How well do poor language scores at ages 3 and 4 predict poor language scores at age 6?

Abstract
The purpose of this study was to describe findings and associations at the group level, and predictive accuracy at the individual level, for three measures of language obtained from a single prospective cohort of US children assessed at three ages during another investigation. Participants comprised all children (n = 414) who had a score at each of three ages (3, 4, and 6 years) on three language measures: Peabody Picture Vocabulary Test-Revised (PPVT); mean length of utterance in morphemes (MLU); and total percentage of phonemes correct (TOTPPC) in a nonword repetition task. At the group level, mean differences and correlations over ages were calculated. At the individual level, the extent to which low scores (1.5 SDs or more below the sample mean) at an earlier age increased the relative risk (R’) of low scores at a later age was calculated. At the group level, scores on all measures increased significantly with age. Earlier and later scores on each measure were significantly correlated; r2 values were generally modest (12-59%). A low score at age 3 did not increase the risk of a low score age 4 or age 6 for any of the measures; a low PPVT score at age 4 significantly increased the risk (R’ = 5.0) of a low PPVT score at age 6. For these generally healthy children and these language measures, predictive accuracy at the individual level was generally poor between the ages of 3, 4, and 6 years. Large, longitudinal studies are needed to identify and validate measures for use in identifying preschool children at risk for later language deficits.

from the International Journal of Speech-Language Pathology