Temporal recalibration in vocalization induced by adaptation of delayed auditory feedback

Abstract

Background: We ordinarily perceive our voice sound as occurring simultaneously with vocal production, but the sense of simultaneity in vocalization can be easily interrupted by delayed auditory feedback (DAF). DAF causes normal people to have difficulty speaking fluently but helps people with stuttering to improve speech fluency. However, the underlying temporal mechanism for integrating the motor production of voice and the auditory perception of vocal sound remains unclear. In this study, we investigated the temporal tuning mechanism integrating vocal sensory and voice sounds under DAF with an adaptation technique.

Methods and findings: Participants produced a single voice sound repeatedly with specific delay times of DAF (0, 66, 133 ms) during three minutes to induce 'Lag Adaptation'. They then judged the simultaneity between motor sensation and vocal sound given feedback. We found that lag adaptation induced a shift in simultaneity responses toward the adapted auditory delays. This indicates that the temporal tuning mechanism in vocalization can be temporally recalibrated after prolonged exposure to delayed vocal sounds. Furthermore, we found that the temporal recalibration in vocalization can be affected by averaging delay times in the adaptation phase.

Conclusions: These findings suggest vocalization is finely tuned by the temporal recalibration mechanism, which acutely monitors the integration of temporal delays between motor sensation and vocal sound.

Figure 1. Proportion of ‘delayed’ responses on average and for each participant in Experiment 1.

Data were fit with a probit function to capture the proportion of ‘delayed’ responses to each SOA (ms) in the simultaneous judgment task. The crossover point of each line and the horizontal line at 50% proportion of ‘delayed’ judgment was taken as the point of subjective similarity (PSS). The PSS calculated for each participant were 182.17, 79.51, 142.26, 59.40, 37.17, and 101.59 ms; the averaged value was 98.41 ms. Red line represents the averaged responses across six participants. P2 data represents the author's response.

Figure 2. Experimental procedure.

(A) Time sequence in a single SOA session: participants were exposed to a uniform SOA (0 ms, 66 ms, or 133 ms) continuously during the adaptation and top-up phases, followed by a test phase where they judged each of seven SOAs in random order. (B) Visual cues presented for producing voice. Numbers on colored circles (red, blue, green and yellow) were presented for 1 s durations as an instructional countdown. Three blue circles meant rest for three seconds, ten green or yellow circles meant vocalization in the adaptation or top-up phases, respectively. Two red circles also meant vocalization but required participants to report their simultaneous judgment.

Figure 3. Change of simultaneous judgment in Experiment 2.

(A) Across participants, average PSS increased as a function of adapted SOA. (B–G) The trends for each participant.

Figure 4. Comparison of mean PSS and difficulty for Experiment 2.

The average PSSs in 0 ms (control), 66 ms, and 133 ms conditions are 67.60 ms, 92.23 ms, and 112.10 ms; the average indices of difficulty are 21.04, 26.82, and 35.60, respectively. Significant differences of PSS were found between control and 66 ms condition (p<.05), control and 133 ms condition (p<.05), and 66 ms and 133 ms condition (p<.01). There were no significant differences of the indices of difficulty.

Figure 5. Temporal recalibration occurs with averaged SOAs.

Probit analysis was used to fit the proportion of stimuli that participants judged as delayed for each SOA (ms) in Experiment 3: 100 ms, 66 ms+133 ms, and 33 ms+166 ms conditions. There was no change in delay judgments as a function of SOA condition.

Figure 6. Sequential analysis of averaged performance in Experiment 1.

Each line represents the result of Probit analysis for each of four sequential blocks that include five sessions. Comparison among the four blocks showed no significant difference. Session 1–5 means the averaged data among first to fifth session, and so on.