Sensorimotor adaptation to feedback perturbations of vowel acoustics and its relation to perception
Schematic diagram of the cycle that occurred during the presentation of one token during an SA experimental run. Refer to Sec. II B for a detailed description.
Diagram of the level of F1 perturbation presented during one experimental session, as a function of epoch number (where an epoch consists of one repetition of each of the 18 words in the corpus). The 65 epochs of an experimental session are divided into four phases (demarcated by dashed vertical lines). From left to right, these phases are base line (epochs 1–15), ramp (epochs 16–20), full perturbation (epochs 21–45), and postperturbation (epochs 46–65). The protocols for two subject groups are shown: those undergoing an upward F1 shift (upper line) and those undergoing a downward F1 shift (lower line).
Produced first and second formant frequencies, normalized to the adjusted base line, as a function of epoch number in feedback words for all subjects. The upper curve corresponds to the normalized F1 for the ten subjects run on the shift-down protocol; the lower curve corresponds to the shift-up protocol. Each data point is the mean value of the nine feedback words across ten subjects (five male, five female). The dashed vertical lines demarcate the phases of the protocol; the dashed horizontal line corresponds to base line values. Normalized F2 values are shown as the dashed curves, which remain close to the base line value of 1.0. The error bars depict the standard error of the mean among ten subjects.
Produced first formant frequency, normalized to the base line, in the feedback words containing the vowel . The top plot shows normalized F1 for the same context, feedback token (“pet”), while the bottom figure shows normalized F1 for the different context, feedback tokens (“get” and “peg”). The axes, data labels and vertical markers are the same as in Fig. 3, except that normalized F2 is not shown.
Full-pert phase formants normalized to base line for all feedback token vowels. Mean first formant values are shown in the upper plot; second formant values, in the lower plot. Values from shift-down subjects are represented by dark bars and from shift-up subjects, by light bars. Error bars show standard error about the mean.
Normalized F0 as a function of epoch number. To maintain consistency with Fig. 3, only feedback vowels are shown. The solid line represents the mean values from the shift-down subjects; the dashed line represents the mean values from the shift-up subjects. The vertical lines demarcate the phases of the experiment.
Example of the adaptive procedure used to estimate jnd. The abscissa shows the presentation number of the given pair, and the ordinate depicts the separation of the different pairs in pert. The text within the figure gives conditions for changes in step size. The staircase terminated after eight reversals.
The adaptive response index is correlated with the jnd score of the base token. The ordinate shows the jnd score (Discrimination Index), while the abscissa shows the adaptive response index. The open circles represent shift-down subjects, while the triangles represent shift-up subjects. Statistics for the regression line are shown in the legend.
A functional diagram of the DIVA model of speech motor control. The feedforward component projects from the speech sound map , and is scaled by weight . The feedback component consists of projections from the auditory and somatosensory error maps, and are scaled by weights and , respectively. The feedforward and feedback projections are integrated by the speech motor cortex to yield the appropriate speech motor commands, which drive the vocal-tract articulators to generate speech sounds.
Normalized F1 as a function of epoch number during the SA protocol in feedback trials: DIVA simulations compared to human subject results. The thin lines shown with standard error bars correspond to the subject SA data (20 subjects). The shaded region corresponds to the DIVA simulations, and represents the 95% confidence interval about the mean. The vertical dashed lines show the experiment phase transitions; the horizontal dashed line indicates base line. The open circles indicate epochs in which the data and the simulation results were significantly different. The black solid curves correspond to high-acuity simulations, while the black dashed curves correspond to low-acuity simulations.
Normalized F1 during the SA protocol in feedback trials: DIVA simulations compared to subject results.
Block diagram of the formant-shifting algorithm used to introduce acoustic perturbations in the SA experiment.
Word list for the SA experiment.
Article metrics loading...
Full text loading...