Fig. 3
Results of computational modeling. a Illustration of the tuning shift (TS) model (Left) and the tuning competition (TC) model (Right). The TS model postulates that attention shifts the tuning of individual neurons (gray curves) towards the attended feature, which results in a uni-modal population response profile (black dashed curve) centered at the attended feature. In particular, when two features are attended, individual neuronal tunings would shift periodically between these two features in a similar manner, thus leading to a rhythmic oscillation of the population response profile center. In contrast, the TC model postulates an ongoing competition between the neuronal populations tuned to the attended features (red and blue curves) in multi-feature attention while the other neurons remain little affected. This leads to a multi-modal population response profile with local peaks at the attended features that rhythmically alternate in dominance. b, c Model comparison results. b Goodness-of-fit (GoF) distributions across all time points pooled from all subjects. c RMSD values averaged across all time points plotted as paired red dots for individual subjects. d The modeled tuning width difference between the two neuronal populations plotted as a function of time (left) and its spectrum (right). The smoothed time course (in black) is overlaid on the unsmoothed time course (in gray). The gray dashed line denotes the statistical threshold for the spectrum (corrected for multiple comparisons)
