Development and validation of the AzBio sentence lists

Area of Science:

• Audiology
• Speech Perception
• Hearing Loss Rehabilitation

Background:

• Assessing speech perception in hearing-impaired individuals and cochlear implant (CI) users requires reliable and equivalent sentence lists.
• Existing lists may not offer sufficient equivalency for longitudinal or cross-condition performance evaluation.

Purpose of the Study:

• To create and validate a new corpus of sentence lists for evaluating speech perception in hearing-impaired listeners and CI users.
• To generate a large number of equivalent sentence lists for consistent performance assessment over time and across different conditions.

Main Methods:

• Developed 33 sentence lists (20 sentences each) from the AzBio corpus, with sentence intelligibility estimated via a five-channel CI simulation and normal-hearing listener scores.
• Validated list equivalency by presenting all lists to 15 CI users, analyzing performance variability using a binomial distribution model.

Main Results:

• The CI simulation predicted 33 lists with a mean intelligibility score of 85% correct.
• Validation with CI users showed no significant score differences for 29 of the 33 lists.
• A binomial distribution model accurately accounted for listener variability and established confidence intervals for score comparisons.

Conclusions:

• A five-channel CI simulation effectively created equivalent sentence lists for speech perception testing.
• 29 of 33 validated lists demonstrated statistical equivalency in CI users, though individual variability exists.
• The binomial model aids in interpreting performance changes, with 15 lists incorporated into the AzBio Sentence Test and eight into the Minimum Speech Test Battery.