The model comparison and selection as well as the full details of the final LMM that best fit the data can be found in the S4 section of the Supplementary Materials. This LMM did not support out hypotheses H1, H2 and H3 regarding the main effects of Sociality and Valence or an interaction effect between sociality and valence, respectively. Specifically, respondents rated alone and social scenarios, as well as positive and negative scenarios with the same likelihood.
However, we generally found support for our hypotheses H4, H5 and H6 concerning the moderating effect of sociality on SCM model. First, there was a three-way interaction between valence, warmth and sociality (H4; F (1, 36597.0) = 183.67, p < .001). For alone situations, there was a social optimism bias for the warm characters (H4a; (M diff = M diff = 9.40%, t (44.5) = 2.15, p = .032) but no bias for the cold characters (M diff = -4.03%, t (44.5) = -.91, p = .361; Figure 1). For social situations, there was a much bigger social optimism bias for the warm characters (H4b; (M diff = M diff = 14.84%, t (44.5) = 3.37, p < .001) and a pessimism bias for the cold characters (M diff = -11.49%, t (44.5) = -2.62, p = .009; Figure 1).
Second, there was a three-way interaction between valence, sociality, and
competence (H5; F (1, 36597.0) = 86.93, p < .001). For alone situations, the respondents displayed an optimism bias for competent characters (H5a; M diff = 10.95%, t (44.5) = 2.50, p = .012) but no bias for the incompetent characters (M diff = -5.57%, t (44.5) = -1.27, p = .204; Figure 2). For social situations, there was no bias toward either competent characters (M diff = 5.49%, t (44.5) = 1.25, p = .212) or incompetent characters (M diff = -2.15%, t (44.5) = .49, p = .624; Figure 2).
Finally, there was a four-way interaction between valence, sociality, warmth, and competence (F (1, 36597.0) = 54.78, p < .001) that generally supported our hypothesized directions (H6). For the warm-competent character (H6a), there was social optimism bias during social situations (M diff = 13.13%, t (23.9) = 3.54, p = .002) but not alone situations (M diff = 8.61%, t (22.9) = 1.71, p = .101; Figure 3). Similarly, respondents manifested an optimism bias towards the warm-incompetent character (H6b) during social situations (M diff = 16.55%. t (23.9) = 4.46, p < .001) but no bias during alone situations (M diff = 10.20%, t (22.9) = 2.02, p = .055; Figure 3). For the cold-competent character (H6c), there was an optimism bias during alone scenarios (M diff = 13.29%, t (22.9) = 2.64, p = .015) but no bias during alone scenarios (M diff = -2.14%, t (23.9) = -.58, p = .569; Figure 3). By contrast, respondents manifested a pessimism bias during both social (M diff = -20.85%, t (23.9) = -5.62, p < .001) and alone scenarios (M diff = -21.34%, t (22.9) = -4.23, p < .001; Figure 3) toward the cold-incompetent character (H6d).
None of our hypotheses about attachment anxiety and attachment avoidance were supported. There was no interaction effect between attachment avoidance and sociality (H7; F (1,36410.0) = .21, p = .649), meaning that respondents rated alone and social events irrespective of the levels of attachment avoidance. Furthermore, there was no interaction between attachment avoidance, sociality and valence (H8; F (1,36410.0) = .22, p = .635), suggesting that there was no optimistic or pessimistic bias for either alone or social events, regardless of the levels of attachment avoidance. Lastly, there was no interaction between attachment avoidance, sociality, valence and competence (H9; F (1,34610.0) = .14, p = .707). However, there was a significant three-way interaction between attachment avoidance, valence and competence that was not predicted (F 1,36406.0) = 6.04, p = .014. Specifically, there was no bias for incompetent characters, regardless of the levels of attachment avoidance (Mean - 1 standard deviation (SD): t (46.4) = -1.16, p = .245; Mean: t (46.4) = -1.23, p = .218; Mean + 1 SD: t (46.4) = -1.26, p = .207). However, there was a significant optimism bias for competent characters whose magnitude decreased as attachment avoidance increased (Mean - 1 SD: t (46.4) = 3.00, p = .003; Mean: t (46.4) = 2.63, p = .009; Mean + 1 SD: t (46.4) = 2.17, p = .030). This goes against the predicted direction in H9. There was also no interaction effect between attachment anxiety and sociality (H10; F (1,34610.0) = .29, p = .592), meaning that respondents rated alone and social events irrespective of the levels of attachment anxiety. Furthermore, there was no interaction between attachment anxiety, sociality and valence (H11; F (1,34610.0) = .36, p = .546), suggesting that there was no optimism or pessimism bias for either alone or social events, regardless of the reported levels of attachment anxiety. Lastly, there was no interaction between attachment anxiety, sociality, valence and warmth (H12; F (1,34610.0) = .32, p = 574).
Exploratory analyses. Because the mental schemes about others triggered by an individual’s attachment style are confined to how responsive others would be to their own needs, social cognition in individuals with an unsecure attachment style is egocentric, revolving around the question of how relevant others are to their own needs and fears 10,22,43−45. By contrast, stereotypical thinking along dimensions of warmth and competence are cognitions and emotions widely held by society that transcend the individual needs and fears of the members of that society 15,16,18. A possible reason why our hypotheses about attachment avoidance and anxiety were not supported could be because dimensions of warmth and competence are not relevant enough for the respondent in order for the attachment styles to manifest. To test this possibility, we ran an exploratory analysis replacing warmth and competence with scores of the Inclusion of Others in the Self (IOS; see Supplementary Materials), which indexes the degree of relevance of others to oneself 46,47. This analysis showed that attachment anxiety and avoidance interact with the degree of self-relevance of the social target to modulate the magnitude of the optimism bias (S5 section of the Supplementary Materials). First, there was an interaction between valence and IOS (F (1,30199.08) = 1062.03, p < .001) with respondents being generally pessimistic toward social targets with low self-relevance (t (45.96) = -4.75, p < .001) but optimistic towards social targets with high self-relevance (t (45.96) = 6.15, p < .001). However, this effect was qualified by further interactions between valence, IOS and sociality of events (F (1,36796.7) = 28.05, p < .001; Figure 4) as well as by three-way interactions with attachment anxiety (F (1,28958.5) = 7.86, p = .005) and attachment avoidance (F (1,31615.8) = 4.07, p = .044), respectively. Respondents were more optimistically biased toward social targets at high scores of IOS when the outcomes were social (M diff = 23.57%; t (45.96) = 4.85, p < .001) than alone (M diff = 15.20%; t (45.96) = 3.90, p < .001; Figure 4).
While the attachment dimensions did not affect the levels of pessimism at low IOS, they buffered the levels of optimism at high IOS. Specifically, respondents were progressively less optimistically biased toward highly self-relevant social targets (Mean + 2 SD) as attachment anxiety increased from very low (t (45.96) = 7.01, p < .001) to average (t (45.96) = 6.15, p < .001) to very high (t (45.96) = 3.88, p < .001; Figure 5) as well as when attachment avoidance increased from very low (t (45.96) = 6.45, p < .001) to average (t (45.96) = 6.15, p < .001) to very high (t (45.96) = 4.46, p < .001; Figure 5). For attachment anxiety, the pattern was driven by an increasing slope for negative events (high IOS: t (190.8) = 3.82, p = .001; Figure 5) but not for positive events (high IOS: t (190.8) = -.37, p = .709; Figure 5). By contrast, the pattern of findings for attachment avoidance was driven by the likelihood estimates for positive scenarios (high IOS: t (190.8) = -2.09, p = .037; Figure 5) but not for negative scenarios (high IOS: t (190.8) = .48, p = .628; Figure 5). These patterns suggest that social optimism bias is maximal among the respondents with a secure attachment style (low attachment anxiety/avoidance) that identify the most with the social targets.