Our understanding of the stability in attachment during the first two decades of life is limited because i) ambiguity concerning the term ‘stability’ and ii) children have not been repeatedly followed up, leaving how attachment develops from childhood to adolescence uncharted. A birth cohort sample from Norway (n = 952) was examined with observational measures of attachment at ages 4, 6, 10, 12, 14, and 16 years to provide data on four aspects of stability: i) stability at the group level: avoidant and disorganized attachment decreased throughout development; ii) stability relative to the group increased with increasing age; secure attachment at ages 4 and 6 was weakly correlated with attachment in adolescence; other attachment strategies evinced no stability; iii) stability relative to oneself followed the above correlational pattern, but two latent classes both best described the individual trajectories of secure and ambivalent attachment; iv) stability of changes: changes in attachment styles from 4 to 6 did not forecast level or change in attachment in adolescence. Finally, I found little support for an early-formed prototype being responsible for stability. In sum, there was little continuity in attachment from childhood to adolescence, and the development in security and ambivalence might both follow two different trajectories.