Fig 1.
Updates of information gain for different initial prediction errors (initial uncertainty = 1.0, noise = 0.5, learning rate α = 0.1).
Fig 2.
Updates of information gain for different initial uncertainties (initial prediction error = 0, noise = 0.5, learning rate α = 0.1).
Fig 3.
Updates of information gain for different initial uncertainties (initial prediction error = 10.0, noise = 0.5, learning rate α = 0.1).
Fig 4.
Relationship between the initial uncertainty and the integrated value of the information gain.
Table 1.
Combination of percussion instruments and percussive sounds.
Fig 5.
Subjectively reported scores for surprise intensities in response to percussive sounds congruent with the instrument shown (i.e., low initial prediction error).
The results for familiar and unfamiliar instruments were compared at every 40 exposures (N = 16).
Fig 6.
Subjectively reported scores for surprise intensities in response to percussive sounds incongruent with the instrument shown (i.e., high initial prediction error).
The results for familiar and unfamiliar instruments were compared at every 40 exposures (N = 16).
Fig 7.
Grand mean ERP waveforms for the four combinations of percussion instruments and percussive sounds at the parietal midline region (Pz).
Open triangles: The onset of film presentation. Solid triangle: The onset of beating sound presentation. The horizontal bars show the time range of 250–600 ms for the P300 latency.
Fig 8.
P300 amplitudes evoked by percussive sounds congruent with the instrument shown (i.e., low initial prediction error).
The results for familiar and unfamiliar instruments were compared at every 40 exposures (N = 15).
Fig 9.
P300 amplitudes evoked by percussive sounds incongruent with the instrument shown (i.e., high initial prediction error).
The results for familiar and unfamiliar instruments were compared at every 40 exposures (N = 15).