Word Retrieval Measures with the AphasiaBank Stimuli: Test-Retest Reliability Proposal for CAC 2014

Boyle, Mary (2014) Word Retrieval Measures with the AphasiaBank Stimuli: Test-Retest Reliability Proposal for CAC 2014. [Clinical Aphasiology Paper]

PDF 504-926-1-RV(Boyle).pdf Download (102kB)

Abstract

Research into treatment for improving word retrieval ability in aphasia is increasingly focused on assessing outcomes at a discourse level. For example, the AphasiaBank project (http://talkbank.org/AphasiaBank/ ) uses a number of tasks to elicit discourses from individuals with aphasia. The discourses can then be analyzed with a set of analysis tools from the Computerized Language Analysis (CLAN) system. MacWhinney, Fromm, Holland, Forbes, & Wright (2010) have suggested that the AphasiaBank tools can be used to study recovery from aphasia and the effects of aphasia treatments. The AphasiaBank protocol is promising because of its ability to quickly and accurately perform a number of analyses that are time-consuming, cumbersome, and vulnerable to error when performed manually. However, except for a report on VOCD, a measure of lexical diversity that is part of the CLAN system (Boyle, 2013), there have been no reports about the test-retest reliability of the various language measures included in CLAN when they are used with the elicitation stimuli that are part of the AphasiaBank protocol. Test-retest reliability refers to the assessment of whether a test produces the same results on repeated application when the participants who are being tested have not changed on the domain that is being measured (Fitzpatrick, Davey, Buxton, & Jones, 1998). Before a measure is used as an outcome assessment, its test-retest reliability must be established, otherwise it is impossible to assert that changes on the measure are related to treatment rather than to spurious, day-to-day variability inherent in the measurement or the behavior it is measuring (Brookshire & Nicholas, 1994; Herbert, Hickin, Howard, Osborne, & Best, 2008). Test-retest reliability is as important for measures used to evaluate impairments as it is for those that measure change, since measures that are not stable will not provide valid or reliable assessments of impairments. Several measures available in the CLAN System can be used to assess word retrieval difficulty. To use CLAN, the discourses must first be transcribed and coded for errors and other behaviors of interest using a format specified in the CHAT Manual (http://talkbank.org/AphasiaBank/). CLAN can then be used to analyze the transcripts for the occurrence of the coded errors as well as for other language parameters. Word-finding problems that can be coded in CHAT include phonemic paraphasias, semantic paraphasias, neologisms, false starts, time fillers, and repetitions. The purpose of this investigation was to provide preliminary information about the test-retest reliability of these measurements in narrative discourses elicited with the AphasiaBank stimuli from speakers with aphasia.

Actions (login required)

View Item