Skip to main content
. 2018 May 22;13(5):e0197263. doi: 10.1371/journal.pone.0197263

Table 1. The average values of the Bayesian information criterion (BIC) based on three feature-based RLs and their object-based counterparts, separately for each of four experiments.

Reported are the average values of BIC over all subjects (mean±s.e.m.) and p-values for comparisons of BIC values between each model and its object-based or feature-based counterparts (two-sided Wilcoxon signed-rank test). The overall best model (feature-based or object-based with decay) and its object-based or feature-based counterpart are highlighted in cyan and brown, respectively.

Model Coupled feature-based Uncoupled feature-based Feature-based with decay Coupled object-based Uncoupled object-based Object-based with decay
# pars. 5 5 6 4 4 5
Exp. 1 903.7±23.1 (p = 0.002) 950.3±17.8 (p = 10−16) 908.8±21.9 (p = 5*10−4) 930.1±17.9 974.7±14.8 934.6±17.4
Exp. 2 989.7±15.4 1019.7±10.6 994.3±15.2 959.3±15.7 (p = 0.054) 1016.1±10.9 (p = 0.64) 957.9±15.6 (p = 0.013)
Exp. 3 687.9±9.6 705.6±10.9 684.6±11.2 685.3±13.9 (p = 0.81) 688.5±11.3 (p = 0.26) 613.3±15.7 (p = 10−16)
Exp. 4 787.8±13.1 (p = 10−16) 789.6±13.7 (p = 2*10−4) 712.3±18.1 (p = 0.017) 844.6±8.3 845.3±7.8 734.4±13.7