Embedded Files
Phase 1: Tone-Based
White noise, 40-Hz modulation frequency
A faint band of increased power appears at 40 Hz. There is a strong band of electrical noise at 60 Hz.
White noise, sinusoidally changing modulation frequency
The modulation frequency changed from 40 Hz to 80 Hz sinusoidally every 10 seconds.
No significant trace of increased power could be observed.
White noise, alternating modulation frequency
The modulation frequency switched between 40 Hz and 80 Hz, staying on each for 10 seconds at a time.
Faint bands of increased power can initially be observed (80 Hz in the first 8 seconds and 40 Hz between 8 and 18 seconds). However, these signals fade off over time, suggesting that adaptation occurs in the brain. Our assumptions of linearity and time-invariance in the brain are thus violated.
Phase 2: Speech-Based
Event-related potentials in T7 electrode
The blue line on the bottom displays the ERP for A) "Yes" and B) "No" stimuli. Above, each row represents one trial, and the color indicates the voltage. There is a clear difference in the response at 200 ms ("Yes" has a sharper, single-peak negative wave, whereas "No" has a wider, double-peak negative wave).
Visualization with PCA
However, visualization with PCA reveals little evidence of clustering for individual trials. There is much overlap between "Yes" and "No" trials for both A) one subject and B) three subjects.
Machine learning classification
As what might be expected from the PCA plot, the machine learning methods were unable to classify between "Yes" and "No" stimuli better than chance. However, there is some evidence that we simply did not have enough data—our three-subject tests performed better than the single-subject tests despite the presence of interpersonal variance.
Credit: Anthony Vasquez
Page updated
Report abuse