Abstract The composition of executive function (EF) in preschool children was examined using confirmatory factor analysis (CFA). A sample of 129 children between 3 and 5 years of age completed a battery of EF tasks. Using performance indicators of working memory and inhibition similar to previous CFA studies with preschoolers, we replicated a unitary EF factor structure. Next, additional performance indicators were included to distinctly measure working memory, set shifting, and inhibition factors. A two-factor model consisting of working memory and inhibition fit the data better than both a single-factor model and a three-factor model. Findings suggest that the structure of EF in preschoolers that emerges from CFA is influenced by task and performance indicator selection. ACKNOWLEDGMENTS This research was supported in part by a grant from the Human Early Learning Partnership to Ulrich Müller and by a grant from Research in Early Education and Childhood Health to Kimberly A. Kerns. Notes 1Similar results emerged when performance was measured in terms of number of correct trials. 2Similar results emerged when performance was measured both in terms of number of correct trials and in terms of number of moves for two- and three-move trials combined. Because illegal moves were not recorded, we were unable to test a ratio of the number of illegal moves to the total number of moves (see Wiebe et al., Citation2008). P-CPT = Preschool Continuous Performance Test; GNG = Go/No-Go; DCCS = Dimensional Change Card Sort; PPVT-III = Peabody Picture Vocabulary Test-Third edition. Note. Variables 4, 5, and 9 are reverse-scored to ease interpretation. P-CPT = Preschool Continuous Performance Test; GNG = Go/No-Go with block number; DCCS = Dimensional Change Card Sort; PPVT-III = Peabody Picture Vocabulary Test-Third edition. † p ≤.10. *p ≤.05. **p ≤.01. 3Scaling and method variance also were similar for Go/No-Go Block 2 hit ratio and Go/No-Go Block 3 hit ratio. However, correlating the error variances between these two variables resulted in nonsignificant values at p > .05. Therefore, the variance values for these two Go/No-Go blocks remained unconstrained for all estimated models. CPT = Continuous Performance Test; DCCS = Dimensional Change Card Sort. Note. The preferred models are italicized. a Lower values indicated better model fit; values with p ≤ .05 indicated that the model did not fit the data better than a saturated model. b Values ≤ 3 indicated good model fit. c Values ≥ 0.95 indicated good model fit; values ≥ 0.90 indicated adequate model fit. d Values ≤.06 indicated good model fit; values ≤.08 indicated adequate model fit. e When comparing models, lower values indicated the better model fit. f Values with p ≤.05 indicated that the simpler model was significantly less satisfactory than the comparatively complex model. g Not positive definite covariance matrix. *p ≤.05. **p ≤.01. P-CPT = Preschool Continuous Performance Test; GNG = Go/No-Go with block number; DCCS = Dimensional Change Card Sort. *p ≤.05. **p ≤.01. ***p ≤.001. 4For exploratory purposes, we conducted a test of invariance comparing younger (< 4 years old; n = 55, M age = 3;8, SD age = 3 months) and older (> 4 years old; n = 74, M age = 4;7, SD age = 5 months) preschoolers. This test indicated strict measurement invariance, χ 2 (84, N = 129) = 87.72, p > .05. In addition, a test of invariance for preschoolers divided into boys (n = 78, M age = 4;3, SD age = 7 months) and girls (n = 51, M age = 4;2s, SD age = 7 months) indicated strict measurement invariance for sex, χ 2 (84, N = 129) = 96.60, p > .05. Additional informationNotes on contributorsMichael R. Miller Michael R. Miller is now at Vanderbilt University, Nashville, TN, USA. Gerald F. Giesbrecht is now at the Behavioral Research Unit, Alberta Children's Hospital, Calgary, AB, Canada. Robert J. McInerney is now at IWK Health Centre, Halifax, NS, Canada.