Fig. 2.

Results for regions of interest (ROIs).

Results for areas V1-3 are shown as bar plots. Bar colours indicate conditions: audiovisual congruent in green, audiovisual incongruent in red, visual only in light grey and audio only in dark grey. Error bars indicate the standard error of the mean adjusted for repeated measurements (Morey, 2008).

• a)
  Classification accuracies for 4-way classification using linear support vector machines (see Materials and methods for details). The dashed line indicates chance level (.25). Stars indicate significantly different decoding accuracies between conditions involving visual stimulation (as indicated by paired t-tests, see Results section for details of respective ANOVAs; *p < .05, **p < .01).
• b)
  Mean signal amplitudes estimated by the GLM. Note that amplitudes were not significantly different between conditions involving visual stimulation in any of the regions of interest. Note that beta maps used for this analysis were not mean corrected (see Materials and methods for details).
• c)
  Pattern reliability as indicated by means of Fischer z-transformed correlation coefficients between patterns for a given stimulus in odd and even runs (see Materials and methods for details). Stars indicate significantly different pattern reliabilities between conditions involving visual stimulation (as indicated by paired t-tests, see Results section for details of respective ANOVAs; *p < .05, **p < .01, ***p < .001).
• d)
  Pattern similarity as indicated by means of Fischer z-transformed correlation coefficients between patterns for different stimuli (see Materials and methods for details). Note that pattern similarities were not significantly different between conditions involving visual stimulation in any of the regions of interest. Patterns are negatively correlated because they were mean corrected across stimuli within each condition (see Materials and methods for details).