The Barnes, Gutfreund, Satterly and Wells study

Of the research outlined in Chapter 10, BGSW's use of the Bristol child language corpus (Wells, 1985) gives them the strongest claim to a representative sample of the population, and naturalistic and spontaneous speech undistorted by the presence of an observer (see Section 2.1). BGSW selected sixteen boys and sixteen girls with a broad range of family background and coded the speech addressed to them by adults in the recording when the child' s MLUS was closest to 1.5 morphemes (Tl). Because intervals between speech sampling were three months in the Bristol Study, exact matching for MLU was not possible and actual values for MLUS ranged from 1.0 to 2.21 (mean = 1.68). The mean age at this point was 2;0 (range l;6–2;9). To assess the children's progress, transcripts from a point nine months later (T2) were used. Children's gains were indicated by residual gain scores, the difference between predicted scores at T2 and actual scores at T2 (see Barnes et al., 1983; O'Connor, 1972). For the auxiliary there were three measures: range of auxiliary meanings, auxiliary frequency, and range of auxiliary forms. Meanings and forms are as defined in Wells (1979a), forms being major forms rather than subforms (see Section 2.2.2). Only the correlations with range of auxiliary meanings are reported.

Strengths and limitations of BGSW

1. Size of population sample: BGSW's sample of thirty-two children is the largest of the published studies and only HG1 and HG2 approach this figure (N = 22). NGG's original sample of fifteen is reduced to two groups of six in the reanalysis (GNG). For FNB, SW, and YK, sample sizes are seven, nine and ten respectively.

2. […]