Table 3.

Factors that influenced the likelihood of dislodging the other reward in experiment 2

Parameter	Estimate	SE	Z	P	95% bounds
					Upper	Lower
Model 1*
Condition	1.25	0.16	1.72	0.085	0.97	1.62
Position	0.90	0.01	−9.11	<0.001	0.88	0.92
Session	0.81	0.02	−9.05	<0.001	0.77	0.85
Trial	1.07	0.14	0.50	0.614	0.83	1.3
Hannah	0.25	0.09	−3.87	<0.001	0.12	0.50
Huey	1.31	0.43	0.82	0.413	0.69	2.4
Jessie	1.11	0.34	0.35	0.724	0.61	2.02
Karin	0.87	0.26	−0.47	0.640	0.48	1.56
Kelly	3.40	1.35	3.08	0.002	1.55	7.40
Martha	1.06	0.45	0.13	0.895	0.46	2.43
Moose	7.99	5.82	2.85	0.004	1.91	33.32
Pepper	2.04	0.62	2.35	0.019	1.12	3.70
Punch	1.49	0.45	1.32	0.188	0.82	2.71
Sandy	1.73	0.57	1.66	0.096	0.91	3.29
Model 2†
Condition	1.06	0.18	0.34	0.732	0.76	1.47
Position	1.03	0.41	0.09	0.931	0.48	2.24
Session	0.83	0.02	−7.21	<0.001	0.79	0.87
Trial	0.90	0.01	−8.13	<0.001	0.88	0.93

^*The binary logistic regression model included categorical variables for individuals as predictors in addition to the variables shown. Variables were coded so that the odds ratios would exceed 1 for condition if actors were more likely to provide the other reward when a recipient was present. Position was coded so that the odds ratio would exceed 1 if the actors were more likely to provide the other reward if it was positioned on the top. For trial and session, the odds ratio would exceed 1 if chimpanzees were more likely to provide the other reward as the experiment progressed within trials or across sessions. An odds ratio of less than 1 for both session and trial indicates learning: the chimpanzees were less likely to provide the other reward across trials within a session and across sessions. The odds ratio for each individual indicates whether these 10 subjects were more or less likely to provide the other reward than was Coco. This analysis was conducted on those trials in which actors chose at least one tray.

^†The binary logistic regression model used clustered robust standard errors to calculate the confidence intervals for predictor variables.