Table 6. Effect sizes, unconditional standard errors, confidence intervals and relative importance for parameters included in the top models investigating players' responses to being punished for defecting in the previous round (player continued to defect = 0, player switched to cooperating = 1).

Player type	Partner cooperated	Parameter	Effect size	SE	Confidence Interval	Importance
Weak	NA	Intercept	1.85	0.19	(-2.21, -1.48)
		Game type (asymmetric)	0.66	0.28	(0.11, 1.22)	1.00
		Partner punished (yes)	0.38	0.34	(-0.29, 1.05)	1.00
		Partner cooperated (yes)	1.86	0.22	(1.42, 2.30)	1.00
		Game type x Partner cooperated	-1.28	0.47	(-2.20, -0.36)	0.72
		Punished punished x Partner cooperated	-0.88	0.50	(-1.85, 0.10)	0.45
		Game type x Partner punished	0.57	0.58	(-0.57, 1.71)	0.27
Strong	No	Intercept	-2.16	0.26	(-2.68, -1.64)
		Game type (asymmetric)	-0.15	0.42	(-1.00, 0.71)	0.22
		Partner punished (yes)	0.17	0.42	(-0.66, 1.00)	0.22
	Yes	Intercept	0.45	0.35	(-0.23, 1.22)
		Game type (asymmetric)	0.01	0.46	(-0.92, 0.89)	1.00
		Partner punished (yes)	-0.87	0.46	(-1.78, 0.02)	1.00
		Game type x Partner punished	2.43	0.87	(0.77, 4.20)	1.00