Table 1.

Summary of results of quality of models using TP and ADB to predict human ratings of similarities of role–filler pairs. AIC, Akaike Information Criterion; BIC, Bayesian Information Criterion.

model	AIC	BIC	estR2
rating∼TP	261 721	261 754	0.58
rating∼ADB	258 088	258 121	0.63