
Second Language Learners of English, Dutch, and German,” Unpub-