Stability of Temperament Profiles in Early Childhood: A Latent Transition Analysis

Abstract

From a variable-centered perspective, early childhood represents a developmental period where temperament is moderately stable. There is evidence, however, of change among individual traits, drawing attention to emerging gaps in our understanding of how temperamental stability translates to the person-level. Additionally, how and when children transition between profiles across early childhood, and how those transitions relate to socioemotional development, is still unclear, as these remain novel research questions within the temperament literature. Therefore, my dissertation project aimed to address the following goals: identify best-fitting structures of early childhood (3-years, 4-years, 6-years) temperament using Latent Profile Analysis; assess structural stability of and latent transitions between enumerated profiles using tests of measurement invariance in a latent transition framework; and investigate differences in middle childhood externalizing behaviors based on transition probabilities across early childhood. Four qualitatively similar temperament profiles were identified at all three ages, Negative/Average, Reserved, Approachable, and Surgent, but an unconditional transition model fit data best between 3-year and 4-years, and 4-years and 6-years, suggesting poor structural stability. Transition analyses using the unconditional models suggested fewer transitions between 3-years and 4-years compared to 4-years to 6-years. The Approachable profile displayed fewest transition patterns during younger ages, followed by the Surgent profile during older ages. Differential relations between 6-year profiles and 9-year externalizing behaviors are also reported. Together, observations elucidate how temperamental stability manifests from a person-centered perspective and raises the question of how important temperamental transitions may be for socioemotional development, above and beyond predictive power of distinct latent profiles.