Table 7.

Linear regression models to determine whether age 8 oral language ability mediated the effect of late talking (not combining words) at age 2 on academic skills at age 8.

Predictor	b	t	F	p
Reading comprehension
1. Direct effect of late talking (model fit)			12.8	.000
Late talking statusa	-15.52	3.7		.000
2. Mediator (oral language at 8) added as
predictor (model fit)			105.5	.000
Oral language at 8	3.92	13.4		.000
Late talking statusa	-6.35	2.2		.031
3. Late talking status as predictor of oral language at
8 (model fit)			8.4	.004
Late talking statusa	-2.34	2.9		.004
Broad math
1. Direct effect of late talking (model fit)			10.6	.001
Late talking statusa	-13.38	3.3		.001
2. Mediator (oral language at 8) added as
predictor (model fit)			69.0	.000
Oral language at 8	4.10	11.0		.000
Late talking statusa	-4.33	1.3		.194
3. Late talking as a predictor of oral language at 8
(model fit)			12.8	.000
Late talking statusa	-2.21	3.5		.000

Notes. The reading comprehension measure was the Passage Comprehension subtest standard score, and the math ability measure was the Broad Math subtest standard score, both of the Woodcock-Johnson Psychoeducational Battery, Revised (WJ-R). Oral language ability at age 8 was the sum of z-scores for Picture Vocabulary and Memory for Sentences subtests of the WJ-R, the Word Definitions subtest of the WASI, and the narrative total score from the SECCYD.

^aLate talking defined by an indicator variable for no word combinations at age 2 versus combining words at age 2.