Comparing connected language elicitation procedures in persons with aphasia: Concurrent validation of the Story Retell Procedure

Background: The Story Retell Procedure (SRP) (Doyle et al., 1998) is a well-described method for eliciting connected language samples in persons with aphasia (PWA). However, the stimuli and task demands of the SRP are fundamentally different from commonly employed picture description, narrative, and procedural description tasks reported in the aphasia literature. As such, the extent to which measures of linguistic performance derived from the SRP may be associated with those obtained from picture description, narrative, and procedural description tasks is unknown. Aims: To assess the concurrent validity of linguistic performance measures obtained from the SRP with those obtained from picture description, narrative, and procedural description tasks by examining the correlations and the magnitude differences across the linguistic variables among the elicitation tasks. Secondarily, we examined the relationship of the percentage of information units per minute (%IU/Min) to other linguistic variables within the SRP and across the other elicitation tasks. Methods and Procedures: This study compared the SRP to six different, frequently used sampling procedures (three sets of picture descriptions, one fairytale generation, one set of narratives, and one set of procedural description tasks) from which the same five verbal productivity, four information content, two grammatical, and two verbal disruption measures were computed. Language samples were elicited from 20 PWA, spanning the aphasia comprehension severity range. Tests of association and difference were calculated for each measure between the SRP and the other sampling methods. Outcomes & Results: Significant and strong associations were obtained between the SRP and the other elicitation tasks for most linguistic measures. The SRP produced either no significant or significantly greater instances of the dependent variable except for the type-token ratio, which yielded a significantly lower value than the other sampling procedures. Conclusions: The findings are interpreted as support for the concurrent validity of the SRP and as evidence that a single form of the SRP will yield a language sample that is generally equivalent in distribution to other sampling procedures, and one that is generally greater in quantity to those typically used to assess connected spoken language in PWA. Additionally, it was found that the %IU/Min metric predicted highly the information content linguistic measures on the SRP as well as on the other elicitation procedures. However, it did not predict well measures of verbal productivity, grammaticality, or verbal disruptions.