Lower Attentional Skills predict increased exploratory foraging patterns

Abstract

When engaged in a search task, one needs to arbitrate between exploring and exploiting the environment to optimize the outcome. Many intrinsic, task and environmental factors are known to influence the exploration/exploitation balance. Here, in a non clinical population, we show that the level of inattention (assessed as a trait) is one such factor: children with higher scores on an ADHD (Attention Deficit/Hyperactivity Disorder) questionnaire exhibited longer transitions between consecutively retrieved items, in both a visual and a semantic search task. These more frequent exploration behaviours were associated with differential performance patterns: children with higher levels of ADHD traits performed better in semantic search, while their performance was unaffected in visual search. Our results contribute to the growing literature suggesting that ADHD should not be simply conceived as a pure deficit of attention, but also as a specific cognitive strategy that may prove beneficial in some contexts.

Figure 1

Illustration of search paths in visual (top, with the stimulus sheet overlaid – Bells test by Gauthier et al. 1989) and semantic (bottom) spaces for two participants, representative of diverse patterns of Inattentive/hyperactive behaviours, one with high (47, left, red) and the other with low (0, right, blue) rating on the ADHD-rating scale. For illustrative purposes only, we here represent the semantic space by means of the t-distributed stochastic neighbor embedding (t-SNE, tsne package in R) algorithm for dimensionality reduction, in order to obtain a two dimensions map of the 300 dimensions vectors from the word2vec google.news model. Insets contain the histograms for the distributions of distances.

Figure 2

Distribution of distances between two consecutive targets in the visual search task (A) and in the semantic search task (B), for children scoring high (pain lines) and low (dotted lines) on the ADHD rating scale (ADHD-rs), according to a median split (m = 9) on the scores. The horizontal bars correspond to the range of a significant difference in the density distributions, assessed by means of bootstrap with a cluster-wise significance level of 0.05–from 0.56 to 0.99, for the visual search task and from 0.87 to 0.99 for the semantic search task. On the semantic search task, the distribution of similarities shows two minor peaks: the first one at 0 corresponds to the immediate repetitions (ex: dog-dog). The second one correspond to the “dog-cat” pairs with semantic similarity of 0.29.