Table 3.

Outputs of different linear mixed model testing the effects of time and zone (touristic or control) on cortisol production for both trial [model (i)] and at each trial [model (ii), model (iii)].

Model	Cortisol ~ Zone + Time + (1|Trial/Aquarium)
(i) June + August	Random effects
		Groups	Name	Variance	Std. Dev.	n
		Aquarium:Trial	(Intercept)	0.055	0.23	20
		SES	(Intercept)	0.004	0.06	2
		Residual		0.17	0.41
		Number of observations: 100
	Fixed effects
		Estimate	Std error	t-value	p-value
	(Intercept)	0.45	0.11	4.1	0.025	*
	Zone:Touristic	0.42	0.13	3.2	0.005	**
	Time	0.08	0.024	3.5	<0.001	***
	Cortisol ~ Zone + Time + (1+Time|Aquarium)
(ii) June	Random effects
		Groups	Name	Variance	Std. Dev.	N
		Aquarium	(Intercept)	0.07	0.26	8
		Time	(Slope)	0.01	0.12
		Residual		0.07	0.27
		Number of observations: 40
	Fixed effects
		Estimate	Std error	t-value	p-value
	(Intercept)	0.34	0.14	2.5	0.03	*
	Zone:Touristic	0.34	0.17	2.02	0.066	.
	Time	0.12	0.05	2.5	0.038	*
	Cortisol ~ Zone + Time + (1|Aquarium)
(iii) August	Random effects
		Groups	Name	Variance	Std. Dev.	n
		Aquarium	(Intercept)	0	0	12
		Residual		0.18	0.43
		Number of observation: 60
	Fixed effects
		Estimate	Std error	t-value	p-value
	(Intercept)	0.52	0.09	5.8	<0.001	***
	Zone:Touristic	0.49	0.11	4.4	<0.001	***
	Time	0.06	0.03	1.8	0.079	.

Models with additive terms (zone + time) and interaction between terms (zone x time) were first compared using the Akaike's Information Criterion (AIC) so that the most parsimonious model was conserved.

^***< 0.001,

^**< 0.01,

^*< 0.05.