Speech intelligibility depends on listener experience: Comparing speech-language pathologists, naïve listeners, and automated recognition

This study examined whether intelligibility judgements differ among experienced speech-language pathologists (SLPs), naïve listeners, and an automated speech recognition system when presented with identical speech stimuli. A total of 87 participants (48 SLPs, 38 naïve listeners, and one automated system) completed a closed-set word identification task and provided continuous intelligibility ratings for a narrative passage. Speech stimuli consisted of synthesised productions incorporating developmental phonological error patterns. Accuracy scores increased from the automated system (M = 56.00%) to naïve listeners (M = 56.32%, SD = 8.6) to SLPs (M = 61.08%, SD = 7.9). Group differences were statistically significant, χ2 (2) = 6.17, p = 0.046, with post hoc analyses indicating differences between SLPs and naïve listeners. Intelligibility ratings also differed across groups, χ2 (2) = 10.11, p = 0.006, with higher ratings observed for SLPs compared to naïve listeners. Findings suggest that listener experience influences intelligibility judgements, with experienced SLPs demonstrating greater accuracy and higher intelligibility ratings than naïve listeners. Performance of naïve listeners was comparable to that of the automated system, with no statistically significant differences observed between these groups. These results highlight the role of listener experience in interpreting disordered speech and have implications for both clinical assessment and the use of automated recognition systems.