Complex executive function over childhood
I’m going to do two entries on executive function. Here, I’m going to summarize an article at length. Next time, I will present multiple specific articles. McGuckian et al. (2023) published “Development of Complex Executive Function Over Childhood: Longitudinal growth curve modeling of performance on the Groton Maze Learning Task” in Child Development. Here’s an edited version of their article (for those of you who hate research design and statistics, I have put in bold critical findings):
This longitudinal study modeled children's complex executive function (EF) development using the Groton Maze Learning Task (GMLT). Using a cohort-sequential design, 147 children (61 males, 5.5–11 years) were recruited from six multicultural primary schools in Melbourne and Perth, Australia. Children were assessed on the GMLT at 6-month intervals over 2-years between 2010 and 2012. Growth curve models describe age-related change from 5.5 to 12.5 years old. Results showed a quadratic growth trajectory on each measure of error—that is, those that reflect visuospatial memory, executive control (or the ability to apply rules for action), and complex EF. The ability to apply rules for action, while a rate-limiting factor in complex EF, develops rapidly over early-to-mid childhood.
Theoretical and empirical models consider executive function (EF) to be hierarchical in structure, consisting of (i) a domain-general module termed cognitive control, which is concerned with coordinating cognitive operations, implementing strategies for problem solving, and monitoring errors and (ii) a sub-set of more specialized (modality-specific) operations including working memory, inhibition, and executive attention. It is widely accepted that EF develops rapidly over childhood and into the adolescent period, in synergy with the heightened demands of learning, schooling, and sociocultural participation. The development of complex EF is, indeed, critical to the acquisition of academic and life skills, predicts academic achievement, and later adaptive function.
Behavioral and cognitive neuroscience studies show that performance of complex visuospatial tasks that involve rule-governed behavior (like maze learning) is dependent on separable cognitive processes that include planning based on prior behavior, feedback from errors, and holding information in memory about the outcomes of prior moves or trials. In the GMLT, performers are instructed to follow a set of rules that govern the type of moves that are allowed in order to learn a hidden maze (within a square grid of tiles) over repeated trials. For the GMLT, the following four rules apply: (i) move one tile at a time; (ii) do not move diagonally; (iii) do not backtrack along the path; and (iv) return to the last correct tile after an error. After each move, visual feedback is temporarily provided on the selected tile (green tick or red cross) to indicate whether the performer has correctly hit on the hidden path or not. If they correctly hit the path, they can make their next selection. If they hit off the path, they are required to return to the last correct tile and select an alternate next tile. A trial is complete when the performer reaches the end of the maze.
A total of 147 typically developing children (61 male) aged between 5.5 and 11 years of age were recruited from six mainstream (public and independent) primary schools in the greater Melbourne and Perth metropolitan areas. The Australian population has ancestries (of at least one parent) of approximately 57% European, 31% Oceanian (including ‘Australian’), 17% Asian, 3% Indigenous, 3% North African and Middle Eastern, 1% Peoples of the Americas, 1% Sub-Saharan African. All children were identified by parents on a developmental questionnaire as being free of any major medical or neurological condition, and none reported an intellectual disability.
Complex EF was assessed using an 8 × 8 version of the GMLT, presented centrally on a 12-inch touchscreen PC. Four primary outcomes were calculated from the GMLT. Rule-break errors was calculated as the total number of times any of the four rules were broken, summed for all five trials. ‘Legal errors’ was calculated as the total number of times a response was in accordance with the rules but to a location that did not lie on the hidden pathway, summed for all five trials. “Total errors’ was calculated as the sum of rule-break errors and legal errors. Response duration was calculated as the total time (in seconds) taken to complete all five trials.
On legal errors, rule-break errors, and total errors, quadratic trends were the best fitting and most parsimonious model in each case. For changes in response duration over childhood, a cubic trend was the best fitting (and more parsimonious) growth function. The pattern of steep decline in rule-break errors between 5.5 and 7 years suggests a period of rapid consolidation of simple action-oriented rules that support learning of a visuospatial navigation task.
A rapid reduction in total response duration was found between 6.5 and 8.5 years followed by further (but more gradual) reduction until 12.5 years. This change in response duration reflects an improvement in performance efficiency and psychomotor speed, in line with (1) the reduction in rule-break and legal error rates, and (2) improvements in fine- and gross-motor capabilities, which facilitate precise and rapid motor actions required to physically interact with the maze quickly.
The quadratic trend that we observed on rule-break errors suggests that the ability to use simple action-oriented rules in a spatial learning task is acquired rapidly over the course of early-to-middle childhood. Children reduced the average number of such errors from 20 at 5.5 years of age, to less than 10 errors by age 7, and to around four errors by 9 years. The ability to apply such rules is thought to be supported by a process of self-reflection and higher-order attentional control. Even use of simple rules, such as those involved in the GMLT, requires a capacity to monitor goal-directed behavior in real-time. In essence, reflection (via sub-vocal verbalization) is thought to enable the child to reference previously learned task rules in order to solve an action problem, while also keeping other features of the task in mind (e.g., accessing the correct path held in working memory. An inability to reflect may also explain the utilization deficiency that occurs with young children, where a gap exists between the ability to learn task rules and call upon them mid-task. The improvement with age may be further aided by the transition from reactive to proactive cognitive control that occurs from 3- to 8-years of age. In the context of the GMLT, it may be that older children are better able to proactively maintain action-oriented rules for future planning of moves, whereas when this cognitive control is enlisted in younger children, rule use is slower and more cognitively demanding, resulting in more rule-break errors and slower overall response time. Taken together, we argue that age-related changes in the ability to enlist simple action-oriented rules through reflection and proactive control, combined with an expanding capacity to buffer visuospatial information in working memory, maintains steady improvement on the GMLT over this critical developmental period.
It follows that performance of visuospatial learning tasks that involve more complex rule structures will follow different growth trajectories to those with simpler rule structures, like the GMLT. First, tasks that involve hierarchical rule structures or more abstract reasoning tend to rely more heavily on lateral regions of the PFC. These higher-order cortical regions develop relatively late, and then exert top-down influence on lower-level (but earlier developing) sensorimotor regions. Second, at a neural level, there is a shift with age and experience from sparse (within network) activation to a more targeted and well-defined grouping of network activation. This changing pattern occurs progressively over childhood and into adolescence. Finally, structural immaturities in the development of EF networks in younger children appear to necessitate a different pattern of neural activation. Put simply, when presented with a complex spatial planning and updating task, younger children appear to enlist networks other than those associated with mature EF—the left parietal cortex is one region where compensatory activation is evident. Taken together, our behavioral modeling of EF abilities suggests that performance of younger children may be supported by these less-efficient compensatory processes that are yet to fully develop. However, higher-level control is rapidly conferred as children enter middle-to-late childhood. With age, the capacity to couple the rule-monitoring functions of the PFC (and cingulate cortex) with lower-level spatial buffering in working memory is likely to underpin advances in complex task performance.
The decline on rule-break errors between 5.5 and 9 years combined with quadratic growth trend over the childhood period suggests rapid acquisition of the ability to enlist simple action-oriented rules in goal-directed behavior (specifically, maze navigation). These behavioral results mirror structural changes in the development of neural networks that underpin working memory and cognitive control, particularly synergies between PFC and cingulate cortex. Steady maturation in such networks supports the ability to integrate the maintenance of visual–spatial information online while solving cognitive tasks. For tasks that involve more complex action-oriented rules (than those examined here), a more protracted period of development may be necessary. The ability to describe these patterns of growth in EF with greater precision has important implications for the design of learning environments for school-aged children; the aim of which should be to present learning tasks that impose constraints on working memory and cognitive control that are scaled appropriately for proximal learning.
What I love about this study is its cross-sequential design, capturing change over a 2-year period from different cohorts. It also defines executive functioning skills, highlights the ages during which we should see most rapid changes, and has important implications for understanding the events developmentally that may adversely impact the development of executive functions.