Measuring the development of inhibitory control: The challenge of heterotypic continuity

Abstract

Inhibitory control is thought to demonstrate heterotypic continuity, in other words, continuity in its purpose or function but changes in its behavioral manifestation over time. This creates major methodological challenges for studying the development of inhibitory control in childhood including construct validity, developmental appropriateness and sensitivity of measures, and longitudinal factorial invariance. We meta-analyzed 198 studies using measures of inhibitory control, a key aspect of self-regulation, to estimate age ranges of usefulness for each measure. The inhibitory control measures showed limited age ranges of usefulness owing to ceiling/floor effects. Tasks were useful, on average, for a developmental span of less than 3 years. This suggests that measuring inhibitory control over longer spans of development may require use of different measures at different time points, seeking to measure heterotypic continuity. We suggest ways to study the development of inhibitory control, with overlapping measurement in a structural equation modeling framework and tests of longitudinal factorial or measurement invariance. However, as valuable as this would be for the area, we also point out that establishing longitudinal factorial invariance is neither sufficient nor necessary for examining developmental change. Any study of developmental change should be guided by theory and construct validity, aiming toward a better empirical and theoretical approach to the selection and combination of measures.

Keywords: Changing measures; Executive function; Inhibitory control; Longitudinal factorial invariance; Meta-analysis; Self-regulation.

Fig. 1

Children’s average percentage-of-maximum score on “early” inhibitory control measures by age. Gray lines represent age ranges with weaker empirical support (i.e., fewer than two studies; line is interpolated if there are no studies at a given age). Dashed gray horizontal lines at 20% and 80% reflect our low and high cutoffs, respectively, for defining the useful age range of each task.

Fig. 2

Children’s average percentage-of-maximum score on “middle” inhibitory control measures by age. Gray lines represent age ranges with weaker empirical support (i.e., fewer than two studies; line is interpolated if there are no studies at a given age). Dashed gray horizontal lines at 20% and 80% reflect our low and high cutoffs, respectively, for defining the useful age range of each task.

Fig. 3

Children’s average percentage-of-maximum score on “late” inhibitory control measures by age. Gray lines represent age ranges with weaker empirical support (i.e., fewer than two studies; line is interpolated if there are no studies at a given age). Dashed gray horizontal lines at 20% and 80% reflect our low and high cutoffs, respectively, for defining the useful age range of each task

Fig. 4

Depiction of useful age ranges of inhibitory control measures based on age ranges from Table 1 and Fig. 1. Dotted lines represent age ranges with weaker empirical support (fewer than two studies). CBQ= Children’s Behavior Questionnaire

Fig. 5

Example baseline longitudinal CFA model of inhibitory control adapted from Widaman et al. (2010) for use with changing measures over time. In the example model, the four measures at each time point (rectangles) serve as manifest variables (indicators) of the latent variable reflecting inhibitory control at each time point (ovals). The example baseline model is an abridged CFA with only three time points for pictorial simplicity. A full test of longitudinal factorial invariance would incorporate all nine measurement occasions from Table 6. See Table 5 for a list of the constraints added to the baseline model to test longitudinal factorial invariance. Constraints are only made within measure. For factor loadings (λ), residual variances (θ), and intercepts (τ), the first numerical digit refers to the measure (e.g., 2 from λ₂₃ = Grass/Snow) and the second digit refers to the occasion of measurement (e.g., 3 from λ₂₃ = Time 3: 42 months). For covariances (σ²), the first numerical digit refers to the occasion of measurement from one end of the arrow and the second digit refers to the occasion of measurement of the other arrow. The covariance of a variable with itself represents that variable’s variance. The triangle represents the estimation of means (intercepts), including the means of the latent factors (α) and the intercepts of the indicators (τ). “mo” = child’s age in months, BD = Bear/Dragon, DN = Day/Night, GS = Grass/Snow, RC = Reverse Categorization, SH = Shape Stroop, SC = Spatial Conflict.