Table 1.

Results of the LMMs testing how the length, width and the formation of subgroups were affected by the playback of close calls and the playback of background noise. Significant effects are highlighted in italics. Owing to the Bonferroni correction for multiple testing, the α level was reduced from 0.05 to 0.017 to account for multiple testing. The time steps are the 2 min intervals within the playback at which the length, width and gap size were estimated. The intercept of the model represents the predicted values for the control playback of background noise at time step 0. The coefficients of the model (estimate) give the difference in the mean of the levels of a categorical explanatory variable and the slope for continuous variables. It also indicates the direction of the effect of a factor (positive or negative) on the response variable.

response variable	explanatory variables	estimate	confidence intervals [0.025, 0.975]	s.e	d.f.	t-value	p-value
length (m)	intercept	16.84	[12.60, 21.09]	2.17	13	7.77	<0.001
	playback ‘cc’	7.84	[4.43, 11.25]	1.74	162	4.51	<0.001
	time step	0.33	[0.12, 0.54]	0.11	162	3.09	0.002
	playback ‘cc’ : time step	−0.26	[−0.56, 0.03]	0.15	162	−1.76	0.080
gap size (m)	intercept	0.39	[−0.17, 0.94]	0.28	33	1.36	0.183
	playback ‘cc’	0.90	[0.27, 1.53]	0.32	162	2.78	0.006
	time step	0.00	[−0.04, 0.04]	0.02	162	−0.13	0.898
	playback ‘cc’ : time step	0.01	[−0.05, 0.06]	0.03	162	0.28	0.778
width (m)	intercept	23.58	[19.07, 28.10]	2.30	23	10.24	<0.001
	playback ‘cc’	−1.58	[−6.27, 3.12]	2.40	162	−0.66	0.511
	time step	−0.05	[−0.33, 0.24]	0.15	162	−0.31	0.755
	playback ‘cc’ : time step	0.15	[−0.26, 0.55]	0.21	162	0.71	0.479