Table 1.

Results and interpretations for our mixed-effects model regressing choice generalization on common reward, format, and serial position

Row	Effect:common reward x	Stats	Interpretation
1	1	b = 0.41, F(1,444) = 28.31, p = 2e-7a	positive CRE, i.e., participants generalize their choice more when the common outcome is rewarded versus non-rewarded
2	Format	F(2,444) = 0.3, p = 0.75	no evidence CRE varies as a function of format
3	Position	b = 0.04, F(1,444) = 0.11, p = 0.738	no evidence CRE varies as a function of serial outcome position
4	Position x format	F(2,444) = 3.08, p = 0.047a	CRE varies across the three formats in different ways for the first/second outcomes
5	Position x (PI versus S)	b = 0.08, t(444) = 0.29, p = 0.771
6	Position x (RI versus S)	b = 0.58, t(444) = 2.25, p = 0.025a	CRE varies as a function of format (for RI versus S, but not for other format pairs) in different ways for the first and second outcomes
7	Position x (RI versus PI)	b = 0.5, F(1,444) = 3.54, p = 0.061
8	First outcome: RI versus S	b = 0.18, F(1,444) = 0.95, p = 0.33	for the first outcome, CRE is similar in standard and retrospective inference
9	Second outcome: RI versus S	b = −0.40, F(1,444) = 4.03, p = 0.045a	for the second outcome, CRE is lower in retrospective inference than in standard

The first column indexes row numbers. The second column specifies the effect. We are only interested in effects involving the common reward and hence this factor is included in all effects. For example, the name “position” corresponds to the double common reward x position interaction. The third column specifies statistical tests for each effect. The fourth column specifies the interpretation of the finding. CRE, common reward effect.

^aSignificant effects.