Do Strong Item-Context Associations Facilitate Hierarchical Rule use in Children?

• In Event: Scaffolding the Development of Successful Learning Across Early and Middle Childhood

Abstract

It is well established that the ability to learn and implement hierarchically organized rules shows a prolonged development trajectory. Previous research has linked hierarchical rule use to the efficiency of working memory output gating mechanisms. Output gating is thought to be supported by stable representations of the higher-order context that guides the selection of relevant lower-order working memory contents, which then drive action selection. Here we asked whether children’s performance on a hierarchically structured task can be improved by scaffolding this rule setting process through strong pre-learned associations between the higher-order context and response-relevant lower-order stimuli.

Forty-six typically developing children between the ages of 7 and 10 years and 41 healthy young adults between 18 and 29 years of age participated in the study. Data were collected online using Labvanced® software. We used a hierarchical task that included sequences of three items—two lower-order stimuli and one higher-order context cue—appearing in random order. The context cue indicated which of the lower-order stimuli were response-relevant. Thus, when the context was presented last, participants had to maintain both preceding stimuli in working memory and only then select the response-relevant stimuli. Note that output gating demands—and hence rule setting demands—were higher when one of the two items had to be selected as compared to globally gating both stimuli. Participants were randomly assigned to one of two versions of the task: In the strong association condition, the context cues (e.g., cat) were associated with the corresponding lower-order items (e.g., fish, milk). In the weak association condition, there were no pre-learned associations between context cues (e.g., morning) and those same lower-order items. We expected rule setting to be facilitated in the strong association condition, resulting in decreased performance costs for selective relative to global output gating. We further hypothesized that children should benefit more from pre-learned context-item associations than adults, resulting in reduced age differences for the strong compared to the weak association condition.

Error rates and RTs were fitted on a single-trial level using (generalized) linear mixed models with the fixed factors age group, gating demand, and association strength as well as random intercepts and random slopes for the participants. There were no overall performance differences between weak and strong association condition (p = .31). For adults, RTs were longer for selective compared to global output gating in the weak association task (p < .001 ) but not in the strong association task (p = .78). A similar pattern was found for error rates. Contrary to our predictions, children’s error rates and RTs differed between selective and global output gating (ps < .001) but were not affected by association strength (ps > .28).

Stronger association between context and lower-order stimuli benefitted output gating performance in adults, as indicated by eliminated RT differences between selective and global output gating. The lack of similar behavioral benefits in children suggests that they might not represent rules in a fully hierarchical manner that “separates” context from lower-order items, causing greater interference between competing WM representations.

Author

• Kerstin Unger, Queens College, CUNY
  Presenting Author