BRAIN
AND
LANGUAGE
7, 201-209 (1979)
The Effect of Syntactic Encoding on Sentence Comprehension in Aphasia HAROLD GOODGLASS,~HEILA MARY R. HYDE,EUGENE
E. BLUMSTEIN,JEAN BERKO GLEASON, GREEN, AND SHEILA STATLENDER
Boston Veterans Administration
Hospital and Boston University
This study compared two levels of syntactic encoding with respect to their effect on aphasics’ auditory comprehension. The same informational content was presented either as an embedded sentence or as two contiguous simple propositions. A similar contrast was utilized to compare sentences containing relational terms of time, comparison, instrumentality, as well as sentences involving conjunction reduction, with parallel messages expressed in expanded, simplified form. A picture-verification paradigm was used. Comprehension was significantly enhanced only in the case of embedded sentences, with a tendency for Broca’s aphasics to show the maximum effects.
Recent years have seen a turning away from the traditional tacit assumption that impairment of auditory comprehension can be treated as a global or undifferentiated impairment, varying only in level of severity from patient to patient. Corresponding to this philosophy, the major variable in the testing of comprehension tended to be the number of informational elements to be decoded. The classical prototype of this approach was Marie’s three-paper test (Weisenburg & McBride, 1964) in which the sheer quantity of information extracted from a multiple command was the yardstick against which comprehension was measured. The recent much more sophisticated Token Test (DeRenzi & Vignolo, 1962) seems to belong to this tradition. At least most users of this sensitive This work was supported in part by the Medical Research Service of the Veterans Administration and by Grants NS 07615 to Clark University and NS 06209 to Boston University. Requests for reprints should be sent to Dr. Harold Goodglass, Psychology Service, V.A. Hospital, 150 S. Huntington Ave., Boston, MA 02130. Sheila E. Blumstein is also at the Department of Linguistics, Boston University. Jean Berko Gleason is also at the Department of Psychology, Boston University. Eugene Green is also at the Department of English, Boston University. 201 0093-934X/79/020201-09$02.00/0 Copyright Q 1979 by Academic Press. Inc. All rights of reproduction in any form reserved.
202
GOODGLASS
ET AL.
instrument have treated it merely as a scale for assessing the global comprehension ability of aphasic patients. This tendency to ignore the multidimensionality of auditory language processing has been in sharp contrast to the study of speech production. Responding to the conspicuously varied clinical features of speech output, expressive disorders have typically been classified as to their focus on word finding (Howes & Geschwind, 1964; Wepman, Bock, Jones, & Van Pelt, 1956), syntax (Goodglass, 1968), or articulatory processes at the phonetic or at the phonemic level (Alajouanine, Ombredane, & Durand, 1939; Blumstein, 1973; Lecours & Lhermitte, 1969). Nevertheless, in more recent years researchers have become increasingly sensitive to the possibility that comprehension of language is not unidimensional but rather involves the complex interaction among lexical, syntactic, morphological, and semantic features of language, and further that deficits in auditory language comprehension may break down systematically in relation to these dimensions. In fact, in a reassessment of the Token Test, Whitaker and Noll (1972) demonstrated that the deep structure linguistic differences for implicit and overt instrumentals was a major factor contributing to performance deficits on this test. One of Luria’s major contributions in aphasia was to point out that the comprehension of what he termed logico-grammatical relationships could be disturbed quite autonomously from the comprehension of lexical content and from those aspects of sentential meaning which could be extracted from straightforward subject-verb-object sentence position (Luria, 1970). Further studies of comprehension in aphasia have indicated that patients have differential abilities in the understanding of such grammatical elements as prepositions, word order, and morphosyntactic units (Parisi & Pizzamiglio, 1970; Lesser, 1974). Not only do aphasics seem to show difficulties which cut across these dimensions, but such difficulties also seem to be selective as to types of aphasia. Goodglass, Gleason, & Hyde (1970) showed that the major subtypes of aphasia had a differential susceptibility to problems in lexical processing, prepositions, and auditory pointing span. Zurif and his co-workers (1972, 1976) have shown that Broca’s aphasics with agrammatism fail to discriminate between sentences which are lexically similar but which differ only in fine syntactic features. Thus, there is fairly clear evidence for dissociation between the auditory processing of lexical vs grammatical features. Further work has shown that agrammatic patients not only have difficulty with grammatical morphemes such as articles and plural but also have difficulty in processing syntactically complex sentences. In particular, Caramazza and Zurif (1978) demonstrated that a group of aphasic patients had difficulty in comprehending single object-relative sentences, e.g., “The dog that the boy is chasing is fat,” and that the Broca’s aphasics most particularly were unable to discriminate these sentences when all semantic cues were
GRAMMATICAL
ENCODING
IN APHASIA
203
kept constant and comprehension depended solely upon syntactic cues. Such deficits in syntactic processing suggest that these patients are unable to do purely syntactic operations which require embedding of two simple propositions. Alternatively, it might be the case that the difficulty in understanding these types of sentences is only secondarily a syntactic one and rather reflects an inability to process a number of different ideas or propositions. In the latter case, the disorder becomes an information processing deficit rather than a grammatical one. The speech production behavior of agrammatic patients would seem to suggest that these patients have the ability to process multiple propositions but their syntactic disorder requires that they be expressed as a concatenation of simple propositions. In particular, these patients commonly express ideas which are normally spoken in more compact syntactically encoded form, as a string of simple juxtaposed but syntactically unrelated utterances (Goodglass, Gleason, Bernholtz, & Hyde, 1972). For example, instead of using the comparative sentence “The man is bigger than the woman,” the patient might say “The man is big and the woman is small.” In this study, we have attempted to determine whether a similar phenomenon might occur on the receptive side. Specifically, would it be easier for an aphasic patient, and especially an agrammatic one, to understand an idea as a series of syntactically simple propositions than as a single syntacically complex sentence? In the former case, the presentation of several propositions demands the processing of more words than in the latter case where the nature of the syntactic rules of English allows the encoding of the same number of propositions but in a more compact form. There are various types of syntactic structures in English which reflect the combination of several simpler propositions, and we attempted to include several of these in the set of sentences used in this study. The first type includes embedding of one simple proposition in another, e.g., “The man greeted by his wife was smoking a pipe.” In this particular type of sentence we introduced another variable, that of plausibility. We reasoned that if a sentence were semantically unlikely, it would be more difficult to understand in grammatically encoded form in contrast to an expanded and simplified grammatical structure. The second type of structure consists of conjoining various syntactic structures, e.g., compound verb phrases as “The boy is walking and reading” and compound noun phrases as “The cup and bowl fell.” The third type of structure consists of various syntactic structures which are all syntactically simple sentences in the sense that no embedding occurs but which contain relational attributes: for example, prepositions of directionality as “The book is to the right of the ashtray” or with-of agency constructions, as “She’s touching the fork with the toothbrush.”
204
GOODGLASS
MATERIALS
ET AL.
AND METHODS
Subjects. Subjects were 22 aphasic veterans from the Aphasia Unit of the Boston Veterans Administration Hospital, all of who sustained unilateral damage to the dominant left hemisphere. Of these patients, 12 were Broca’s aphasics and 5 each were Wernicke and Conduction aphasics. These diagnoses were based upon the performance profiles of the patients on the Boston Diagnostic Aphasia Examination (Goodglass & Kaplan, 1972) and overall consensus reached at the Aphasia Rounds where all aspects of the patient’s neurological, psychological, and language status were reviewed. Subjects were also classified on the basis of their auditory language comprehension into two groups, those whose performance fell above the test mean and those whose performance fell below the mean on the auditory comprehension subtest of the aphasia battery. Of the 10 Broca’s aphasics, 6 were above and 4 below the mean. In the case of both the Wernicke and conduction aphasics, 2 were above and 3 were below the mean. Stimuli Two subtests were designed. The first subtest consisted of six subject-embedded constructions which varied along two dimensions, grammatical encodedness and plausibility, resulting in four versions of each theme. Plausible, grammatically encoded: Plausible expanded: Implausible, grammatically encoded: Implausible, expanded:
The man greeted by his wife was smoking pipe. The man was greeted by his wife, and he we smoking a pipe. The woman greeting her husband was smokin a pipe. The woman greeted her husband and she wa smoking a pipe.
The four versions of each of the six themes were interspersed in a random sequence to make a 24-item subtest. Corresponding to each theme was a four-picture response card (see Fig. 1). One of the
FIG.
tence.
1.
Example of picture-verification
item for embedded versus nonembedded
sen-
GRAMMATICAL
pictures
was
correct
for
ENCODING
the plausible
versions
(either
for the implausible
in grammatically
encoded
or expanded
(grammatically
encoded
or expanded
another
form), The labeled
and the remaining two served as foils. second subtest consisted of 48 items grouped into four “Before-After” consisted of six items of the following
The boy The third prepositions
boy
Two
is walking
is walking category, of direction
The The The sentence
versions
encoded and once in expanded form. The truck arrived before the bus. (grammatically truck arrived first and then the bus. (expanded)
The second category labeled “compounds” subjects, compound verbs, and compound
book book
and labeled “on,”
and
reading
he is reading, “Directionality,” “under,”
adjectives
With-of
before
a noun:
agency:
consisting The three
of types
left
encoded) (expanded).
two sentences of constructions
is a small round is a small table
panded). She’s
touching
the
versus
policeman The grammatically randomized to produce was drawn for each
is not
encoded and expanded sentences a subtest containing a total of48 of the sentence types, containing
for both the expanded serving as foils.
and
The
for
table and
fork
with
She’s
got
compact
the
corresponding to the particular
the second subtest. examiner first exposed test item aloud, test item.
and
to insure card for
required
the subject
each were:
of
three
the
toothbrush and
it (expanded). the policeman is strong and
(enthe
(expanded).
for each of the sentence types were items. A four-picture response card one picture corresponding to the
versions
In order a test
three
toothbrush
of the
sentence,
Subjects were tested individually in a quiet testing room, subtests on 2 separate days in the same order to all subjects. One subject, a Wernicke
not complete choices, the
the
(encoded) versus it is round (ex-
Procedure. given
compound
versus
she’s touching the fork with The fireman is stronger than coded) versus The fireman
Comparative:
choice pictures
in which e.g., versus
There There
(encoded)
correct three
encoded)
The
(expanded) included sentences containing “to the right of,” e.g.,
is to the right of the ashtray (grammatically is on the right and the ashtray on the
fourth category was a composite types, for a total of six items.
versus
of nine items represented,
(grammatically
and
categories. The first category. type. given once in grammatencoded).
consisted objects were too.
205
APHASIA
form),
ically
was correct
IN
the subject’s the subject‘s to point
familiarity inspection, to the picture
the
remaining
I and 2 being aphasic, did with then
the foul read the
corresponding
ANALYSIS OF RESULTS The two subtests were analyzed separately by means of analyses of variance. The ANOVA for subtest 1 was based on two between-subject variables (diagnosis and comprehension level) and two within-subject variables (grammatical encodedness and plausibility). Results of this analysis are summarized in Table 1. The only significant effect is a main effect for encodedness (F (1,16) = 5.47, p < .03). This is due to higher performance scores for syntactically expanded sentences in comparison to the encoded sentences. It is worth noting that plausibility had no effect on picture verification; in fact, the implausible items showed a slight
206
GOODGLASS
ET
AL.
TABLE I PERCENTAGE CORRECT FOR PLAUSIBLE EMBEDDED AND EXPANDED
VERSUS IMPLAUSIBLE SENTENCES’ Expanded
Compact Plausible Broca
+ + -
Wernicke Conduction
67 50 50 50 56 59
+ -
a Plus and minus mean.
Implausible
signs indicate
71 50 67 44 61 67 auditory
Implausible
Plausible 79 75 59 50 78 50
comprehension
scores
79 59 59 67 66 50
above
and below
the test
superiority in performance. There is a near significant interaction for Group x Auditory comprehension level x Syntax (F (2,16) = 2.69, p < .lO). Inspection of the table indicates that this is due to the fact that the Broca’s aphasics, particularly the low comprehenders, gain the most from the expanded as opposed to the embedded versions, while the Conduction aphasics gain the least. The second analysis of variance was based on two between-subject variables (diagnosis and comprehension level) and two within-subject variables (syntactic type and encodedness). Table 2 summarizes the mean percentage correct for each diagnostic group by syntactic type and encodedness. Results indicate a significant main effect only for syntactic type (F (3,45) = 5.07, p < .005). This result is due to the fact that Before-After constructions were the most difficult to comprehend while compound verbs and nouns were the easiest. All other effects, in particular, groups, comprehension level, and encodedness were nonsignificant, although there was a tendency in the expected direction for differences in performance between high and low auditory comprehension subjects (F (1,15) = 2.50, p = .14). Finally,, although there was a nonsignificant MEAN
TABLE 2 PERCENTAGE CORRECT FOR GRAMMATICALLY CODED VERSUS EXPANDED FORMS OF SENTENCES B/A
Broca Wernicke Conduction Across
groups
Compact Expanded Compact Expanded Compact Expanded
36.4 61.6 62.8 45.8 41.5 41.8 48.3
Compounds 76.6 76.6 67.3 76.3 73.4 61.3 71.9
(COMPACT)
Direction
Other
62.5 54.2 41.5 58.3 55.3 63.8 55.9
55.2 66.6 54.0 62.5 55.5 66.8 60.1
GRAMMATICAL
ENCODING TABLE
NUMBER OF SUBJECTS HAVING SIONS, ON EXPANDED VERSIONS, GRAMMATICAL STRUCTURES With-of Both
207
APHASIA
3
MORE SUCCESS ON GRAMMATICALLY ENCODED VEROR EQUAL SCORES ON Bow FOR THREE TYPES OF
Comparative
agency
0%
IN
adjectives
Both 100% Both 50% Expanded
I 4 9
2 5
0 II
2 10*
7 2
Encoded
6
2
I
*Statistically
1
Double
significant
(p