Do not like him”, “He is mean”, etc.) was coded as 21. A score of 0 was assigned if the child gave no response or a response that was too ambiguous to code one way or the other. Then, we computed an aggregate Absolute APTO-253 manufacturer Valence Index by averaging the code of the responses to each question. For the contrastive questionnaire, we computed a Relative Valence Index (RVI) based on toddlers’ responses to comparative questions: each child’s response was coded as +1 if the child responded in favor of the pro-social agent or in disfavor of the antisocial agent, 21 if he or she gave the opposite responses. A score of 0 was assigned if the child gave no response or a response too ambiguous to code one way or the other. Here again, we computed an aggregate Relative Valence Index by averaging the code of the response to each question. The RVI was between 21 and +1, a positive value indicating that the “pro-social agent” was globally evaluated more positively than the “anti-social agent”, and vice-versa for a negative RVI. The average AVI for each agent (pro-social vs. anti-social) across toddlers is shown in Figure 1.A. To start, we were interested to know if, irrespective of counterbalancing effects, toddlers found the pro-social agent to be positive and the anti-social agent to be more negative. To this end, we constructed a separate linear model for each agent type using the AVI as the dependent variable. Included in the model were the three counterbalancing factors (actor, order of actions and order of patient) and the resultant intercept was tested against zero. There were no effects of counterbalancing factor on either the pro-social or anti-social agents Furthermore, we found that the AVI obtained for the prosocial agent was slightly negative but not significantly BLU-554 site different from zero (AVI = 20.01, SE = 0.09, F(1,27) ,1, p.0.1, gp2 = 0.001) whereas, the AVI obtained for the anti-social agent was negative and significantly below zero (AVI = 20.19, SE = 0.09, F(1,27) = 6.29, p,.05, gp2 = .34). We then ran a repeated-measures ANOVA to examine the difference between pro-social and anti-social AVI. This analysis indicated that the AVI differed significantly as a function of the agent, (F(1,12) = 10.71, p,.02, gp2 = .47) such that the anti-social agent is evaluated more negatively than the pro-social one. No effects of counterbalancing factor or of participant sex were found in these analyses. The average RVI across toddlers is shown in Figure 1.B and was analyzed using a linear model with the three counterbalancing factors (actor, order of actions, and order of patient) as between subject factors. As above, the intercept was tested against zero. This analysis revealed an intercept significantly above zero (RVI = 0.26, SE = 0.09, F(1,27) = 5.14, p,.05, gp2 = .25) showing that toddlers evaluated the pro-social agent significantly more positively than the anti-social agent. Again, there were no effects ofPLOS ONE | www.plosone.orgFigure 1. Average indexes for 29 months-olds. A. Average Absolute Valence Index for 29-months-old toddlers computed over the responses to the individual questions regarding the pro-social agent and the anti-social agent separately. *p,.05 (between subjects),**p, .02 (within subjects). A positive score indicates a positive verbal statement toward the agent and vice versa for negative scores. The error bars correspond to one between-subject standard error above and below the mean. B. Average Relative Valence Index computed.Do not like him”, “He is mean”, etc.) was coded as 21. A score of 0 was assigned if the child gave no response or a response that was too ambiguous to code one way or the other. Then, we computed an aggregate Absolute Valence Index by averaging the code of the responses to each question. For the contrastive questionnaire, we computed a Relative Valence Index (RVI) based on toddlers’ responses to comparative questions: each child’s response was coded as +1 if the child responded in favor of the pro-social agent or in disfavor of the antisocial agent, 21 if he or she gave the opposite responses. A score of 0 was assigned if the child gave no response or a response too ambiguous to code one way or the other. Here again, we computed an aggregate Relative Valence Index by averaging the code of the response to each question. The RVI was between 21 and +1, a positive value indicating that the “pro-social agent” was globally evaluated more positively than the “anti-social agent”, and vice-versa for a negative RVI. The average AVI for each agent (pro-social vs. anti-social) across toddlers is shown in Figure 1.A. To start, we were interested to know if, irrespective of counterbalancing effects, toddlers found the pro-social agent to be positive and the anti-social agent to be more negative. To this end, we constructed a separate linear model for each agent type using the AVI as the dependent variable. Included in the model were the three counterbalancing factors (actor, order of actions and order of patient) and the resultant intercept was tested against zero. There were no effects of counterbalancing factor on either the pro-social or anti-social agents Furthermore, we found that the AVI obtained for the prosocial agent was slightly negative but not significantly different from zero (AVI = 20.01, SE = 0.09, F(1,27) ,1, p.0.1, gp2 = 0.001) whereas, the AVI obtained for the anti-social agent was negative and significantly below zero (AVI = 20.19, SE = 0.09, F(1,27) = 6.29, p,.05, gp2 = .34). We then ran a repeated-measures ANOVA to examine the difference between pro-social and anti-social AVI. This analysis indicated that the AVI differed significantly as a function of the agent, (F(1,12) = 10.71, p,.02, gp2 = .47) such that the anti-social agent is evaluated more negatively than the pro-social one. No effects of counterbalancing factor or of participant sex were found in these analyses. The average RVI across toddlers is shown in Figure 1.B and was analyzed using a linear model with the three counterbalancing factors (actor, order of actions, and order of patient) as between subject factors. As above, the intercept was tested against zero. This analysis revealed an intercept significantly above zero (RVI = 0.26, SE = 0.09, F(1,27) = 5.14, p,.05, gp2 = .25) showing that toddlers evaluated the pro-social agent significantly more positively than the anti-social agent. Again, there were no effects ofPLOS ONE | www.plosone.orgFigure 1. Average indexes for 29 months-olds. A. Average Absolute Valence Index for 29-months-old toddlers computed over the responses to the individual questions regarding the pro-social agent and the anti-social agent separately. *p,.05 (between subjects),**p, .02 (within subjects). A positive score indicates a positive verbal statement toward the agent and vice versa for negative scores. The error bars correspond to one between-subject standard error above and below the mean. B. Average Relative Valence Index computed.