Table 1. Summaries of the mixed-effects regression models presented in Figure 3a: Models for male speakers’ sibilant tokens. Models were made using a treatment coding scheme with affricate (vs. fricative) and /i/ (vs. /a/) as reference levels

	Word-initial tokens
	Lax		Tense
	Estimate (Std.E)	p-value	Estimate (Std.E)	p-value
(Intercept)	4,926 (415)	<.0001	5,008 (367)	<.0001
Manner(vs. Fricative)	−414 (407)	0.315	82 (416)	0.845
Birthyear(year 2008=0)	0 (14)	0.987	−6 (13)	0.629
Vowel (vs. /a/)	−202 (428)	0.64	−31 (494)	0.95
Manner×Birthyear	−3 (14)	0.819	10 (15)	0.484
Manner×Vowel	3,395 (475)	<.0001	3,857 (588)	<.0001
Vowel×Birthyear	−15 (15)	0.317	17 (16)	0.31
Manner×Birthyear×Vowel	17 (17)	0.317	8 (20)	0.69

	Word-medial tokens
	Lax		Tense
	Estimate (Std.E)	p-value	Estimate (Std.E)	p-value
(Intercept)	5,652 (529)	<.0001	5,172 (347)	<.0001
Manner(vs. Fricative)	–1,151 (586)	0.06	−224 (335)	0.505
Birthyear(year 2008=0)	5 (23)	0.823	5 (12)	0.685
Vowel (vs. /a/)	26 (568)	0.964	593 (363)	0.106
Manner×Birthyear	−8 (24)	0.742	−2 (12)	0.864
Manner×Vowel	3,842 (602)	<.0001	3,846 (463)	<.0001
Vowel×Birthyear	−6 (23)	0.787	−2 (12)	0.9
Manner×Birthyear×Vowel	20 (25)	0.404	26 (16)	0.108