Spectral moments (SMs, Forrest et al., 1988) are widely used to characterize fricatives. Shadle (2023) described the problems of using SMs and proposed a set of alternative parameters (APs) based on articulatory and aerodynamic models. The current study examines how well the APs from Shadle (2023) and SMs predict listener ratings of the goodness of children’s word-initial /s/ and /ʃ/ productions. Seventy-six adult listeners rated a total of 1014 /s/- and /ʃ/-initial words produced by 64 children of 2.5–3.5 years of age along a visual analog scale. For each fricative, the first four SMs were calculated for the middle portion of the fricative, as were the filter parameters in Shadle (2023): frequency and amplitude of the main peak in the 3–8 kHz range, and the frequency and amplitude of the minimum in the 0.55–3 kHz range. While correlation and mixed-effect regression analyses show that both sets of measures robustly predict adult ratings, the model with APs yielded a slightly lower AIC, indicating a better model fit. Moreover, the APs better predict the narrow transcriptions of children’s /s/ and /ʃ/ productions. Together, these findings provide external validation for Shadle’s APs.
Wong et al. (Tue,) studied this question.