Fig 1.
The effect of FPC2 and FPC3 on the mean f0 contour of the connective (left: FPC2; right: FPC3). The solid line in each panel represents the mean f0 contour (normalized f0 in semitones over time in ms); the “+” line illustrates the outcome curve after adding one standard deviation of the FPC to the mean f0 curve; the “-” line shows the outcome curve after subtracting one standard deviation of the FPC from the mean f0 curve. The x-axis denotes time in ms, and the y-axis represents normalized f0 in semitones.
Fig 2.
The f0 contours of the vowel /əʊ/ in the connective so for the two stimuli in pair 5.
The lines are color-coded as black for the objective causality prosody condition (“_O” in the legend) and grey for the subjective causality prosody condition (“_S” in the legend). The dotted lines show the original f0 contours (“original”); the dashed lines represent the resulting f0 contours after FPCA (“reconstructed”); the solid lines display the resulting f0 contours after manipulation, where the values of FPC2 and FPC3 in Eq 1 were changed to the mean values estimated by the linear regression (“manipulated”).
Table 1.
The acoustic details of the connective so in each prosodic condition.
Fig 3.
A graphic description of the timeline of a trial.
Table 2.
Model formulas.
Fig 4.
Prior predictive distributions generated by m0, the intercept model, under different priors.
Fig 5.
The posterior distributions of model parameters of the most complex Bayesian mixed-effects model (m3) with the least informative prior for the constant effect of prosody.
The model contains the constant effect of prosody, which consists of two levels: subjective and objective (coded as 1 and 0, respectively). Additionally, the model incorporates a varying slope of prosody by participant and pair. The plot shows the estimated mean (as represented by circles) as well as the lower and upper bounds of the 95% (depicted by thinner lines) and 50% (thicker lines) credibility intervals for each model parameter of interest. These parameters include the random effect terms sd(prosodyS) by participant and pair, the intercept, and the constant effect of prosody.
Table 3.
The Bayes factors for each model term, computed using three different priors.
Fig 6.
The posterior distribution of the slope for prosody for each participant, averaged over the levels of pair, as visualized using the posterior mean (the black dots) and the percentile-based 95% and 50% posterior intervals (the thin and thick horizontal lines, respectively).
Each line in the figure represents a participant with the participant’s ID denoted on the y-axis.