The Development of Relational Reasoning: An Eyetracking Analysis of Strategy Use and Adaptation in Children and Adults Performing Matrix Completion

Abstract

Relational reasoning is a key component of fluid intelligence and important predictor of academic achievement. Relational reasoning is commonly assessed using matrix completion tasks, in which participants see an incomplete matrix of items that vary on different dimensions and must select an item that best completes the matrix based on the relations among items. Performance on such assessments increases dramatically across childhood into adulthood. However, despite widespread use, little is known about the strategies associated with good or poor matrix completion performance in childhood. This study examined the strategies children and adults use to solve matrix completion problems, how those strategies change with age, and whether children and adults adapt strategies to difficulty. We investigated 6- and 9-year-old children and adults performing matrix completion while undergoing eyetracking to infer strategy use. Across all ages, scanning across matrix rows and columns predicted good performance, and higher rates of consulting potential answers predicted poor performance, indicating that optimal matrix completion strategies are similar across development. Indices of good strategy use increased across childhood. As problems increased in difficulty, children and adults increased their scanning of matrix rows and columns, and adults and 9-year-olds also qualitatively shifted strategies to rely more on consulting potential answers. Adapting strategies to matrix difficulty, particularly increased scanning of rows and columns, was associated with good overall performance in both children and adults. These findings underscore the importance of both spontaneous and adaptive strategy use in individual differences in relational reasoning and its development.

Development of reasoning performance in Raven’s matrices is grounded in the development of effective strategy use.

“…Overall, the proportion of time spent looking at the matrix and the proportion of time before first toggle showed the expected increase with age, and the expected relations with other measures. This is in line with data from analogy tasks suggesting that young children GONTHIER, HARMA, AND GAVORNIKOVA-BALIGAND 10 pay less attention to logical rules (Starr et al, 2018;Thibaut & French, 2016) and are more affected by distracting response options (Guarino et al, 2022;Thibaut et al, 2010aThibaut et al, , 2010b, and with data from matrix tasks suggesting that young children tend less to engage in structured visual exploration of the matrix supporting rule inference (Chen et al, 2016;Niebaum & Munakata, 2021). Taken together, and leaving aside the results for toggle rate, these data provide unambiguous evidence that the use of constructive matching increases with age.…”

“…Comparatively lower effort in young children to look back at the matrix to confirm a solution could be reinforced by our paradigm: with only half of the item displayed at one time, participants had to make a purposeful effort to move the mouse and click to toggle between the matrix and responses. As in our prior study using the same paradigm with adults (Rivollier et al, 2021), this led to considerably lower number of toggles and toggle rate on average than in comparable studies using eye-tracking (e.g., Chen et al, 2016;Niebaum & Munakata, 2021). Likewise, having only half the item displayed at a given time may have limited quick lookbacks and placed more DEVELOPMENT OF REASONING STRATEGIES 13 constraints on working memory to keep elements of the matrix in mind while looking at the responses, and vice versa.…”

“…On the other hand, older children did not spend more time on the items, and they made more toggles between matrix and responses, making the pattern ambiguous; more usual indices of strategy use were not reported in this study. The preprint study of Niebaum and Munakata (2021) found converging results in a group of 81 children, with more structured visual exploration and a lower toggle rate in 9-year-olds than in 6-year-olds; but 6-year-olds were close to chance performance and indices of strategy use were mostly unrelated to performance in the group of 9-yearolds, raising the question of whether strategy use can be adequately measured in these age groups.…”

“…In the same way that various abilities may contribute differently to different versions of the task depending on their requirements, it may also be the case that various abilities contribute differently at different ages (e.g., see Demetriou et al, 2014). In particular, the use of constructive matching may be more constrained by the basic ability to perform the task in the youngest children: indeed, children around 5-6 years old tend to perform around chance in complex matrix tasks (Chen et al, 2016;Niebaum & Munakata, 2021), presumably because of difficulty mentally representing relations between elements of the matrix (see Chen et al, 2016). Conversely, use of constructive matching for older children may depend more on ability to infer the rules, metacognition, or yet other abilities.…”

Cognitive segmentation and fluid reasoning in childhood

“…Response elimination involves the comparison of features between the matrix and response alternatives first, in order to eliminate options and select the correct solution. Children who use constructive matching perform better than those who use response elimination, and some individuals shift strategy to response elimination for more complex items (Chen et al, 2016;Niebaum & Munakata, 2021). This suggests that children shift from relying on response elimination strategies to increasingly using constructive matching.…”