Verb Production in Aphasia: Testing the
Division of Labor between Syntax and
Semantics
Julia Thorne, M.A.1 and Yasmeen Faroqi-Shah, Ph.D., CCC-SLP1
ABSTRACT
Some individuals with aphasia preferably use semantically general
light verbs, whereas others prefer semantically specific heavy verbs. This
study aimed to test Gordon and Dell’s “division of labor” hypothesis that
light versus heavy verb usage depends on syntactic and semantic processes,
respectively. In a retrospective analysis of data from the AphasiaBank
corpus, narrative language of neurologically healthy individuals and
individuals with aphasia was analyzed for the proportion of light verbs
used, and its relationship with narrative measures of syntactic and semantic
sophistication and verb naming scores was examined. In individuals with
aphasia, light verb usage was positively correlated with a syntactic measure
(developmental sentence score) and negatively associated with two semantic measures (idea density and verb naming). For healthy individuals, the
number of verbs per utterance, which is a measure of syntactic complexity,
predicted light verb use. These findings suggest that light verb usage in
aphasia observes an inverse relationship with syntactic and semantic
abilities, supporting the division of labor hypothesis.
KEYWORDS: Aphasia, light verbs, narratives, verb complexity
Learning Outcomes: As a result of this activity, the reader will be able to (1) identify the difference between
light and heavy verbs; (2) discuss how syntactic impairments manifest in aphasia and influence light verb
retrieval; and (3) describe ways of measuring syntactic and semantic abilities in narrative speech.
D
ifficulties in retrieving and processing
verbs have been reported across a variety of
neurologic conditions, including Parkinson dis-
ease, progressive supranuclear palsy, motor
neuron disease, and particularly in aphasia.1–3
In aphasia, verb retrieval struggles are reported
1
Department of Hearing and Speech Sciences, University of
Maryland, College Park, Maryland.
Address for correspondence: Yasmeen Faroqi-Shah, Ph.
D., CCC-SLP, Department of Hearing and Speech Sciences, University of Maryland, 7251, Prienkert Drive,
College Park, MD 20742 (e-mail: [email protected]).
A Look to the Future: Big Data in Neurorehabilitation;
Guest Editor, Yasmeen Faroqi-Shah, Ph.D., CCC-SLP
Semin Speech Lang 2016;37:23–33. Copyright # 2016
by Thieme Medical Publishers, Inc., 333 Seventh Avenue,
New York, NY 10001, USA. Tel: +1(212) 584-4662.
DOI: http://dx.doi.org/10.1055/s-0036-1571356.
ISSN 0734-0478.
23
24
SEMINARS IN SPEECH AND LANGUAGE/VOLUME 37, NUMBER 1
more frequently than noun retrieval deficits.4 In
a review of 280 individual scores, 75% of the
persons with aphasia (PWA) performed more
poorly on verbs compared with nouns.4 The
reason for the vulnerability of verbs in aphasia is
unclear, and therefore poses an impediment for
targeted intervention of word retrieval deficits
in aphasia. Intervention of verb retrieval deficits
is crucial for subsequent treatment of sentence
production because verbs form the core of a
sentence.
One account of verb deficits is that verbs are
more complex and hence more vulnerable to
brain damage. However, this argument is
undermined by the numerous reports of double
dissociations between verb and noun retrieval in
individuals and in group studies.5–8 A neuroanatomical explanation for verb deficits has been
elusive because they have been associated with a
variety of lesion locations.4,9 A third account of
verb deficits in aphasia is that these are a
component of a broader linguistic deficit, particularly agrammatism.10–13 Agrammatic language is characterized by impoverished
sentence structure and functional morphology
along with a paucity of verbs. However, verbspecific deficits have been found in persons who
do not produce agrammatic language, such as
those with fluent aphasia, and not all agrammatic individuals show verb deficits.4,14 One
problem in examining the relationship between
verb deficits and agrammatism is that both of
these symptoms are viewed as dichotomous; a
PWA either has or does not have agrammatism
or does or does not have a verb deficit. This
categorical classification is problematic because
there is no agreed upon objective measure (or set
of cutoff scores) for either symptom. Therefore,
in the present study, we examine the relationship between grammatical abilities and verb
usage by treating these as continuous variables.
WITHIN-VERB DIFFERENCES IN
APHASIA
To elucidate the source of verb deficits in
aphasia, investigators have examined dimensions along which verbs vary, such as transitivity, imageability, instrumentality, and noun
homophony.15–20 The logic is that these variables denote representational complexity of
2016
verbs, thus potentially influencing verb breakdown in aphasia. Syntactic complexity of a verb
is often represented by verb argument structure
(VAS), which refers to the number of arguments a verb requires and the number of
different argument alternations the verb
takes.21 For instance, verbs such as sleep and
bark require a single argument, which is the
agent of the action (the dog slept/barked). Other
verbs may require two (e.g., wash, push) or three
(e.g., give, keep) arguments. Still other verbs can
alternate between transitive and intransitive (or
dative) contexts (e.g., the ice cream melted versus
the candlemaker melted the wax). Some studies
have found a preferential usage of verbs with
simpler VAS by persons with agrammatic
aphasia.18,19,22,23 However, a corpus-based
study of 173 narratives from the AphasiaBank
database found no effect of argument structure
complexity or difference in usage between
PWA and healthy controls.24,25
Along the semantic dimension, complexity
of verbs has been investigated with variables such
as imageability, manipulability, and verb weight.16
Variables such as imageability and manipulability
refer to a single semantic attribute, whereas verb
weight denotes the overall semantic specificity or
complexity of a verb. At one extreme are light
verbs, a specific subset of semantically underspecified verbs whose meaning can vary widely
according to context.26 Heavy verbs are the more
specific, semantically complex verbs, usually including all verbs that are not considered light. For
example, jog, bake, and donate are heavy whereas
go, make, and give are light. Other criteria used to
identify light verbs are those that are highly
frequent,27 often grammaticalized cross-linguistically (i.e., verbs that were once lexical but became
closed-class morphemes, such as auxiliary
verbs),28 and take a diverse number and variety
of complements.29,30 Thus, whereas light verbs
are semantically underspecified, these are syntactically more complex. In contrast, heavy verbs are
used in a rather limited variety of syntactic
constructions, hence are, in general, syntactically
less diverse.
The contrasting semantic and syntactic complexities of heavy and light verbs were formalized
by Gordon and Dell into a connectionist model of
sentence production called the “division of labor”
hypothesis.31 According to this model, a division
VERB PRODUCTION IN APHASIA/THORNE, FAROQI-SHAH
of labor between syntactic and semantic mechanisms guides lexical activation during language
production. For instance, a speaker’s conceptual–
semantic intentions may increase the activation of
lexical nodes that correspond to specific events or
entities. Simultaneously, a syntactic network is
also activated that contributes information about
which types of words are syntactically appropriate
at that particular point in the utterance, such as
verb, noun, or auxiliary. In the case of verbs, a
situation with high syntactic constraints and
underspecified conceptual–semantic constraints
will yield selection of light verbs (e.g., give),
whereas higher conceptual–semantic activation
will result in retrieval of a heavy verb (e.g., donate).
In Gordon and Dell’s connectionist model, a
lesioning of the syntactic system will result in a
selective impairment of light verbs in aphasia and
a reliance on heavy verbs in language production.
This syntactic lesioning is analogous to the
syntactic impairment found in agrammatic aphasia. Conversely, an impairment to the semantic
system with the syntactic system remaining intact
would result in a scarcity of heavy verbs and an
overuse of light verbs. This is illustrated in the
Appendix with language samples from two
PWA; Scale02a exclusively produces heavy verbs
and fragmented utterances. In contrast, Star03a
produces several complete utterances but mainly
light verbs. The syntax–semantics division is a
promising theoretical account of verb impairment
in aphasia that we test in the present study.
Previously, the influence of verb weight on
retrieval in aphasia has been examined using
sentence or story completion tasks where participants fill in a missing verb,8,19,32 and by
analyzing narrative language.19,33 Across these
studies, of 40 individuals with aphasia tested
with the sentence- or story-completion paradigm, 28 showed a numerical (but not statistically significant) advantage for heavy verb
naming, whereas the remaining 12 demonstrated comparable performance for heavy and light
verbs. In narrative language, 15 of 19 participants with aphasia showed an advantage for
heavy verbs, whereas the remaining 4 showed
an advantage for light verbs. Crucially, across
these studies, there was no consistent relationship between the presence of agrammatism and
light verb use. Hence current evidence is inconclusive regarding the use of light verbs and the
division of labor between syntactic and semantic
impairments.
In PWA, patterns of light verb use could be
explained in ways other than the division of
labor hypothesis. Notably, light verbs are semantically diverse, or highly variable in meaning
across contexts, which poses a challenge for
persons with semantic access deficits.34,35
Hence, contrary to the division of labor, the
semantic diversity of light verbs predicts an
association between semantic deficit and limited
light verb use. Another account is that the heavy
verb preponderance seen in most individuals
with aphasia was also found in healthy adults.19
Therefore, it is possible that individuals with
aphasia who are less impaired overall exhibit this
typical pattern, whereas an increased proportion
of light verbs is seen in the more severely
impaired. Heavy verbs might be more difficult
than light verbs for more severely impaired
individuals because their enriched semantic
representations may simply present an increased
processing load. A factor that may favor light
verb use is their earlier age of acquisition and
higher usage frequency than heavy verbs.27
Thus, it is possible that the heavy–light differences are not an effect of syntax–semantics as
Gordon and Dell propose,31 but a result of other
factors such as semantic diversity, aphasia severity, or lexical frequency.
To summarize, although the difference in
semantic complexity of verbs could have a
crucial influence on verb retrieval and sentence
production, only four studies have examined
this issue, with inconsistent findings. Two ways
in which the present limitations can be addressed are to use larger sample sizes that would
allow for more statistically robust effects and to
consider verb use and syntactic and semantic
abilities as continuous rather than categorical
measures. Hence people are not classified a
priori as agrammatic or verb impaired.
THE PRESENT STUDY
The purpose of this study is to test the division
of labor hypothesis model as well as alternative
accounts of light and heavy verb distinctions in
PWA.31 Additionally, the neurotypical pattern
of light and heavy verb use is unclear: both a
preponderance of heavy verbs and comparable
25
26
SEMINARS IN SPEECH AND LANGUAGE/VOLUME 37, NUMBER 1
heavy–light distribution has been reported.19,33
Therefore, we first sought to establish the
pattern of heavy and light verb use in the
narrative language of neurologically healthy
English-speaking adults and PWA. We predicted that less than half of the verbs used by
neurologically healthy individuals would be
light.19 According to Gordon and Dell’s
model,31 aphasic individuals are expected to
show different patterns of light verb production
based on their relative impairment of syntactic
and semantic processes. Therefore, we predicted that, as a group, PWA would be more
variable (i.e., greater variance) in the proportion
of light verbs used compared with neurologically healthy individuals.
Second, to test the central predictions of
the division of labor hypothesis,31 we examined
if measures of syntactic ability or semantic
ability predict the proportion of light verbs
used in narrative speech. The two outcomes
that would support the division of labor are an
association between a lower use of light verbs
and low performance on syntactic measures,
and high performance on semantic measures. If
the increased semantic diversity of light verbs
presents problems for individuals with semantic
access deficits, then lower semantic performance would be associated with lower proportion of light verbs. If aphasia severity exerts a
stronger influence on verb production, then
there would be an inverse relationship between
the proportion of light verbs used and aphasia
severity—that is, less severe individuals would
be more consistent with the neurotypical pattern of lower light verb use.
The previous questions were investigated
with narrative language, to avoid confounds associated with elicitation of verbs in isolation,
which may favor the more imageable heavy verbs.
Measures of semantic and syntactic performance
were also obtained from the same narrative
language samples to be consistent with the context in which heavy and light verbs are produced.
Numerous measures of semantic and syntactic
ability in narrative language have been proposed
in prior literature, in studies of both language
acquisition and aphasia.36–40 After reviewing a
variety of possible measures, three different measures each were selected for syntactic and semantic
analysis. One measure of syntactic ability was the
2016
proportion of grammatical utterances, which is
commonly used to describe agrammatism and
quantifies the syntactic impairment.19 The number of verbs per utterance captures syntactic
complexity because utterances with more than
one verb, such as embedded clauses, are syntactically complex. As a global measure of syntactic
ability, developmental sentence scoring was selected.37 Although this measure has been traditionally used to quantify children’s syntactic
development, it can be used to provide an overall
view of any individual’s morphosyntactic abilities
because it takes into account utterance complexity, grammatical accuracy, and the use of function
words.37
Semantic measures included vocabulary
density (VOCD),38 which measures the diversity of words used by a speaker. Ratios of
different words to total word tokens, such as
type-token ratio are correlated with the length
of the language sample, and hence VOCD
calculates the probability of new vocabulary
given a sample length and yields a more stable
measure of lexical diversity.41,42 Second, we
quantified the lexical completeness of the narrative. For stories that are commonly retold as a
means of assessing narrative language, such as
the Cinderella story used in previous studies of
light verb usage and in this study,19,42 a core
lexicon of key words that make the story
semantically complete has been published.40
So we measured the proportion of words
from this core lexicon that occurred in each
narrative. The third semantic measure was
propositional idea density (ID),37 which has
been used in the analysis of the Nun Study of
Aging and Alzheimer disease.43,44 It is calculated as the number of propositions (nonmodal
verbs, adverbs, adjectives, prepositions, and
subordinating conjunctions) expressed over total number of words in a sample.
METHODS
This study was a retrospective analysis of data
available on AphasiaBank,45 an online database
of transcriptions of discourse produced by
PWA secondary to a cerebrovascular accident
as well as neurologically healthy individuals.
Both aphasic and nonaphasic participant data
on AphasiaBank was gathered according to a
VERB PRODUCTION IN APHASIA/THORNE, FAROQI-SHAH
prespecified protocol, with scores on the Western Aphasia Battery–Revised (WAB-R),45 the
short form Boston Naming Test–2nd Edition
(BNT),46 the Verb Naming Test (VNT) from
the Northwestern Assessment of Verbs and
Sentences–Revised,47 and the nonstandardized
AphasiaBank Repetition Test24 available from
each aphasic participant, as well as narrative
language samples from all participants. The
narrative language elicitation protocol AphasiaBank is described by MacWhinney et al.24
This study used a sample of discourse from each
participant elicited by the retelling of the Cinderella story rather than those elicited by picture
description, given that stories generate richer
language than single pictures and to be consistent with past research.19,32,33,42,48
Participants
Participants were included whose transcripts
contained at least 100 words to obtain a representative language sample. This criterion
yielded 164 monolingual PWA (86 male, 78
female, mean age ¼ 61.0) and a control group
of 166 monolingual English speakers without
aphasia (76 male, 90 female, mean age ¼ 63.3).
Participants with aphasia were more than
1-year post–cerebrovascular accident, control
participants had no history of neurologic impairment, and no participants had a history of a
cognitively deteriorating condition. Experimental and control groups were matched for
age (t [328] ¼ 1.42, p > 0.05), years of education (t [328] ¼1.07, p > 0.05) and gender
distribution (Fisher exact test, p > 0.05).
Proportion of Light Verbs
Light verbs were defined based on several
factors, including those that are previously
defined as light in at least two empirical
studies (developmental and aphasia literature), are highly frequent in English corpora,50,51 take diverse noun complements,30 and
share the feature of frequently grammaticalizing cross-linguistically.28 The final list of
nine light verbs is as follows: come, do, get,
give, go, have, make, put, and take. Auxiliaries,
modals, and copulas are neither considered as
light or heavy verbs due to their grammatical
functions. The FREQ program in CLAN was
used to obtain counts of light and heavy verbs.
The total number of light verbs was then
divided by the total number of verbs in the
sample to obtain the proportion of light verbs.
Syntactic and Semantic Measures
We used three syntactic measures: proportion
of grammatical errors, number of verbs per
utterance, and developmental sentence score
(DSS); and three semantic measures: VOCD,
proportion of words from the Cinderella core
lexicon, and propositional ID. The EVAL
program in CLAN was used to obtain the
number of utterances with grammatical errors
and the average number of verbs per utterance. The DSS was obtained from the
KidEVAL program.† To obtain the proportion of words from the Cinderella core lexicon
present in each narrative sample, the 10 most
frequent nonlight verbs and 10 most frequent
nouns were selected from the core lexicon of
Data Analysis
Narrative samples in the AphasiaBank database
have already been previously transcribed wordfor-word in the CHAT format by the researchers who were involved in the originally collected
samples. Words have been tagged with morphosyntactic roles and coded for aphasic errors
such as semantic paraphasias and neologisms.
Utterances have been coded with utterancelevel errors, including the presence of grammatical errors. Computerized language analysis
(CLAN) was used to conduct automated analyses of the narratives, as described later.49
†
Another syntactic measure used in language acquisition
research, the Index of Productive Syntax (IPSyn) measures
the presence of 56 syntactic structures without weighting for
age of acquisition.61 However, the IPSyn score is
traditionally based on 100-utterance samples, an unfeasible
length for the aphasic Cinderella narratives used in this
study, which averaged to of 37 utterances (range was 10 to
87). IPSyn scores were more strongly correlated with
narrative sample length (rs ¼ 0.525, p < 0.01) than DSS
scores (rs ¼ 0.376, p < 0.01). Therefore, DSS was selected
in this study as a global indicator of syntactic ability.
Grammatical errors in the AphasiaBank transcripts were
originally represented by the [þ gram] code, whereas DSS
counted grammatical errors with the [] code. To accurately
calculate DSS, transcript codes were altered to reflect
grammatical errors with the [] code.
27
28
SEMINARS IN SPEECH AND LANGUAGE/VOLUME 37, NUMBER 1
MacWhinney et al. 40 Both nouns and verbs
were chosen to prevent possible impact of
noun- or verb-specific lexical impairments
on proportional scores, and light verbs were
excluded as they do not contribute informational content to the Cinderella story. The
total number of word types from this core
lexical set present in each sample was
counted for each participant using the
FREQ program. VOCD and ID were calculated automatically from KidEVAL and
EVAL respectively. ID calculation uses the
Computerized Propositional Idea Density
Rater software, 54 which is available in
CLAN. 49
Separate multiple regression analyses were
used to identify predictors of light verb usage in
the neurologically healthy and aphasia groups.
Five predictor variables were used for the
healthy group: the number of verbs per utterance, DSS, VOCD, the proportion of the core
lexicon present, and ID for their narrative
samples. In addition, predictor variables for
the aphasic group included scores on the
BNT, the VNT, and the AphasiaBank
Repetition Test (part II.B, the total number
of words correct).
2016
RESULTS
There was no significant correlation between the
total number of verbs and number of light verbs in
the data, precluding effects of overall sample
length on the syntactic or semantic measures
(aphasic group, rs ¼ 0.10, n ¼ 164, p > 0.05;
control group, r ¼ 0.05, n ¼ 166, p > 0.05).
The WAB-R Aphasia Quotient and the proportion of light verbs used were not significantly
correlated (r ¼ 0.013, p > 0.05), suggesting that
the severity of aphasia did not influence the
proportion of light verbs used. There was also
no difference between WAB aphasia classification
(Broca’s n ¼ 23, Wernicke’s n ¼ 15, conduction
n ¼ 30, anomic n ¼ 72, and nonaphasic [i.e.,
aphasia severity low enough to preclude classification on the WAB-R] n ¼ 20) and proportion
of light verbs (F [4, 155] ¼ 1.14, p > 0.05).
Light Verb Proportions
The mean proportion of light verbs used by
controls was 0.38 (standard deviation ¼ 0.09),
with a normal distribution (Shapiro-Wilks test,
W ¼ 0.99, p > 0.05, see Fig. 1). The proportion
of light verbs used by aphasic individuals was
similar to the control group but with the wider
Figure 1 Histograms of the proportion of light verbs produced by the aphasic and control groups.
VERB PRODUCTION IN APHASIA/THORNE, FAROQI-SHAH
dispersion (mean ¼ 0.39; standard deviation
¼ 0.17; Levene’s test for equality of variances,
F ¼ 47.16, p < 0.01) and normal distribution
(W ¼ 0.99, p > 0.05). Due to unequal variances,
the groups were compared with an independent
samples Mann-Whitney U test, which revealed
no significant difference in proportion of light
verbs between the two groups (U
[298] ¼ 13,129, Z ¼ 0.56, p > 0.05).
Predictors of Light Verb Use
For neurologically healthy persons, The linear
regression model with five predictor variables
was statistically significant (F [5, 160] ¼ 2.91,
p < 0.05, R2 ¼ 0.08, R2Adjusted ¼ 0.05). Only
one variable, the number of verbs per utterance,
emerged as a significant predictor of light verb
proportion (t [165] ¼ 2.86, p < 0.01) with
a negative relationship. That is, as the
number of verbs per utterance (i.e., syntactic
complexity) increased, the number of light
verbs used decreased.
For the aphasic group, the nine predictor
model was statistically significant (F [9,
154] ¼ 4.47, p < 0.01, R2 ¼ 0.21, R2Adjusted
¼ 0.16). Three predictors were significant, DSS
(t [163] ¼ 2.64, p < 0.01), ID (t [163] ¼
2.01, p < 0.05), and VNT scores (t [163] ¼
2.50, p < 0.05). DSS was positively associated with light verb proportion, meaning that a
higher DSS score (i.e., higher syntactic ability)
was related to increased use of light verbs. ID was
negatively associated with light verb proportion,
indicating a higher ID score (i.e., greater semantic
ability) was related to decreased use of light verbs.
Thus, both DSS and ID varied in a direction that
was consistent with the Gordon and Dell model.31 The VNT score was also negatively associated with light verb proportion, indicating that
better verb naming was associated with lower
occurrence of light verbs in narrative language.
Correlations among Syntactic and
Semantic Measures in Aphasia
DSS correlated significantly with both other
syntactic measures (for the number of verbs per
utterance, r ¼ 0.77, p < 0.001; for the proportion of grammatical utterances, r ¼ 0.35,
p < 0.001, using a level of 0.001). ID signifi-
cantly correlated with VOCD (r ¼ 0.32,
p < 0.001, using a Bonferroni adjusted a level
of 0.001), but not with the proportion of the
core lexicon used (r ¼ 0.108, p > 0.001,
using a level of 0.001).
Standardized Difference Score
For individuals with aphasia, in accordance
with the Gordon and Dell model,31 it was
suspected that a measure of relative syntactic
or semantic impairment might provide more
insight into the proportion of light verbs used
rather than absolute measures of syntactic or
semantic ability because most individuals
exhibit varying degrees of both semantic and
syntactic impairments. Hence a difference score
was calculated for each aphasic individual by
first converting their ID and DSS scores into
scores (z scores), and then subtracting the
standardized DSS score from the standardized
ID scores. A positive difference score indicated
relative greater syntactic impairment, a negative
score indicated relative semantic impairment,
and a score of 0 indicated equal severity of
semantic and syntactic impairment.
A Spearman correlation analysis between
the standardized difference score and the proportion of light verbs used showed a significant
small negative association between the two
variables (rs ¼ 0.17, p < 0.05). That is,
for persons with relatively greater syntactic
impairment, the proportion of light verbs
decreased, as predicted by Gordon and Dell.31
DISCUSSION
The main goal of this study was to test Gordon
and Dell’s division of labor hypothesis of light
verb use.31 This was achieved by identifying the
pattern of verb production in neurologically
healthy and aphasic individuals’ narratives and
comparing the use of light verbs with measures
of syntactic and semantic ability for both
groups. The main findings were that aphasic
and neurologically healthy individuals produced similar average proportions of light verbs
(38% of all verbs), but individuals with aphasia, as a group, were more variable. For individuals with aphasia, the proportion of light verbs
showed no relationship to aphasia severity or
29
30
SEMINARS IN SPEECH AND LANGUAGE/VOLUME 37, NUMBER 1
subtype of aphasia. Narrative language measures of syntactic and semantic ability were
significantly associated with the proportion of
light verbs used. These findings will be discussed in the following sections.
Predictors of Light Verb Use
Neurologically healthy individuals showed an
association between increased light verb use and
increased syntactic complexity, as represented
by verbs per utterance. That is, individuals using
more syntactically complex language also used
more heavy verbs. This finding is inconsistent
with Gordon and Dell’s proposal,31 and it
suggests that individuals who produced syntactically complex language more often produced
semantically specific verbs. In other words,
some neurologically healthy speakers produced
overall more sophisticated narrative language
than others.
For individuals with aphasia, four findings were consistent with Gordon and Dell’s
proposal31: light verbs’ positive correlation
with DSS and negative correlations with
ID, ID/DSS difference score, and VNT.
The latter test mainly elicits heavy verbs
such as wash and pour.47 These results extend
previous studies that considered syntactic
impairment as a categorical variable by focusing on persons with agrammatism.19,33 Here
we use a large data set to show that grammatical ability in aphasia is a continuum that is
consistently associated with light verb use. A
novel aspect of this study is to show the
inverse relationship between semantic ability
and light verb usage in aphasia. Prior support
of the inverse relationship between semantic
ability and light verb use came from persons
with probable Alzheimer disease.19
Although six measures were used to measure syntactic and semantic ability in narratives, light verb usage was positively
correlated with only one syntactic measure,
the DSS, and negatively correlated with one
semantic measure, ID. A possible explanation
for why DSS and ID were associated with
light verb use but not other measures of
syntactic and semantic ability is that the
calculations of ID and DSS are more global
compared with the other four syntactic and
2016
semantic measures. For example, DSS takes
into account the proportion of grammatical
utterances (grammatical accuracy), the number of verbs per utterance (syntactic complexity), and occurrence of various functional
morphemes, all of which capture symptoms
of agrammatic aphasia. The DSS gives greater
weight to morphosyntactic forms that are
acquired later,37 and presumably these are
syntactically more complex and problematic
for persons with agrammatic aphasia.12,52,53
In contrast, the other syntactic measures
(proportion of verbs per utterance and grammatical utterances) measure one aspect of
agrammatism. Similarly for semantic measures, ID, calculated as the average number of
propositions controlling for sample length, is
a more global measure than VOCD or the
proportion of the core lexicon present. However, a cautionary note is that ID has been
suggested to be sensitive to syntactic as well as
semantic production because its count of
propositions includes content words and select function words: verbs, prepositions, adjectives,
adverbs,
and
coordinating
conjunctions.54
Clinical Implications of the Study
Verb retrieval in healthy speakers and those with
aphasia is complex and incompletely understood.
Some factors that were known to contribute to
verb retrieval were imageability,55 semantic features such as manipulability,56,57 and frequency.34
This study contributes to our understanding of
verb retrieval by empirically supporting verb
weight as a relevant dimension of verb retrieval
and by providing empirical support to the division
of labor hypothesis. The practical relevance of this
study is that it provides clinicians who work with
PWA additional clues about the underlying nature of deficit in their clients. The language
samples in the Appendix illustrate the association
between syntactic structure and heavy verb usage.
Clinicians can also use automated calculations of
objective measures such as DSS and ID,24 not
only to capture overall syntactic and semantic
abilities but also to examine their influence on
verb retrieval. Remediation of verb retrieval deficits is important because, as pointed out earlier,
these are much more frequent than noun retrieval
VERB PRODUCTION IN APHASIA/THORNE, FAROQI-SHAH
impairments in aphasia.4 Furthermore, verbs
form the framework of a sentence; when a verb
is retrieved from the mental lexicon, its VAS
influences which arguments (or other entities)
need to be strung together to formulate a complete sentence. Hence addressing verb retrieval in
PWA is functionally significant. Several studies
have found improvement in sentence production
following verb retrieval intervention.62
There are a few points of caution in
applying the findings of this study to theorizing and clinical practice. First, although this
study showed a relationship between light
verb retrieval in narrative language and the
VNT, other studies have shown poor or
modest correspondence between narrative
and confrontation naming tasks.3,58–60 Furthermore, although we argued in favor of
treating syntactic and semantic abilities as
continuous variables, one could make similar
case for verb weight as a continuous variable
(and not light versus heavy categories).
Maouene et al have attempted to address
this concept, quantifying verb weight continuously as a function of the diversity of complements following the verb.30 Other
important factors, however, could be factored
into quantification of the “lightness” of a verb,
such as its frequency and the extent of its
grammaticalization. By creating a system for
quantifying verb weight continuously, future
studies could examine the relationship between syntactic ability, semantic ability, and
verb weight in narrative language more
precisely.
To conclude, overall this study supports
the view that, in aphasia (but not in neurologically healthy adults), there is a trade-off between light verb/syntactic abilities and
semantic abilities in narrative language. This
study contributes to the understanding of factors underlying verb specific deficits in aphasia,
a problem that has proved complex and remains
poorly understood. It is clear that a multitude
of factors are involved in the lexical retrieval of
verbs: word frequency, imageability, morphological complexity, and now semantic complexity have all been shown to play a role across
studies of verb retrieval, and further study of
these factors will help fully comprehend the
representations and processes underlying the
production of verbs.
REFERENCES
1. Daniele A, Giustolisi L, Silveri MC, Colosimo
C, Gainotti G. Evidence for a possible
neuroanatomical basis for lexical processing of
nouns and verbs. Neuropsychologia 1994;32(11):
1325–134
2. Bak TH, O’Donovan DG, Xuereb JH, Boniface S,
Hodges JR. Selective impairment of verb processing associated with pathological changes in Brodmann areas 44 and 45 in the motor neurone
disease-dementia-aphasia syndrome. Brain 2001;
124(Pt 1):103–120
3. Zingeser LB, Berndt RS. Retrieval of nouns and
verbs in agrammatism and anomia. Brain Lang
1990;39(1):14–32
4. Mätzig S, Druks J, Masterson J, Vigliocco G. Noun
and verb differences in picture naming: past studies
and new evidence. Cortex 2009;45(6):738–758
5. Caramazza A, Hillis AE. Lexical organization of
nouns and verbs in the brain. Nature 1991;
349(6312):788–790
6. Damasio AR, Tranel D. Nouns and verbs are
retrieved with differently distributed neural systems. Proc Natl Acad Sci U S A 1993;90(11):
4957–4960
7. McCarthy RA, Warrington EK. Evidence for
modality-specific meaning systems in the brain.
Nature 1988;334(6181):428–430
8. Berndt RS, Mitchum CC, Wayland S. Patterns of
sentence comprehension in aphasia: a consideration
of three hypotheses. Brain Lang 1997;60(2):
197–221
9. Aggujaro S, Crepaldi D, Pistarini C, Taricco M,
Luzzatti C. Neuro-anatomical correlates of
impaired retrieval of verbs and nouns: interaction
of grammatical class, imageability and actionality.
J Neurolinguist 2006;19(3):175–194
10. Fillenbaum S, Jones L, Wepman J. Some linguistic
features of speech from aphasic patients. Lang
Speech 1961;4:91–108
11. Myerson R, Goodglass H. Transformational grammars of three agrammatic patients. Lang Speech
1972;15(1):40–50
12. Saffran EM, Berndt RS, Schwartz MF. The quantitative analysis of agrammatic production: procedure and data. Brain Lang 1989;37(3):440–479
13. Goodglass H. Agrammatism in aphasiology. Clin
Neurosci 1997;4(2):51–56
14. Edwards S, Bastiaanse R. Diversity in the lexical
and syntactic abilities of fluent aphasic speakers.
Aphasiology 1998;12(2):99–117
31
32
SEMINARS IN SPEECH AND LANGUAGE/VOLUME 37, NUMBER 1
15. Bastiaanse R, Jonkers R. Verb retrieval in action
naming and spontaneous speech in agrammatic and
anomic aphasia. Aphasiology 1998;12(11):951–969
16. Bird H, Howard D, Franklin S. Why is a verb like
an inanimate object? Grammatical category and
semantic category deficits. Brain Lang 2000;72(3):
246–309
17. Jonkers R, Bastiaanse R. The influence of instrumentality and name-relation to a noun on verb
comprehension in Dutch aphasic speakers. Aphasiology 2006;20(1):3–16
18. Kim M, Thompson CK. Verb retrieval in agrammatism. Brain Lang 2000;74:1–25
19. Kim M, Thompson CK. Verb deficits in
Alzheimer’s disease and agrammatism: implications
for lexical organization. Brain Lang 2004;88(1):1–20
20. Stavrakaki S, Alexiadou A, Kambanaros M, Bostantjopoulou S, Katsarou Z. The production and
comprehension of verbs with alternating transitivity by patients with non-fluent aphasia. Aphasiology
2011;25(5):642–668
21. Grimshaw J. Argument Structure. Cambridge,
MA: MIT Press; 1990
22. Bastiaanse R, Hugen J, Kos M, van Zonneveld R.
Lexical, morphological, and syntactic aspects of
verb production in agrammatic aphasics. Brain
Lang 2002;80(2):142–159
23. Thompson CK, Tait ME, Ballard KJ, Fix SC.
Agrammatic aphasic subjects’ comprehension of
subject and object extracted Wh questions. Brain
Lang 1999;67(3):169–187
24. Macwhinney B, Fromm D, Forbes M, Holland A.
AphasiaBank: methods for studying discourse.
Aphasiology 2011;25(11):1286–1307
25. den Ouden D, Malyutina S, Richardson JD. Verb
Argument Structure in Narrative Speech: Mining
the AphasiaBank. Tuscon, AZ: Academy of Aphasia; 2015
26. Pinker S. Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, MA:
The MIT Press; 1989
27. Clark EV. Discovering what words can do. In:
Farkas D, Jacobsen WM, Todrys KW, eds. Paper
presented at: Paracession on the Lexicon, Chicago
Linguistics Society; April 14–15, 1978; Chicago, IL
28. Ninio A. Model learning in syntactic development:
intransitive verbs. Int J Biling 1999;3:111–131
29. Hollebrandse B, van Hout A. Aspectual bootstrapping via light verbs. In: Weissenborn J, Dittmar N,
Penner Z eds. Issues in the Theory of Language
Acquisition: Essays in Honor of Jürgen Weissenborn:. Bern, Switzerland: P. Lang; 1998
30. Maouene J, Laakso A, Smith LB. Object associations of early-learned light and heavy English verbs.
First Lang 2011;31(1):109–132
31. Gordon JK, Dell GS. Learning to divide the labor:
an account of deficits in light and heavy verb
production. Cogn Sci 2003;27(1):1–40
2016
32. Breedin SD, Saffran EM, Schwartz MF. Semantic
factors in verb retrieval: an effect of complexity.
Brain Lang 1998;63(1):1–31
33. Barde LHF, Schwartz MF, Boronat CB. Semantic
weight and verb retrieval in aphasia. Brain and
Language 2006;97(3):266–278
34. Hoffman P, Rogers TT, Ralph MA. Semantic
diversity accounts for the “missing” word frequency
effect in stroke aphasia: insights using a novel
method to quantify contextual variability in meaning. J Cogn Neurosci 2011;23(9):2432–2446
35. Mirman D, Britt AE. What we talk about when we
talk about access deficits. Philos Trans R Soc Lond
B Biol Sci 2013;369(1634):20120388
36. Templin MC. Certain Language Skills in
Children: Their Development and Interrelationships. Minneapolis, MN: University of Minnesota
Press; 1957
37. Lee L. Developmental Sentence Analysis: A
Grammatical Assessment Procedure for Speech
and Language Clinicians. Evanston, IL: Northwestern University Press; 1974
38. Malvern D, Richards BJ. A new measure of lexical
diversity. In: Ryan, Wray A, eds. Evolving Models
of Language. Clevedon, UK: Multilingual Matters;
1997:58–71
39. Turner A, Greene E. (1977). The Construction
and Use of a Propositional Text Base. Boulder, CO:
Institute for the Study of Intellectual Behavior,
University of Colorado; 1977
40. MacWhinney B, Fromm D, Holland A, Forbes M,
Wright H. Automated analysis of the Cinderella
story. Aphasiology 2010;24(6–8):856–868
41. McCarthy PM, Jarvis S. Voc-D: a theoretical and
empirical evaluation. Lang Test 2007;24(4):459–488
42. Berndt RS, Haendiges AN, Mitchum CC, Sandson
J. Verb retrieval in aphasia. 2. Relationship to sentence processing. Brain Lang 1997;56(1):107–137
43. Kemper S, Greiner LH, Marquis JG, Prenovost K,
Mitzner TL. Language decline across the life span:
findings from the Nun Study. Psychol Aging 2001;
16(2):227–239
44. Snowdon DA, Greiner LH, Markesbery WR.
Linguistic ability in early life and the neuropathology of Alzheimer’s disease and cerebrovascular
disease. Findings from the Nun Study. Ann N Y
Acad Sci 2000;903:34–38
45. Kertesz A. Western Aphasia Battery-Revised. San
Antonio, TX: Pearson; 2006
46. Goodglass H, Kaplan E, Barresi B. Boston Diagnostic Aphasia Examination, 3rd ed. Philadelphia,
PA: Lippincott Williams and Wilkins; 2001
47. Cho-Reyes S, Thompson CK. Verb and sentence
production and comprehension in aphasia: Northwestern Assessment of Verbs and Sentences
(NAVS). Aphasiology 2012;26(10):1250–1277
48. Fergadiotis G, Wright HH. Lexical diversity for
adults with and without aphasia across discourse
VERB PRODUCTION IN APHASIA/THORNE, FAROQI-SHAH
49.
50.
51.
52.
52.
53.
54.
55.
elicitation
tasks.
Aphasiology
2011;25(11):
1414–1430
MacWhinney B. The CHILDES Project: Tools
for Analyzing Talk, Vol. 1: Transcription Format
and Programs. Mahwah, NJ: Erlbaum; 2000
Wilson MD. The MRC psycholinguistic database:
machine readable dictionary, version 2. Behav Res
Methods Instrum Comput 1988;20(1):6–11
Davis M. The Corpus of Contemporary American
English: 450 million words, 1990 present. Retrieved in 2008 from http://corpus.byu.edu/coca/
Grodzinsky Y. Language deficits and the theory of
syntax. Brain Lang 1986;27(1):135–159
Grodzinsky Y. Language deficits and the theory of
syntax. Brain Lang 1986;27(1):135–159
Goodglass H, Christiansen JA, Gallagher R. Comparison of morphology and syntax in free narrative
and structured tests: fluent vs. nonfluent aphasics.
Cortex 1993;29(3):377–407
Brown C, Snodgrass T, Kemper SJ, Herman R,
Covington MA. Automatic measurement of propositional idea density from part-of-speech tagging.
Behav Res Methods 2008;40(2):540–545
Bird H, Lambon Ralph MA, Patterson K, Hodges
JR. The rise and fall of frequency and imageability:
noun and verb production in semantic dementia.
Brain Lang 2000;73(1):17–49
APPENDIX
Selected Transcripts of the Cinderella
Story
The following two transcripts illustrate the
almost exclusive use of either heavy or light
verbs, and the associated syntactic wellformedness. Error codes in transcripts have
been deleted to improve readability. Verbs are
italicized.
Participant Scale02a, Using Mainly
Heavy Verbs
Um, middle-aged woman and, um,
mid-uh, early twenties um, uh, no, um,
10 years old. . . . And um, older gentleman.
And uh, a paint, a painting sunlight, beautiful sunlight. . . . Uh, next, um, um, next
um, a bad um, um, older woman and um,
mid-twenties dark, dark, stern fighting and
lovely uh, girl. . . . Um, um, older women
and um, mid-twenties um, uh, cursing and
stuff. . . . Next um, um, fairy, uh, father
56. Arevalo A, Perani D, Cappa SF, Butler A,
Bates E, Dronkers N. Action and object processing in aphasia: from nouns and verbs to the effect
of manipulability. Brain Lang 2007;100(1):
79–94
57. Vinson DP, Vigliocco G. A semantic analysis of
grammatical class impairments: semantic representations of object nouns, action nouns and
action verbs. J Neurolinguist 2002;15(3–5):
317–351
58. Williams SE, Canter GJ. The influence of situational context on naming performance in aphasic
syndromes. Brain Lang 1982;17(1):92–106
59. Pashek GV, Tompkins CA. Context and word
class influences on lexical retrieval in aphasia.
Aphasiology 2002;16(3):261–286
60. Mayer J, Murray L. Functional measures of naming
in aphasia: word retrieval in confrontation naming
versus connected speech. Aphasiology 2003;17(5):
481–497
61. Scarborough HS. Index of productive syntax. Appl
Psycholinguist 1990;11:1–20
62. Edmonds LA, Babb M. Effect of verb network
strengthening treatment in moderate-to-severe
aphasia. Am J Speech Lang Pathol 2011;20(2):
131–145
[fairy godmother] uh . . . Um, cot [god] . . .
A woman, uh, father [fairy godmother].
Next um, horses and um, uh, a chariot
um, uh, horses and ride to the uh, castle.
And um, uh, gaily um, um, waltzing the,
what called, polka and um, uh, music.
Participant Star03a, Producing Only
Light Verbs
Uh, the story was um, (a) bout
Cinderella. And she uh, was uh, having
a different stuff uh, with the maid of the
house. And the, the lady in the house uh,
with uh, her two uh, thing were, were
angry at Cinderella. And so uh, then
they had a uh, uh, difference with Cinderella. And Cinderella had uh, some differences with them. And uh, uh, so, and uh,
Cinderella had a big uh, um, big deal with
uh, the prince. And Cinderella was doing
okay. But then she was, she had their
deal... out. And she had, the uh, prince,
uh had them. All of them had the table,
table and, and so forth.
33