François Bergeron1, Aurore Berland2, Elizabeth M Fitzpatrick3, Christophe Vincent4, Annie Giasson5, Kevin Leung Kam6, Walid Chafiq6, Thibaut Fanouillère6, Dominique Demers1. 1. a Department of Rehabilitation, Medical Faculty , Université Laval , Québec , QC , Canada. 2. b Department of Speech and Language Pathology , Université de Toulouse II , Toulouse , France. 3. c School of Rehabilitation Sciences , Université d'Ottawa , Ottawa , ON , Canada. 4. d Department of Otology & Neurotology , CHRU de Lille , Lille , France. 5. e Department of Audiology , Hôpital Régional Chaleur Regional Hospital , Bathurst , NB , Canada. 6. f Faculte des Sciences de Montpellier , Université de Montpellier , Montpellier , France.
Abstract
Objectives: The goal of this work was to develop and normalise an international French version of the AzBio sentence test. Design: A corpus of 1000 sentences was generated. These sentences were recorded with four talkers and processed through a four-channel cochlear implant simulation. The mean intelligibility for each sentence achieved by 16 normal-hearing listeners was computed. The consecutively ordered 165 sentences from each talker rendering an average score of 85% were sequentially assigned to 33 lists of 20 sentences. All lists were presented to 30 normal-hearing and 25 hearing-impaired listeners in order to verify their equivalency. Thirty normal-hearing adults were also recruited to assess the test's psychometrics and define norms. Results: The results of the list equivalency validation study showed no significant differences in percent correct scores for 30 sentence lists. A binomial distribution model was used to estimate the 95% critical differences for each potential percentage score. Normalization data showed an average performance between 96% and 99% with a very low standard deviation. Conclusions: With a set of 30 lists, researchers and clinicians can use the FrBio to evaluate a large number of experimental conditions; changes in performance over time or across conditions can then be tracked.
Objectives: The goal of this work was to develop and normalise an international French version of the AzBio sentence test. Design: A corpus of 1000 sentences was generated. These sentences were recorded with four talkers and processed through a four-channel cochlear implant simulation. The mean intelligibility for each sentence achieved by 16 normal-hearing listeners was computed. The consecutively ordered 165 sentences from each talker rendering an average score of 85% were sequentially assigned to 33 lists of 20 sentences. All lists were presented to 30 normal-hearing and 25 hearing-impaired listeners in order to verify their equivalency. Thirty normal-hearing adults were also recruited to assess the test's psychometrics and define norms. Results: The results of the list equivalency validation study showed no significant differences in percent correct scores for 30 sentence lists. A binomial distribution model was used to estimate the 95% critical differences for each potential percentage score. Normalization data showed an average performance between 96% and 99% with a very low standard deviation. Conclusions: With a set of 30 lists, researchers and clinicians can use the FrBio to evaluate a large number of experimental conditions; changes in performance over time or across conditions can then be tracked.