Consistency in computerised and standard testingin aphasia and healthy controls

Susan Edwards *

School of Psychology & Clinical Language Sciences, University of Reading, Reading RG6 6AL, UK

Introduction

Aspects of aphasic comprehension in one participant were investigatedusing traditional and computerised versions of the same tests. The aphasicscores from both modes of testing were compared. Results were then com-pared with those from a group of healthy controls. The computerised ver-sions provided reaction times (RT) and scores allowing results to beexamined for consistency and an assumed hierarchy of processing diffi-culty associated with sentence types and anaphoric expressions aboutwhich, there are conflicting claims. In a recent treatment study (Edwards& Tucker, 2006), clinical testing of aphasic comprehension and speed ofresponse revealed non-significant variation in performance prior to treat-ment but significant improvement after treatment. Such results need to beinterpreted in the light of healthy control data.

Subjects

The aphasic subject (CB) was a 53 year old male, 3 years post-onset ofa left middle cerebral thrombosis. He was judged to have a moderate tosevere Broca’s aphasia, with occasional comprehension difficulties, usingclinical observation and assessment (Goodglass, Kaplan, & Barresi,2001). Fifteen control subjects were recruited from a panel of healthyelderly volunteers: age range 70–86 years (mean = 78) and years of educa-tion range between 9 and 19 years (mean = 13.3).

Procedures

The sentence comprehension task from the Verb and Sentence Test:VAST (Bastiaanse, S., & Rispens, 2002) and a test of pronoun/reflexivecomprehension (Edwards & Varlokosta, in press) were given. CB wastested with the standard and a computerised version of the tests. As thereare existing control data for the standard version of both tests, the controlparticipants were tested with the computerised versions only. Thisemployed sound files, a touch screen (for the VAST test) and aresponse-button panel for the pronoun/reflexive test.

Sentence comprehension was tested using a picture selection task. CBheard the target, one of 4 sentence types, while four pictures were dis-played in booklet form (standard procedure) or on the computer screen.

The pictures comprised the target and 3 distractors. The task was to touchthe picture on the page (or on the screen) that matched the sentence(N = 40).

A truth verification task was used to explore comprehension of pro-nouns (N = 12), reflexives (N = 12) and the use of these anaphors in sen-tences with quantified antecedents (N = 12 for each condition). A fifthsentence type, Exceptional Case Marking, was also included (N = 8) butnot discussed in this report. Each picture (N = 56) was displayed and asentence heard that either matched or did not match the picture. The par-ticipant was required to say yes or no (in the standard procedure) or presseither a yes or a no key in the computerised version. Prior to testing it wasestablished that CB had reliable yes/no responses. Sentences were ran-domly ordered in both tests and for both presentations.

Results

Sentence comprehension

Neither scores nor reaction times for CB (Table 1) show an agram-matic pattern and, unexpectedly, active and subject clefts have a longerRTs than object clefts and passives. Scores are similar in both conditionsalthough depressed for passives in the computerised version.

Unlike CB, the 15 healthy control subjects made no errors and yieldeda predicted pattern in terms of reaction times. Actives had the shortestRTs, followed by subject clefts, passives and finally object clefts. Somecontrols were slower on both cleft sentence types. Speed across controlsvaried.

Pronoun comprehension

Scores on both modes of the truth verification task were similar for CBapart for the noun + quantified antecedent (QA) (Table 1). In the mis-match, correct condition he had fastest RTs for sentences containingreflexives, followed reflexives with QAs, then pronouns, then pronounswith QA. Rankings in the correct match condition and the incorrect mis-match condition deviate from this ranking. Some controls made errors inthe mismatch pronoun condition and in the pronoun and reflexive QAconditions. Speed of their response in the mismatch and the match condi-tion varied with neither condition consistently slower. Rank order acrossthe conditions was similar but with considerable individual variation inoverall speed.

doi:10.1016/j.bandl.2007.07.111

* Fax: +44 118 378 8142.

E-mail address: llsedwds@rdg.ac.uk

www.elsevier.com/locate/b&l

Brain and Language 103 (2007) 8–249

Discussion

CB’s relatively poor comprehension of passive sentences is exacerbatedin the computerised test yet these sentences do not have the longestresponses. It could be that accuracy is sacrificed for speed. RTs for thefour sentence types do not relate to the distribution of complexity assumedfor Broca’s aphasia. Slightly fewer errors are made in the computerisedversion of the truth verification task and the impact of quantification var-ies across the two conditions. There was no clear pattern in RTs and noconsistent advantage in either the mis-match or match condition. Whereasthe presence of a QA depressed reaction times where responses were incor-rect, the same was only true for QA + reflexive condition in the correctresponses. Variation across mode of testing for the aphasic patient andconsiderable variation in RTs in individual controls underlines the care

needed when using a single clinical test for diagnosis or when comparingpathological performance to ‘normal’ behaviour.

References

Bastiaanse, R. Edwards, S. & Rispens, J. (2002). The Verb and Sentence

Test. Harlow: Thames Valley Publishers.Edwards, S. & Varlokosta, S. (in press). Pronominal reference and

agrammatic comprehension. Journal of Neurolinguistics.

Edwards, S., & Tucker, K. (2006). Verb retrieval in fluent aphasia: Aclinical study. Aphasiology, 20, 644–675.

Goodglass, H., Kaplan, E., & Barresi, B. (2001). The assessment of aphasia

and related disorders (3rd ed.). Baltimore: Lippincott, Williams &Wilkins.

Table 1Scores and reaction times for aphasic participant

Active Subject cleft Object cleft Passive

1. Sentence comprehension # correct N = 40Standard 8/10** 9/10** 8/10* 7/10Computer 8/10 8/10 7/10 4/10RT ms 5062 5625 4083 4570

Pronoun Reflexive Pronoun + qu. antecedent Reflexive + qu. antecedent

2. Comprehension mismatch conditionStandard 4/6 5/6 2/6 3/6Computer 5/6 5/6 5/6 2/6

3. RT correct responsesRT on mismatch/match 3428/3177 2291/4099 3866/2992 2539/4509

4. RTms incorrect responsesMismatch/match 2563/0 3578/0 5087/0 6156/2609

* Includes o1 correct response after repetition.** Includes o2 correct responses after repetition.

194 Abstract / Brain and Language 103 (2007) 8–249