Figure 1.
Illustration of the paradigm and systematic matrix of the four trial categories.
Cue-sound combinations STA (frequent cue and tone, exemplarily displayed as high-pitched), A, V and VA are shown with their different probabilities. At least two standards precede each deviant as shown for type STA, the relevant tone is colored. The trial starts with the presentation of the high or low note symbol (cue), followed by one of the two tones (target) after an SOA of 600 ms. Type V and type A represent situations with contradictory auditorily and visually induced predictions whereas type VA means to violate redundant predictions from the two modalities.
Figure 2.
Grand-averaged, unfiltered ERPs of the complete trial (nose referenced).
Cue and tone onset are marked for all four trial categories (green: frequent cue before frequent tone, STA; blue: rare cue before rare tone, A; red: rare cue before frequent tone, V; purple: frequent cue before rare tone, VA). Left: Visual ERPs to the cues are best observable in occipital electrodes (here Oz). Right: Please note that the data were not filtered to clearly show the CNVs which are influenced by the cue probabilities (more pronounced for the rare cue, type A and V). Tones were presented with an onset 600 ms after the trial started, eliciting auditory ERPs in fronto-central regions. Negative is plotted upwards.
Figure 3.
Filtered (1.3–100 Hz bandpass filter) and grand-averaged auditory ERPs and difference waves (nose referenced).
Left: Auditory ERPs, elicited by the four types of cue-sound combinations (green: STA, meaning no violation; blue: A, meaning violation of auditory-auditory regularity; red: V, meaning violation of visual-auditory regularity; purple: VA, meaning concurrent violation of both regularities). Right: For every deviant the respective deviant-minus-standard difference waveform is shown (same filter setting and color code). The prediction error signals of MMN, IR and IRMMN correspond to the marked time window of 105–130 ms for the deviant types A, V and VA, respectively. Negative is plotted upwards.
Figure 4.
Topographic, SCD and VARETA SPM maps of the error signal components.
Left: Potential maps (nose reference) show the scalp distributions of the difference data for MMN, IR and IRMMN (105–130 ms). Middle: SCDs for the three components of IR, MMN and IRMMN are compatible with sources in auditory areas of the temporal cortices. Right: SPMs of the source reconstructions of the three prediction error signals MMN, IR and IRMMN (p<.01). Coronal slices which contain the centre of gravity of the inverse solution are displayed (all within the Superior Temporal Gyrus). The Y-coordinates represent the location of the coronal slice in MNI space.
Figure 5.
Illustration of the additive model.
Panel A: The difference waveform (black line) of measured IRMMN (VA-STA) and the sum of the mean amplitudes of IR (V-STA) and MMN (A-STA) shows values around zero µV until the end of the time window of prediction error signals. Negative is plotted upwards. Together with the similar voltage distributions of measured and modelled component (panel B), this supports the assumption of functional independence of IR and MMN. Panel C: Bar graphs for potential and PCD data, showing amplitudes (left: mean amplitude of ROIs lateral) and PCDs (right) for modelled (IR+MMN) and measured (IRMMN) concurrent violation of sound prediction. For VARETA source analysis, mean PCD values of the centers of gravity were averaged for both hemispheres in the time window of 105–130 ms. Error bars represent the standard error of mean.
Figure 6.
RTs (left) and accuracy data (right).
Tested pairs with their significance levels are tagged (STA: frequent cue before frequent tone vs. A: rare cue before rare tone; V: rare cue before frequent tone vs. VA: frequent cue before rare tone).