PP508 - Behavioral and Neural Correlates of Phonetic Categorization in Adults with Cochlear Implants

Abstract:
Postlingually deafened cochlear implant (CI) adults have large variability in speech perception abilities. While behavioral measures remain the gold standard for determining speech perception abilities, an objective measure is needed for patients who are unable to provide reliable behavioral responses. Auditory event-related potentials (ERP) were obtained with a double oddball paradigm to produce the mismatch negativity (MMN) response, which has been shown to have associations with phonetic categorical perception at the group level. The results indicated that the MMN is associated with categorical perception at the individual level. These findings are clinically relevant to determining appropriate follow-up care post-implantation.

Summary:
The overall aim of this research was to determine whether the double-oddball mismatch negativity (MMN) paradigm (that produces a neural response amplitude ratio) predicted behavioral phonetic categorization in addition to speech perception performance in adult CI patients. Using electrophysiological and behavioral measures, this study determined the relationship between neural measures of phonetic categorization to determine if these factors are predictive of speech perception in quiet and noise for adult CI users. This study used a double oddball paradigm for electrophysiological measures, which differs from the classic single oddball paradigm in that the MMN is elicited using two different deviant stimuli in the same block. By using the double oddball paradigm, a listener’s MMN response evoked by a within phoneme category deviant can be compared to the MMN evoked by an across phoneme category deviant, producing an MMN ratio response. This study employed a cross-sectional design to evaluate the brain function and speech-perception abilities of 10 individuals with hearing loss as well as 6 normal hearing listeners. A combination of behavioral measures of auditory, speech and language and physiologic measures of brain activity were examined, and these measures were identical or similar to measures that are routinely administered in clinical settings. Stimuli utilized for the identification and discrimination tasks were created with Adobe Audition, and recorded materials were utilized for speech testing in quiet and in noise.

The data collected from behavioral and objective measures were analyzed with appropriate statistical tests including linear mixed effect regression models and analysis of variance (ANOVA). Linear mixed effects regression was used to assess relationships between behavioral and electrophysiological measures. Individual behavioral identification data was fit using a probit function and the slope, boundary, and boundary width of the function was analyzed. Individual behavioral discrimination data was analyzed for overall percent correct, taking into account all possible responses categories and a bias-free estimate of perceptual sensitivity (d’) was computed. Word and sentence recognition scores were scored for percent correct and were used in the analyses.

For clinical purposes, an objective test needs to be sensitive to individual differences. This study explores a novel double-oddball paradigm that produces a clinical MMN amplitude criterion that is predictive of behavioral outcomes at the individual level. CI recipients with strong behavioral phonetic categorization skill will demonstrate larger neural responses to MMNs elicited with the across category vs. within category deviant (e.g. AC:WC amplitude ratio > 1.0). Conversely, CI recipients with small MMN ratio responses are expected to have poorer perception of phonetic contrasts compared to those with larger MMN ratios. Examining the relationship between behavioral measures of speech perception (in quiet and noise) and neural measures of phonetic categorization will determine if MMNs are a valuable clinical tool when assessing speech perception in individuals with CIs. The results of this study suggest the AC:WC MMN amplitude ratio is a significant predictor of phonetic categorization skill and that having a typical phonetic category boundary was predictive of better speech recognition scores.