There is a widespread assumption that children go through two stages in their acquisition of sociolinguistic variation and a stable adult vernacular. First, they acquire variation that mirrors their adult female caregiver's speech (transmission). Then, at around age four or five, they depart from this model and accelerate ongoing sound changes by incrementally using more innovative variants as they age (incrementation). These changes are collectively known as ‘vernacular reorganisation’ (Labov 2001). The tipping point that is thought to trigger the switch from transmission to incrementation is the shift a child makes from “the caregiver-dominated norms of the home to the peer-dominated norms of the wider world” when they enter the school system (Smith & Holmes-Elliott 2022: 98).

In this paper, we investigate the contemporary accuracy of the vernacular reorganisation model. We explore a longitudinal corpus of early child speech (3;11 years to 5;5 years; 131 children) consisting of a story retell task carried out in Christchurch early childhood education centres. The vocalic production of these children is interpreted with respect to changes in progress in the community, measured using the QuakeBox corpus (18-85 years; 251 talkers). Using linear mixed-effects regression models, fit within a Bayesian framework, we find that children appear to start from a more conservative position in the New Zealand English short front vowel shift than would be expected from caregiver transmission. That is, the children 'lag' behind the youngest female speakers in the QuakeBox corpus. We explore these results in light of our understanding of the changing social world of children since Labov's initial proposal of the vernacular reorganisation model.