Table S2.

Comparison between different models of social decision

ObservationFixed predictorRandom predictorAICLogLikP
Choice preference in offer J-SMutual gaze rate in offer J-SChoice preference in offer A-SActor and partner identity1,178−585<0.001
Choice preference in offer J-SActor and partner identity1,254−625
Choice preference in offer A-SActor's blink rate in offer A-nSActor's and partner's blink rate in offer A-SChoice preference in offer J-SActor and partner identity674−350<0.001
Choice preference in offer A-SActor and partner identity704−332
Choice preference in offer A-SActor's blink rate in offer A-nSActor's and partner's blink rate in offer A-SChoice preference in offer J-SActor and partner identity674−350<0.001
Choice preference in offer A-SActor's and partner's blink rate in offer A-SChoice preference in offer J-SActor and partner identity886−439
Choice preference in offer A-SActor's blink rate in offer A-nSActor's and partner's blink rate in offer A-SChoice preference in offer J-SActor and partner identity674−350<0.001
Choice preference in offer A-SActor's blink rate in offer A-nSActor's blink rate in offer A-SChoice preference in offer J-SActor and partner identity690−340
Choice preference in offer A-nSActor's blink rate in offer A-nSActor and partner identity657−324<0.001
Choice preference in offer A-nSActor and partner identity684−339
Mutual gaze rate in offer J-SActor's blink rate in offer A-SActor and partner identity−192100<0.001
Mutual gaze rate in offer J-SActor and partner identity−15178
Actor's blink rate in offer A-SPartner's blink rate in offer A-SActor and partner identity−282145<0.05
Actor's blink rate in offer A-SActor and partner identity−278143
  • The theoretical likelihood ratio test was used to estimate the validity of the mixed effects analyses on session-by-session performances. Here, null models with only the random effects are compared with the models of interest shown in Table 1, to test whether the latter have significantly better predictive value. The AIC (Akaike information criterion) and the Loglik (Log-likelihood) are measures of the relative quality of statistical models for a given set of data. Gray cells highlight the difference predictors used to evaluate the model specified in the row above it.