Is speech categorization consistency achieved in part through lexical knowledge?

Speech categorization is a gateway for downstream language processes. Recent evidence from work with the Visual Analog Scaling (VAS) task (Kapnoula et al., 2017) underscores the importance of categorization consistency (trial-by-trial variability around the mean function) over slope (long-term category structure) as a critical predictor of real-world outcomes such as L1 language ability and L2 language learning (Kim et al., under review; Honda et al., 2024). However, the mechanisms that contribute to categorization consistency are unclear. Given that higher-level factors like the lexicon can stabilize the percept by cleaning up lower-level perceptual noise (Luthra et al., 2021), we examined the relationship between categorization consistency and the lexical status of the percept. We tested adult American English listeners (n = 48, data collection ongoing) with the two sets of VAS tasks: one with word continua (e.g., beach-peach) and the other with nonword continua (e.g., beag-peag). Preliminary results showed that listeners’ categorization consistency for the word continuum is significantly higher than for the nonword continuum. However, we did not find a significant difference in slope. This suggests that categorization consistency (but not slope) can be stabilized by listeners’ lexical knowledge, supporting “auditory/phonological clean-up” driven by lexical chunks.