1 of 1

Unreliable auditory feedback leads to decreased sensitivity to auditory errors

Taijing Chen, Ben Parrell (Mentor), Communication Science and Disorders

Background

Result

Figure 1: results from the 5 responders in both the normal and noisy conditions

​

Exposure (450 trails)

Test (240 trails)

Normal

Veridical condition (Figure 2)

  • Auditory perturbation: no
  • Reliable
  • Auditory perturbation: large
  • Past studies show such perturbation will lead to online compensatory speech production

​

Noisy

Noisy condition (Figure 3)

  • Auditory perturbation: small, unpredictable
  • Less reliable

Experiment Design

Figure 2: illustration of normal condition

Figure 3: illustration of noisy condition

Noisy exposure

Normal exposure

F1 compensation to downshift perturbation in normal condition

F1 compensation to upshift perturbation in normal condition

F2 compensation to downshift perturbation in normal condition

F2 compensation to upshift perturbation in normal condition

Discussion

Acknowledgement

  • 21 participants in total; 5 participants showed a significant compensation to the perturbation in the test phase (responders) (Figure 4)
  • The average amount of compensation of the 5 responders is reduced under the noisy condition compared with the one under the normal condition (Figure 1)

Support Hypothesis 2:

  • Exposure phase: perturbing participants’ auditory feedback with a small amount 🡪 auditory feedback reliability
  • feedback reliability 🡪 perturbation compensation
  • Consistent with Bayesian Integration model

Figure 4: F1 compensation by subject

Deep thanks to Dr. Parrell for mentoring me through the project, and to Elizabeth Markland and Kathleen Zarnott who are working on this project with me. Great thanks to the Undergraduate Research Scholar Program to give me the opportunity to work on this project.

Responders

Bayesian Integration Model: prediction + sensory feedback

Prediction

Final Estimate

Sensory Feedback

Hypothesis 1: rely more on own sensory feedback

Test

Test

Hypothesis 2: rely more on the prediction

Compensation (F1)