Modeling spoken word recognition performance by pediatric cochlear implant users using feature identification

Area of Science:

• Linguistics
• Auditory Neuroscience
• Speech-Language Pathology

Background:

• Understanding spoken word recognition in pediatric cochlear implant (CI) users is crucial for developing effective interventions.
• Existing psycholinguistic models of word recognition need evaluation for their applicability to this population.
• Assessing the relationship between different speech perception measures is important for comprehensive evaluation.

Purpose of the Study:

• To computationally evaluate psycholinguistic theories of spoken word recognition in children using cochlear implants.
• To investigate the interrelations between closed-set and open-set speech perception measures.
• To compare the performance of cochlear implant coding strategies (MPEAK vs. SPEAK) within these models.

Main Methods:

• Developed a software simulation of phoneme recognition using feature identification scores.
• Created two lexical access simulations: one with early phoneme decisions, one with delayed decisions.
• Applied simulated performance to behavioral data from the Phonetically Balanced Kindergarten and Lexical Neighborhood Tests.

Main Results:

• Open-set word recognition was accurately predicted by feature identification scores.
• No significant performance differences were found between MPEAK and SPEAK cochlear implant users.
• The model with delayed phoneme decisions during lexical access best predicted word recognition ability.

Conclusions:

• Spoken word recognition in pediatric CI users appears to follow the same fundamental cognitive processes as in normal-hearing individuals.
• Both closed-set feature identification and open-set word recognition provide valuable, related insights into language processing.
• Collecting both types of data can enhance clinical intervention strategies for children with cochlear implants.