Investigating the structure and etiology of temperament is key to understanding how children interact with the world (Kagan, 1994). Although these topics have yielded an abundance of research, fewer studies have employed observational data during middle childhood, when unique environmental challenges could influence temperament development. To address this gap, Israeli twin children were observed at Age 6.5 (N = 1,083, 564 families; 50.6% females) and again at Age 8-9 (N = 768, 388 families; 52.0% females; 611 children from 322 families had data from both ages). Temperament was assessed globally by trained coders and, at Age 8-9, also by the experimenter who interacted with the child. We examined whether Rothbart et al.'s (2000) three-factor model, according to which temperament includes the domains negative affect, positive affect/surgency, and effortful control, emerges from the data. In addition, we considered a bifactor model, where a fourth global factor accounts for all behaviors' commonality. Across the two ages and rating methods, confirmatory factor analyses supported the bifactor model. The global factor's loadings suggested that it reflects children's expressiveness. Adding this factor changed the associations between the other factors and enabled differentiation between surgency and positive affect. This suggests that in observational settings that capture temperament impressions holistically, children's expressiveness affects other traits' behavioral displays. Twin models revealed genetic influences for most traits. Importantly, twin models revealed shared-environmental influences for negative affect and expressiveness, which modestly contributed to temperament consistency across ages. These findings shed light on temperament traits' interrelatedness and stress the importance of the shared environment to temperament development during middle childhood. (PsycInfo Database Record (c) 2024 APA, all rights reserved).