webster et al thematic structure

8/8/2019 Webster Et Al Thematic Structure

1/23


2/23


3/23


4/23

(1997) describe these conceptual conditions as the core meaning of the verb. Alongside

this core meaning, Marshall (1995) suggests the semantic representation also contains

semantic selection restrictions that constrain the lexical items that can fulfil particular

arguments. For example, for the verb ``to eat'', the lexical item fulfilling the thematicrole of Patient must be edible, and the Agent must be animate. Semantic deficits affect

both single word and sentence processing and as semantic information is considered

central to both production and comprehension, semantic impairments will affect both

modalities.

In comprehension, semantic deficits result in the confusion of items closely related in

meaning. On testing, people therefore choose distractors that are semantically related to

the target word (Berndt, Mitchum, Haendiges, & Sandson, 1997b). In production,

semantic deficits result in semantic paraphasias in single word and sentence contexts. In

addition, verb deficits may also result in a reliance on single phrases with limited use of

sentence structure, the omission of the verb, or the production of light verbs (Berndt,

Haendiges, Mitchum, & Sandson, 1997a; Breedin & Martin, 1996). Light verbs are high-frequency verbs with low semantic content (like ``do'', ``get'', and ``have'') that

resemble auxiliaries. If specific nouns cannot be retrieved within the sentence frame,

there may be an excessive reliance on pronouns (Gleason, Goodglass, Obler, Green,

Hyde, & Weintraub, 1980) or obligatory arguments may be omitted as the words cannot

be produced.

Impairments in the production of nouns and verbs may also be due to a deficit in their

phonological production (Caramazza & Hillis, 1991); this would not affect comprehen-

sion and should have a reduced impact on sentence production. Clients with phonological

impairments are generally able to produce a sentence, with the nouns in the correct order

even if the verb is omitted (Fink, Martin, Schwartz, Saffran, & Myers, 1992). Obligatory

arguments may, however, still be omitted if the words constituting those argumentscannot be accessed.

As stated previously, the PAS specifies the number and type of arguments associated

with a verb. Verbs differ in terms of their PAS; they differ in the number of arguments,

the thematic roles specified by those arguments, and the number of different PAS

arrangements. Some verbs have obligatory (fixed) argument structures, i.e., they can only

be used in a particular argument structure. For example, ``admire'' can only be used in a

two-argument structure ``the woman admired the painting''. Other verbs have variable

(optional) argument structures, i.e., they can be used in more than one PAS arrangement.

For example, ``bake'' can be used in a one-argument ``she is baking'' or a two-argument

``the boy is baking biscuits'' structure. It is difficult to obtain information about the

number and type of PAS arrangements associated with a particular verb.In addition to the production of the PAS, sentences often contain non-arguments. Non-

arguments are not specified by the verb; they give additional information about the verb,

the participants within the situation, and the perspective of the speaker (Black & Chiat,

2003). Non-arguments typically give information about time, manner, or place. By

definition, they never specify obligatory information and can occur alongside any verb. It is

sometimes difficult to distinguish between arguments and non-arguments within a sen-

tence. Garrett's model gives no account of the production of non-arguments; the pro-

duction of the functional-level representation revolves only around the PAS. Argument

status has been shown to influence processing, although results are currently mixed

regarding whether arguments or non-arguments are preferentially processed. Shapiro and

Levine (1990) reported increased retention of verb arguments, whereas Byng and Black(1989) suggested that non-arguments were easier for some people with aphasia to produce.

PRODUCTION OF THEMATIC STRUCTURE 49


5/23

Thompson, Lange, Schneider, and Shapiro (1997) investigated the production of PAS

by subjects with aphasia during conversational tasks. They concluded that compared with

normal subjects, subjects with aphasia produced verbs with simple argument structures,

in terms of fewer PAS arrangements and fewer arguments. Subjects were unable to produce all of the argument structure arrangements associated with a verb, particularly

those requiring sentential complements. Systematic investigation of PAS production in

more constrained tasks is more limited. It is difficult to assess access to PAS information

and the construction of the argument structure without relying on lexical selection,

thematic role assignment, and the subsequent mapping onto syntactic structure.

The predicted impact of PAS difficulties depends on whether the creation of the PAS

relies only on lexically specified information or is a distinct process that also uses

information from the conceptual representation, as suggested by Schwartz. In both cases,

problems with the PAS may result in a limited use of sentence structure, the omission of

obligatory arguments, the addition of inappropriate arguments, or the production of

arguments fulfilling inappropriate thematic roles. If the creation of the PAS relies only onlexically specified information then sentence production difficulty will be related to the

verbs used and should not be influenced by the task. If the creation of the PAS is

influenced by the conceptual representation, then task presentation may also affect per-

formance.

Levelt (1989) suggests that PAS information is part of the verb's semantic repre-

sentation. As a consequence, it has been proposed that semantic verb deficits have a more

disruptive effect on sentence production than phonological verb deficits (Berndt et al.,

1997a). Berndt et al. (1997b) thus suggest that giving the verb in spoken and written form

(giving the verb's semantic representation) should provide access to PAS information and

facilitate more accurate sentence production. As semantic representations are considered

central to comprehension and production, problems in sentence production would beaccompanied by comprehension difficulties, as the person with aphasia no longer has

access to the number and type of arguments that are needed alongside the verb. The

central nature of PAS information has been suggested by the results of a therapy study

targeting access to PAS information (Webster & Whitworth, 2001). Therapy targeting

comprehension resulted in significant gains in sentence production. However, therapy

focused on improving meta-linguistic awareness and it is difficult to determine whether it

was this that resulted in the generalisation from comprehension to production, or the

central nature of PAS information. If the production of PAS is a distinct process, dis-

sociations may be seen between access to a verb's semantic representation and the

creation of the argument structure. In addition, giving the verb may not result in more

accurate sentence production. If lexically specified information is used alongside con-ceptual information in the creation of the PAS, task constraints may influence perfor-

mance. In spontaneous speech and when given only the verb, the person with aphasia

must generate the idea themselves and that is the only cue to the argument structure. In

constrained picture tasks, the picture depicts the participants in the event and this may

provide a clue to the number of arguments within the sentence (thus supporting sentence

production to a much greater extent).

Thematic role assignment associates lexical items with thematic roles specified in the

PAS. This process places items in the form of ``who is doing what to whom/what''

(Whitworth, 1994). In non-reversible sentences, e.g., ``the man drives the car'', real-

world knowledge and semantic selection restrictions govern the thematic roles that

``man'' and ``car'' must fulfil. In the production of reversible sentences, e.g., ``the boypushes the girl'', such information is not available as the items can fulfil more than one

50 WEBSTER, FRANKLIN, HOWARD


6/23

role. Deficits in thematic role assignment are thus seen most clearly in the production of

reversible sentences, with the production of reverse role errors; e.g., for the target sen-

tence ``the boy is pushing the girl'', ``the girl is pushing the boy'' is produced. Thematic

role assignment also becomes more complex as the number of arguments within thesentence increases. Thematic role assignment deficits thus also result in the omission of

arguments in sentences with an increased number of obligatory arguments (Schwartz,

Fink, & Saffran, 1995; Whitworth, 1994).

Thematic role assignment precedes the specification of word order; the same

functional-level representation can be translated into sentences with different surface

forms. It is, however, difficult to assess thematic role assignment without relying on

the subsequent mapping of this thematic form onto syntactic form. Mapping deter-

mines how the thematic roles are realised within syntactic structure. For some sen-

tences, mapping is transparent as the sentences have canonical word order, e.g., in an

active sentence, the thematic role of Agent becomes the subject of the sentence. In

other sentences, e.g., passives, the mapping process is more complex as the sentenceshave non-canonical word order (Schwartz, Linebarger, Saffran, & Pate, 1987). As with

thematic role assignment, mapping deficits are reported to result in difficulties with

reversible sentences (Schwartz et al., 1980). Byng (1988) suggested that mapping diffi-

culties also result in verb comprehension deficits for reverse role verbs; verbs that dif-

fer in the way they map their thematic roles onto syntax, for example, ``buy'' and

``sell''.

Within the current literature, the distinction between thematic role assignment and

mapping is not clearcut. Terms are often used interchangeably, testing has not been

specific enough to distinguish deficits, and therapy has often targeted both processes.

Within the studies, however, it has been proposed that thematic role assignment and

mapping are central processes involved in sentence comprehension and production,resulting in parallel deficits (Jones, 1986; Saffran et al., 1980; Schwartz et al., 1980).

Reverse role errors in production are mirrored in the selection of reverse role distractors

in sentence comprehension. There is also significant evidence that sentence production

improves as a consequence of therapy targeted at improving thematic role assignment and

mapping in comprehension (for example, Jones, 1986; Byng, 1988).

The systematic assessment of subjects with aphasia has shown selective deficits in

some aspects of performance. Subjects LK and JS, described by Breedin and Martin

(1996), showed a double dissociation between access to semantic information and the-

matic role information. Similarly, subject PB, described by Marshall et al. (1997) showed

retained access to semantic information despite impaired thematic role assignment.

Studies have also shown selective deficits in access to syntactic sub-categorisationinformation (Breedin & Martin, 1996; Marshall, Pring, & Chiat, 1998). However, no

studies have attempted to investigate the independence of the processing of the predicate

argument structure.

AIMS OF STUDY

This study compared the performance of four people with aphasia who had difficulties

producing the thematic structure of sentences. They were assessed on tests of single word

and sentence production and comprehension. In this way, the processes responsible for

the production of thematic structure were investigated in order to see whether discrete

deficits could be identified. If discrete deficits were identified, this would provide somesupport for the sub-processes proposed by Schwartz (1987).



7/23

METHOD

Subjects

Four people with aphasia participated in this study; these are described in Table 1. GW,KD, and TJ were aphasic following a single left hemisphere CVA. JM's aphasia resulted

from surgery to clip a left middle cerebral artery aneurysm. They were at least 6 months

post-onset at the time of testing and had sufficiently good functional comprehension to

understand task instructions. They were selected from an analysis of their sentence

production during narrative speech (Webster, 1999). Within this study, the performance

of the people with aphasia was compared to a group of 20 normal controls (mean age 54.8

years). A comprehensive analysis of their production of thematic, syntactic, and phrasal

structure was carried out. The people with aphasia were considered to differ from the

normal group on each parameter if their score fell outside two standard deviations of the

normal mean, GW, JM, KD, and TJ differed significantly from the normal control group

on measures related to the production of thematic structure. They all had a lower meanthematic complexity score than the normal group. The mean thematic complexity score

was a weighted mean which reflected the range and complexity of the argument struc-

tures produced (see Webster et al., 2001). JM, KD, and TJ produced an increased pro-

portion of utterances with an undetermined thematic structure (single words and phrases

containing no verb) and a reduced proportion of complex three argument structures. None

of the people with aphasia produced complex utterances with thematic embedding. GW

produced a higher percentage of utterances where obligatory arguments had been omit-

ted. Table 2 compares the results on these parameters for GW, JM, KD, and TJ with those

of the normal group.

TABLE 1

Client details

Subject Sex Age when first

tested

Time post-onset when

first tested

Previous occupation

GW M 41 8 years Aerial rigger

JM F 62 18 months Psychiatric sister

KD M 49 6 months Shipyard worker

TJ M 57 2 years University lecturer

TABLE 2

Production of thematic structure

Normal mean

score

Normal range

(2 s.d.)

GW JM KD TJ

Mean thematic complexity 3.15 2.743.56 2.47* 2.56* 2.00* 1.17*

% Undetermined thematic structure 2.54 08.45 2.35 12.5* 46.15* 91.3*

% 1 argument 12.83 2.9122.75 14.71 21.88 10.26 0

% 2 argument 58.02 41.3774.67 52.94 62.50 41.03 8.7*

% 3 argument 20.28 7.4833.08 8.82 3.13* 0.03* 0*

% Thematic embedding 6.33 016 0 0 0 0

% Omission of obligatory arguments 0.15 01.09 8.82* 0 0 0

* Indicates outside normal range (2 s.d. from the mean of the normal group).



8/23

Procedure

The people with aphasia were tested on a battery of published tests, unpublished tests,

and tasks specifically designed for this study. For these tasks, normal control subjectswere recruited from a group of volunteers aged from 49 to 85; they had no history of

brain damage and showed no signs of dementia. Not all of the normal control subjects did

all of the assessment tasks. Tests of single word and sentence production were accom-

panied by comprehension, anagram, and grammaticality judgement tasks in order to

investigate the relationship between comprehension and production. Due to the difficulty

of assessing processes independently of one another (this will be considered further in the

discussion), performance on systematically varied tasks was contrasted. Within this

study, deficits in thematic role assignment and mapping were not distinguished due to the

difficulty of eliciting sentences with non-canonical word order.

Description of assessments usedAccess to semantic information for nouns and verbs was assessed using single word

retrieval and comprehension tasks. The Verb and Noun (VAN) Test (Webster & Bird,

2000) was used to compare noun and verb retrieval. A short video clip of an action or

object was shown and people were asked to describe what was happening (verbs) or what

it was (nouns). In both cases, people were asked to respond using one word. Performance

of the people with aphasia was compared to that of a group of 17 normal control subjects

(aged 4969). The comprehension of nouns was assessed using the spoken and written

word to picture matching sections of the Comprehensive Aphasia Test (CAT, Swinburn,

Porter, & Howard, in press). Verb comprehension was investigated using the verb

comprehension video described in Byng (1988). Video depictions of the target verb and

an unrelated or related distractor were presented on a split screen and participants wereasked to match a spoken word to the corresponding depiction. The video contrasted three

types of verb: reverse role verbs, for example ``buy'' and ``sell'', reverse action verbs, for

example ``throw'' and ``catch'', and reverse direction verbs, for example ``fall'' and

``rise''.

Access to PAS information, thematic role assignment, and mapping in sentence

production were assessed by contrasting production of one-, two-, and three-argument

structures with picture stimuli in Thematic Roles in Production (TRIP, Whitworth, 1996)

and without picture stimuli in a sentence generation task. In the sentence generation task,

subjects were given a verb and asked to produce a sentence. No picture stimuli were

given. The stem form of the verbs were presented randomly in written form and read

aloud by the researcher. The subjects were asked to produce a sentence containing the``action word'' in any form. A total of 74 verbs were divided into five groups depending

on the argument structures associated with the verb. The five groups were:

(i) Group 1. 15 verbs with an obligatory one-argument structure, e.g., sunbathe,

die.

(ii) Group 2. 15 verbs with an obligatory two-argument structure, e.g., admire, enjoy.

(iii) Group 3. 14 verbs with optional two- and three-argument structures, e.g., blame,

instruct.

(iv) Group 4. 15 verbs with optional one- and two-argument structures, e.g., bake,

marry.

(v) Group 5. 15 verbs with optional one-, two-, and three-argument structures, e.g.,write, teach.



9/23

The verb classifications used were taken from the syntactic classifications found in the

Celex database (Baayen, Piepenbrock, & Gulikers, 1995) and confirmed using the dic-

tionary. Due to the difficulty in finding verbs, groups of verbs were not matched for

frequency or length. The groups of verbs varied quite significantly in mean frequency andlength; this seemed to reflect the fact that the more possible verb argument arrangements,

the greater frequency of use and the shorter the word. The mean frequency of the verbs in

each group and their mean length in syllables and phonemes are shown in Table 3.

Three scores were calculated for each of the people with aphasia and compared to the

results of a group of 22 normal subjects (aged 2471). First, the number of sentences

containing the target word as a verb (Target as Verb) was calculated. Verbs that produced

no response or sentences that did not contain the target as a verb were excluded from the

rest of the analysis. Normal subjects produced an average of 72.9 sentences containing

the target as verb. Second, the percentage of sentences with an inappropriate argument

structure (due to the omission of arguments or the presence of an inappropriate argument)

was calculated. Normal subjects produced a very low percentage of sentences with aninappropriate PAS (mean 0.90%, range 04.1%). Finally, the percentage of sentences

containing non-arguments was calculated. This was the percentage of sentences with an

appropriate argument structure that also contained non-arguments. Normal subjects

produced a high percentage of sentences with co-occurring non-arguments (mean 35.8,

range 21.952.1%). Non-arguments were normally produced alongside optional and

obligatory one- and two-argument structures. Following an analysis of all the verbs,

particular difficulties with a group of verbs was noted. The 45 verbs from groups one,

two, and four from the sentence generation task were then also used in an anagram task.

Phrases that formed appropriate arguments were presented alongside distractor phrases

and the people with aphasia were asked to produce an appropriate sentence using some or

all of the phrases.The TRIP assessment contrasted the retrieval of high-frequency nouns in isolation and

in one-, two-, and three-argument structures. The assessment is a delayed repetition task

so all of the stimuli were modelled prior to the person with aphasia being given the

picture and asked to describe what was happening. Sentences were scored for the retrieval

of nouns and verbs and the thematic completeness of the sentence. The normal control

group in the original study scored 95100% on all measures. In a separate task, the people

with aphasia were presented with the TRIP sentence pictures and asked to order the

sentence components in an anagram task. The anagram task assessed the ability to assign

TABLE 3Mean frequency and length of verb types

Verb Type Obligatory 1

argument

Obligatory 2

argument

Optional 2 and 3

argument

Optional 1 and 2

argument

Optional 1, 2,

and 3 argument

Mean

frequency

1.27

(0.72)

1.60

(0.55)

1.27

(0.46)

1.76

(0.76)

2.12

(0.77)

Mean length

syllables

2.33

(0.98)

2.13

(0.64)

1.93

(0.59)

1.67

(0.72)

1.27

(0.46)

Mean length

phonemes

5.40

(2.00)

5.33

(1.23)

5.13

(1.73)

4.33

(1.91)

3.60

(1.24)

Frequency of the lemma in log frequency per million.Standard deviations in brackets.



10/23


11/23

from that of the group of normal controls. Chi-square tests were used to look at each

individual's performance across conditions within a single assessment. Following a

description of results for each person, a brief interpretation of their difficulties will be

given.

Results for GW

GW's single word retrieval was characterised by retained access to nouns but impaired

access to verbs. His verb retrieval (37/54) differed significantly from normal subjects,

t(16) = 5.13, p < .001, and was characterised by the production of a noun semanticallyrelated to the target verb, e.g., ``hands'' for ``holding''. His comprehension of nouns on

TABLE 4

Predicted patterns of performance

Test Semantic deficit PAS deficit Thematic role assignment and/or mapping deficit

1. Verb and noun test Impaired retrieval of

verbs and/or nouns

Retained Retained

2. CAT comprehension Selection of semantic

distractors

Retained Retained

3. Verb video Selection of semantic

distractors

Retained Possible difficulty with

reverse role verbs

4. Sentence generation

task

Noun deficits may result

in the omission of

arguments. Verb deficits

may be bypassed due toprovision of verb. No

impairment in anagram

version

Inappropriate PAS

structures. Reliance on

simple argument

structures. Difficulty withmore complex argument

structures. Remains

difficult with anagrams

Omission of arguments in

complex argument

structures

5. Thematic roles in

production

Verb deficits may be

evident. Effect of noun

retrieval deficits

minimised by stimuli. No

impairment in anagram

version

Difficulty with more

complex argument

structures


more complex argument

structures. Difficulty with

reversible sentences

6. Production of


Noun and verb deficits

may be evident in free

production but no

impairment when given

lexical items

Difficulty with more

complex argument

structures. Reduced

problems when given the

lexical items


more complex argument

structures. Reverse role

errors present

7. Comprehension of


Selection of lexical

distractors

Retained Selection of reverse role

distractors

8. Grammaticality

judgement

a) Semantic anomalies Accepts semantic

anomalies

Able to reject semantic

anomalies

Able to reject semantic

anomalies

b) Mapping anomalies Able to reject mapping

anomalies

Able to reject mapping

anomalies

Accepts mapping

anomalies

c) PAS anomalies Able to reject PAS

anomalies

Accepts PAS anomalies Able to reject PAS

anomalies



12/23


13/23


14/23

w2 (3) = 9.1, p = .027. He showed good retrieval of nouns in isolation but omitted words

and produced semantic paraphasias in sentences. The semantic paraphasias all involved

the production of ``man'', ``boy'', ``girl'', and ``woman''. His verb retrieval in sentences

was also characterised by semantic paraphasias. GW omitted arguments and producedsentences with a simplified argument structure (generally two-argument structures

instead of three-argument structures); this resulted in reduced thematic completeness

scores. He produced two reverse role errors for two of the two-argument structures, e.g.,

``the horse is pushed by the man'' instead of the ``horse is pushing the man''. He also

made one reverse role error in the TRIP anagram task.

GW's production of reversible sentences in the free production condition resembled

his performance on TRIP; he produced reverse role errors and omitted obligatory argu-

ments. His reverse role errors were always attempts at passive sentences. The omission of

arguments occurred in both two- and three-argument structures, e.g., ``the boy is

throwing to the girl'' instead of ``the boy is throwing the ball to the girl''. When given the

lexical items, his omission of arguments was eliminated but he continued to producereverse role errors in the production of two-argument structures. He made no errors in the

production of three-argument structures.

GW scored 42/70 on the reversible sentence comprehension test, selecting a higher

proportion of reverse role compared to lexical distractors (Binomial test, p = .002). He

made reverse role errors on active, passive, and locative sentences. GW differed from the

normal controls on all of the grammaticality judgement tasks: semantic anomalies, t(17)

= 3.44, p = .003, mapping anomalies, t(17) = 9.28, p < .001, PAS anomalies, t(13) = 7.05,

p < .001. In each case, his impaired performance was a consequence of an increased

number of false positive responses (acceptance of a sentence with an anomaly as correct).

Interpretation. GW was impaired in his ability to retrieve verbs and understandthem. This may indicate a central semantic impairment. His difficulties in rejecting

semantic anomalies in the grammaticality judgement tasks may also reflect these

semantic difficulties. His verb retrieval difficulties were evident in single word and

sentence contexts. GW experienced difficulty in producing the right arguments for verbs,

both in terms of the number of arguments and their appropriateness. The majority of his

errors involved the omission of obligatory arguments. GW did not have difficulty with

noun retrieval in single word tasks. His word retrieval difficulties in sentences are,

therefore, likely to be due to PAS and/or thematic role assignment difficulties. A deficit

in accessing PAS information is suggested by his performance on the PAS anagram and

grammaticality judgement tasks. GW did not have access to knowledge about what

arguments were required alongside a verb. However, GW also seems to have a deficit ineither thematic role assignment and/or mapping. He had difficulties with reversible

sentences in both production and comprehension and failed to reject mapping anomalies.

The omission of obligatory arguments in sentence production and his reliance on simple

argument structures is likely to reflect this combination of PAS and thematic role

assignment difficulties. GW thus has difficulties with all three of the sub-processes

involved in the production of thematic structure.

Results for JM

JM, like GW, differed significantly from the normal group in her retrieval of verbs, t(16)

= 3.97, p = .001, but showed no difficulty with nouns. She produced a combination ofsemantic and phonemic paraphasias. She did not have any difficulties with the com-



15/23

prehension of nouns on the CAT and made only one error on the verb video. In the

sentence generation task, she was able to produce the target as a verb in a sentence but

produced 12.5% of sentences with an inappropriate argument structure. JM occasionally

omitted both internal and external arguments in two-argument structures but the majorityof her errors were the production of an additional inappropriate phrase with obligatory

one-argument verbs, e.g., ``the baby is crawling the floor'', ``we disagree the fish tank''.

She produced a range of one-, two-, and three-argument structures and produced a similar

percentage of non-arguments as the normal subjects. JM performed poorly on the ana-

gram task, selecting appropriate arguments to produce a sentence on only four occasions.

She could not reject the distractor arguments and despite prompting tried to produce a

sentence containing all of the components.

Figure 2 shows JM's performance on TRIP. She successfully retrieved the nouns in

isolation and in sentences. She produced semantic paraphasias for some of the target

verbs in sentences. She generally produced thematically complete and appropriate sen-

tences, although she produced additional, inappropriate arguments for two of the one-argument structures, e.g., ``the boy swimming the sea'' instead of ``the boy is swim-

ming''. She made no errors on the anagram version of TRIP. In the production of

reversible sentences, she made only one error, abandoning the sentence due to verb

retrieval difficulties.

JM scored 63/70 on the reversible sentence comprehension test, selecting a mixture of

reverse role and lexical distractors across sentence types. On the grammaticality judge-

ment tasks, she was able to identify mapping anomalies but differed from normal controls

Figure 2. JM's performance on TRIP.



16/23


17/23

sentence, e.g., ``I inform . . . I inform . . . no no I inform''. Following multiple repetitions

and apparent difficulty in retrieving an appropriate word, he would give up and accept the

sentence he had produced. KD produced only one- and two-argument structures and

produced only a small percentage of utterances containing non-arguments. When given aselection of arguments in the anagram task, KD always produced an appropriate sentence.

Figure 3 shows KD's production of sentences on TRIP. KD successfully retrieved the

nouns in isolation but had difficulty in retrieving the nouns in sentences, w2 (3) = 44.84,

p < .001. His errors in sentences were a mixture of semantic paraphasias, omissions,

production of the nouns in isolation, and the use of pronouns. Sentences were not the-

matically complete due to the omission of verbs and a failure to create an argument

structure, e.g., ``the sheep . . . bread . . . bread . . . the man . . . the given . . . the no girl''.

These difficulties were evident in the production of two- and three-argument structures.

He also made a reverse role error in the production of a two-argument structure and four

reverse role errors in the TRIP anagram task. Similarly, in the production of the reversible

sentences, he produced reverse role errors on the two-argument structures in the free production and lexical conditions. He made no errors on the three-argument structures.

KD showed some impairment in the comprehension of reversible sentences (score 57/70).

He selected a higher proportion of reverse role distractors compared to lexical distractors

(Binomial test, p = .0025). His errors were predominantly in the comprehension of

passive and locative sentences. KD's performance on the grammaticality judgement tasks

was comparable to normal subjects; he was able to identify semantic, mapping, and PAS

anomalies accurately.

Figure 3. KD's performance on TRIP.



18/23

Interpretation. KD has word retrieval difficulties affecting both nouns and verbs but

retained comprehension. He produced some phonemic paraphasias, suggesting that a

phonological impairment may be contributing to his word retrieval difficulties. He

experienced no difficulty in detecting semantic or PAS anomalies in grammaticalityjudgement tasks. KD's difficulty in producing an appropriate argument structure seemed

to be the result of his word retrieval difficulties. He seemed aware that the sentences he

produced were not complete, sometimes omitting the argument and sometimes producing

only a determiner and omitting the main noun. His word retrieval difficulties may also

account for the low percentage of utterances containing non-arguments. When given

possible arguments in the anagram task, he could select and order the arguments

appropriately, again suggesting intact access to PAS information. In addition, however,

some of the omitted arguments may be a consequence of thematic role assignment and/or

mapping difficulties. KD omitted arguments in sentences in TRIP, despite being able to

retrieve the same nouns in isolation. He also produced some reverse role errors in sen-

tence production and selected reverse role distractors in the sentence comprehension task.He did not, however, have difficulty in identifying mapping anomalies in the gramma-

ticality judgement task. KD thus has a problem with thematic role assignment and/or

mapping that is resulting in poor production of thematic structure. In addition, he presents

with word retrieval difficulties that are in part due to a difficulty in accessing phonology.

Results for TJ

TJ, like KD, showed an impairment in both noun and verb retrieval on the VAN: verbs,

t(16) = 7.44, p < .001; nouns t(16) = 5.01, p < .001. Nouns were retrieved more accurately

than verbs, w2 (1) = 5.83, p = .016. He produced semantic paraphasias for both nouns and

verbs, e.g., ``pen'' for ``pencil'', ``viewing'' for ``watching''. TJ scored within thenormal range on noun comprehension in the CAT but made seven errors on the verb

comprehension video. He selected semantic distractors on all three verb types.

TJ also showed significant difficulties in the retrieval of nouns and verbs in sentences.

Within the sentence generation task, he differed from the normal control subjects in his

ability to use the target as a verb in a sentence, t(21) = 7.86, p < .001, and in the

production of sentences with an appropriate argument structure. TJ's errors were pre-

dominantly the omission of external arguments with optional two- and three-argument

verbs. Like KD, he also produced sentences in which the internal argument consisted

only of a determiner. TJ produced one- and two-argument sentences and a low proportion

of sentences containing non-arguments. He had no difficulties selecting arguments to

produce a sentence in the anagram version of the task.Figure 4 shows TJ's performance on TRIP. He showed a significantly different pattern

of noun retrieval in single word and sentence contexts, w2 (3) = 43.01, p < .001. There

was also a significant difference between one-, two-, and three-argument structures, w2

(2) = 19.86, p < .001; he made more errors as the complexity of the argument structure

increased. Similarly, his thematic completeness score deteriorated as the complexity of

the argument structure increased, w2 (2) = 13.30, p = .001. TJ's difficulty in producing

nouns within sentences and producing thematically complete sentences seemed to stem

from his failure to retrieve an appropriate verb. Sentences were abandoned at the point of

the verb and thus some of the verb's arguments were omitted, e.g., for ``the girl kicked

the snake'' he said ``the snake is . . . the snake is . . . no''. When given the lexical items in

the anagram version, TJ was able to order the words to produce the target sentence. TJwas unable to attempt the two- and three-argument reversible sentences in the free



19/23

production condition. When given the lexical items, he still showed some difficulty in

producing the words due to perseveration on previous items. However, he did not produce

reverse role errors.

In the comprehension of reversible sentences, TJ scored 60/70, making a combination

of reverse role and lexical errors. He showed particular difficulty on locative sentences.

On the grammaticality judgement tasks, he was able to identify semantic, mapping, and

PAS anomalies, performing within the normal range.

Interpretation. TJ, like KD, experienced difficulty with both noun and verb retrievalbut also made errors in the comprehension of verbs. However, he was able to identify

semantic anomalies in the grammaticality judgement task. TJ's noun retrieval difficulties

seem to account for his omission of obligatory arguments in sentence production. On the

sentence generation task, he seemed aware that arguments were necessary but was unable

to produce words to realise those arguments. When given a selection of arguments, he

was able to produce an appropriate sentence and he was able to identify PAS anomalies in

the grammaticality judgement task. This would indicate intact access to PAS information.

In TRIP when he was not given the verb, his word retrieval difficulties often resulted in a

complete failure to produce a sentence. When given the lexical items in an anagram task,

again he had no difficulty in producing the sentence. Although he made a number of

errors in sentence comprehension, particularly with locative sentences, there was nodistinct pattern of errors and he was able to detect mapping anomalies in grammaticality

Figure 4. TJ's performance on TRIP.



20/23

judgement. This would suggest that his ability to assign thematic roles and map them

onto syntactic structure is intact (at least for verb predicates). These results suggest that

access to PAS and thematic role assignment are intact, but his word retrieval difficulties

are so severe that he cannot produce thematic structure. These word retrieval difficultiesseem to have a semantic origin.

DISCUSSION

The aim of this study was to investigate the processes involved in the production of the

functional-level representation, via the analysis of the patterns of performance of four

individuals with aphasia, thought to have difficulties in the production of thematic

structure. The performance of the four individuals on a battery of tests was contrasted

with a view to determining the nature of their underlying impairment. The identification

of discrete impairments would provide evidence of sub-processes in the production of

thematic structure. This section will discuss the performance of the four clients as well ashighlighting methodological issues related to the assessments used.

The performance of the four individuals suggests that different underlying impair-

ments are responsible for their poor production of thematic structure. They present with

different patterns of impaired and retained performance, compared to normal subjects,

and when impaired, different patterns of error.

All four individuals exhibited difficulty with verb retrieval. For two, GW and TJ,

the origin of these difficulties was a central semantic impairment resulting in difficul-

ties with both the comprehension and production of verbs. For JM and KD, it

appeared that phonological difficulties were contributing to their difficulties with verb

retrieval. The four individuals varied to the extent that other deficits were present

alongside difficulties with verb retrieval. GW presented with poor access to PAS infor-mation and thematic role assignment and/or mapping difficulties. JM had poor access

to PAS information. KD's performance seemed to result from a combination of word

retrieval difficulties and impaired thematic role assignment and/or mapping. TJ's per-

formance was a consequence of impaired access to semantic information, resulting in

severe word-finding difficulties.

There is evidence from the results of dissociations between various aspects of pro-

cessing. Within the performance of JM and KD, a double dissociation between the

processing of PAS and thematic role assignment and/or mapping is evident. JM had poor

access to PAS information but was able to assign thematic roles and map them onto

syntactic structure. KD had apparently good access to PAS information but made reverse

role errors in production and comprehension, suggesting an impairment in thematic roleassignment and/or mapping. TJ shows access to PAS information and retained thematic

role assignment, despite a deficit in accessing semantic information. These results con-

firm the independence of thematic role assignment from other processes involved in the

production of thematic structure as found by Breedin and Martin (1996) and Marshall et

al. (1997). The dissociations also suggest that PAS processing is a discrete aspect of

production that can be selectively impaired. Despite poor access to the semantic repre-

sentation of the verb, TJ's knowledge of PAS was retained (although he had difficulty

realising the argument structure in spoken production). JM showed poor access to PAS

information and impaired creation of the argument structure despite no semantic diffi-

culties. In none of the clients did just giving the verb result in accurate sentence pro-

duction. There is thus evidence from the above studies and this study that poor productionof thematic structure may be a consequence of semantic deficits, impaired processing of



21/23

PAS or impaired thematic role assignment. This provides evidence for the distinct sub-

processes suggested by Schwartz (1987).

The results of this study also provide evidence for the central nature of PAS infor-

mation and thematic role assignment. GW and JM's deficits in the production of PASwere associated with impaired performance on the anagram and grammaticality judge-

ment tasks. GW and KD's deficits in thematic role assignment (in terms of the production

of reverse role errors) were associated with an increased proportion of reverse role errors

in comprehension. KD did not, however, present with difficulties identifying mapping

anomalies in the grammaticality judgement task.

Determining the precise extent to which PAS and thematic role assignment difficulties

contribute to the omission of arguments and the reliance on simple argument structures is

not straightforward. The impact of JM's difficulty in accessing PAS information was

minimised by the presence of a picture that presumably gave her clues about the number

of arguments. In comprehension, the provision of the sentence and the picture seemed to

enable her to bypass her PAS impairment. The impact of PAS deficits may also beminimised by the type of argument structure tested; most of the structures elicited were

simple one-, two-, and three-argument structures. The current assessments did not assess

the clients' abilities to use the verbs in all of their PAS forms and using sentential

complements. The PAS difficulties seen in conversation by Thompson et al. (1997) may,

therefore, not have been identified in this study.

It is difficult to determine the impact of thematic role assignment difficulties on

sentence production in terms of the omission of arguments and a possible reliance on

simple argument structures. As both GW and KD had additional impairments that would

result in argument omission or a reliance on simple structures, we cannot be sure which

impairment is responsible for the surface symptoms.

One interesting pattern of performance observed within the study is the apparentdissociation between the ability to understand verbs (i.e., discriminate them from related

verbs) and the ability to identify semantic anomalies in grammaticality judgement.

Marshall (1995) suggested that semantic selection restrictions are encoded within the

verb's lexical semantic representation and thus would be accessed alongside the verb's

core meaning. GW presented with difficulties with verb comprehension and the identi-

fication of semantic anomalies. He also produced semantically anomalous arguments

within the sentence generation task. TJ and JM, however, present with selective diffi-

culties. TJ had a verb comprehension deficit that was more severe than that of GW, but

despite this performed within normal limits on the grammaticality judgement task. JM, on

the other hand, had no difficulties with verb comprehension but failed to reject semantic

anomalies. GW and JM both have difficulties accessing PAS information and it may bethat their difficulties in accessing semantic selection restriction information reflect this

impairment. This would suggest that PAS information not only specifies the number and

type of arguments needed alongside the verb but also information about the lexical items

that can fulfil those arguments. Further investigations would be necessary to confirm this

association and its distinction from the core meanings of verbs. The sentence generation

task and grammaticality judgement tasks within this study used the same verbs and it may

be that the co-occurrence of deficits reflects particular difficulty with those verbs. Further

investigations of comprehension are needed with a broad range of verbs. In addition,

tasks investigating the relationship between verbs and nouns that can fulfil various the-

matic roles are required.

The results of this study suggest that difficulties in producing the thematic structure ofsentences may be a consequence of different underlying impairments. The same surface



22/23

symptoms in connected speech (reliance on simple argument structures, production of

structures with an undetermined argument structure, and the omission of arguments) are

manifestations of different impairments. We have shown that, for example, the omission

of arguments can result from not being able to retrieve a word, not knowing that theargument is necessary, or not being able to assign a thematic role to the argument. If

treatment is to be targeted at the underlying impairment, therefore, it needs to be pre-

ceded by detailed assessment on constrained tasks. Only by an analysis of performance

on these tasks can the precise nature of the impairment be determined.

Manuscript received 7 July 2003

Manuscript accepted 14 November 2003

REFERENCES

Baayen, R. H., Piepenbrock, R., & Gulikers, L. (1995). The CELEX Lexical Database (Release 2) [CD-ROM].

Philadelphia, PA: Linguistic Data Consortium, University of Pennsylvania.

Berndt, R. S., Haendiges, A. N., Mitchum, C. C., & Sandson, J. (1997a). Verb retrieval in aphasia II:

Relationship to sentences processing. Brain and Language, 56, 107137.

Berndt, R. S., Mitchum, C. C., Haendiges, A. N., & Sandson, J. (1997b). Verb retrieval in aphasia I: Char-

acterizing single word impairments. Brain and Language, 56, 68106.

Black, M., & Chiat, S. (2003) Linguistics for clinicians. London: Arnold Publishers.

Black, M., Nickels, L., & Byng, S. (1991). Patterns of sentence processing deficit: Processing simple sentences

can be a complex matter. Journal of Neurolinguistics, 6, 79101.

Breedin, S. D., & Martin, R. C. (1996). Patterns of verb impairment in aphasia: An analysis of four cases.

Cognitive Neuropsychology, 13, 5191.

Byng, S. (1988). Sentence processing deficits: Theory and therapy. Cognitive Neuropsychology, 5, 629676.

Byng, S., & Black, M. (1989). Some aspects of sentence production in aphasia. Aphasiology, 3, 241263.Caramazza, A., & Hillis, A. E. (1989). The disruption of sentence production: Some dissociations. Brain and

Language, 36, 625650.

Caramazza, A., & Hillis, A. E. (1991). Lexical organization of nouns and verbs in the brain.Nature, 349, 788790.

Fink, R. B., Martin, N., Schwartz, M. F., Saffran, E. M., & Myers, J. L. (1992). Facilitation of verb retrieval

skills in aphasia: A comparison of two approaches. Clinical Aphasiology, 21, 263275.

Garrett, M. F. (1980). Levels of processing in sentence production. In B. Butterworth (Ed.), Language pro-

duction I. New York: Academic Press.

Gleason, J., Goodglass, H., Obler, L., Green, E., Hyde, M., & Weintraub, S. (1980). Narrative strategies of

aphasic and normal speaking subjects. Journal of Speech and Hearing Research, 23, 370383.

Jones, E. V. (1986). Building the foundations for sentence production in a non-fluent aphasic. British Journal of

Disorders of Communication, 21, 6382.

Levelt, W. J. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press.

Levelt, W. J. M. (1992). Accessing words in speech production: Stages, processes and representations. Cog-

nition, 42, 122.

Marshall, J. (1995). The mapping hypothesis and aphasia therapy. Aphasiology, 9, 517539.

Marshall, J., Chiat, S., & Pring, T. (1997). An impairment in processing verbs' thematic roles: A therapy study.

Aphasiology, 11, 855876.

Marshall, J., Pring, T., & Chiat, S. (1998). Verb retrieval and sentence production in aphasia. Brain and

Language, 63, 159183.

Saffran, E. M., Schwartz, M. F., & Marin, O. S. M. (1980). The word order problem in agrammatism II:

Production. Brain and Language, 10, 263280.

Schwartz, M. F. (1987). Patterns of speech production deficit within and across aphasia syndromes: Application

of a psycholinguistic model. In M. Coltheart, G. Sartori, & R. Job (Eds.), The cognitive neuropsychology of

language (pp. 163199). Hove, UK: Lawrence Erlbaum Associates Ltd.

Schwartz, M. F., Fink, F. B., & Saffran, E. M. (1995). The modular treatment of agrammatism. Neuro-

psychological Rehabilitation, 5, 93127.

Schwartz, M. F., Linebarger, M. C., Saffran, E. M., & Pate, D. S. (1987). Syntactic transparency and sentenceinterpretation in aphasia. Language and Cognitive Processes, 2, 85.



23/23

Schwartz, M. F., Saffran, E. M., & Marin, O. S. M. (1980). The word order problem in agrammatism: I.

Comprehension. Brain and Language, 10, 249262.

Shapiro, L. P., & Levine, B. A. (1990). Verb processing during sentence comprehension in aphasia. Brain and

Language, 38, 2147.

Swinburn, K., Porter, G. & Howard, D (in press). The Comprehensive Aphasia Test. Hove, UK: Psychology

Press.

Thompson, C. K., Lange, K. L., Schneider, S. L., & Shapiro, L. P. (1997). Agrammatic and non-brain damaged

subjects' verb and verb argument structure production. Aphasiology, 11, 473490.

Webster, J. (1999). A semantic and syntactic analysis of aphasic speech. Unpublished PhD thesis, University of

Newcastle, UK.

Webster, J., & Bird, H. (2000). VAN: The Verb and Noun Test. Northumberland, UK: STASS Publications.

Webster, J., Franklin, S., & Howard, D. (2001). An investigation of the interaction between thematic and phrasal

structure in nonfluent agrammatic subjects. Brain and Language, 78, 197211.

Webster, J., & Whitworth, A. (2001). AL: Accessing the predicate argument structure. In S. Byng, K. Swinburn,

& C. Pound (Eds.), The aphasia therapy file (Vol. 2). Hove, UK: Psychology Press.

Whitworth, A. (1996). Thematic Roles in Production (TRIP). London: Whurr Publishers Ltd.

Whitworth, A. B. (1994). Thematic role assignment in word retrieval deficits in aphasia. Unpublished PhD

Thesis, University of Newcastle, UK.


webster et al thematic structure

Documents