Effect of speech recognition test presentation on cochlear implant user performance.

Everyone was tested with their own device and in their own listening setup (both ears, one ear, or implant plus hearing aid), so the group mixes very different listeners and the averages hide that. Participants were recruited by convenience at two centres, with no comparison group and no randomisation; each person was compared against themselves across the four conditions. Live-voice testing cannot be blinded, and one audiologist spoke for every participant, so the live-voice results reflect that person's voice and accent and may not transfer to another tester. Children from age five and adults were tested together, and the results are not broken down by age. Recorded speech in noise produced low scores (a mean of 30.8% for sentences) and live voice in quiet produced high ones (a mean of 69.2% for words), so the top and bottom of the scale may be constrained. The authors themselves caution that similar average scores across conditions may still reflect different listening demands. This was a single assessment, so it says nothing about whether either presentation mode predicts how someone hears in daily life or how much they benefit from their implant over time.

The paper states that there was no financial support ('nothing to declare') and gives no other conflicts of interest. It does not say who designed the study or wrote it up.

Do not read this as a reason to switch to live-voice testing. The higher live-voice scores mean a clinic that tests by voice may overestimate how well a patient hears, and the authors recommend recorded materials precisely because a tester's voice, accent and health change from session to session.