When less is more: costs and benefits of varied vs. fixed content and structure in short-term task switching training

Abstract

Training variability has been brought forward as one possible moderator for wider scale transfer effects in cognitive training. However, little is known about which aspects of task variability are important for optimizing training outcomes. This study systematically examined the impact of variability in the different task components on outcome measures, here manipulating content (whether the task stimuli remained fixed or changed between blocks) and the deeper structural task configuration (task sequence: whether the task sequence was fixed or random). Short-term task switching training was implemented with one of four training variability conditions: fixed content\fixed structure; fixed content\ random structure; varied content\fixed structure and varied content\varied structure. The experiment consisted of a baseline block, seven training blocks (learning phase), followed by two transfer blocks, one with fixed and one with random task structure, respectively. In the learning phase, more rapid training gains were observed in the fixed content as compared to varied content. Interestingly, training with fixed content resulted in a trend for costs when transferred to a novel task switching context. In contrast, moderate transfer gains were noted in the varied content condition, manifested specifically on switch trials. These results suggest that task (content) variability is one of the means to improve positive transfer and avoid negative transfer. Additionally, and in agreement with the wide literature on training, this finding suggests that conditions that prevent training gains are in fact beneficial for learning generalization.

On How to Be Flexible (or Not): Modulation of the Stability-Flexibility Balance

“…In a recent study, participants had to switch between the same two tasks across several blocks. Even though it showed a steep learning curve, substantial costs emerged when participants were confronted with two new tasks in the end (Sabah, Dolk, Meiran, & Dreisbach, 2018). We therefore assume that frequent forced switching motivates participants to keep the respective tasks in a highly active state (for related evidence, see Schneider, 2015).…”

Abstract sequential task control is facilitated by practice and embedded motor sequences

“…The simplest explanation for this lack of effect is that task switching improvements were general across the conditions and not specific to practiced abstract task sequences, such that task switching improved in Novel sequences as well. This idea is supported by the finding that task switching practice effects generalize to cognitive control tasks that involve similar processes (Sabah et al, 2019), and suggests that training benefits are not strictly domain specific.…”

“…Switch costs in non-sequential tasks are reduced, but not eliminated, with practice (Berryhill & Hughes, 2009;Stoet & Snyder, 2007;Strobach et al, 2012). Reductions in switch costs are hypothesized to reflect improvements in shifting attention to a new task set, inhibiting the irrelevant task set, or retrieving a new goal state (Hirsch et al, 2018;Sabah et al, 2019). When task switching is incorporated into an abstract task sequence, we assume that these same processes will still be optimized.…”

“…Due to extensive work showing that humans can improve at other types of sequences (i.e., motor sequences) with practice (Nissen & Bullemer, 1987;Robertson, 2007;Verwey et al, 2010), and evidence that behavioral manifestations of control costs associated with the execution of hierarchically structured tasks are reduced with practice (Berryhill & Hughes, 2009;Sabah et al, 2019;Strobach et al, 2012), we hypothesized that performance of the abstract task sequences would improve with practice. Specifically, we hypothesized that improvements in performance would be specific to the first position in the sequence, and the initiation costs that are indicative of the cognitive control engaged to perform abstract task sequences would be reduced.…”

“…Specifically, we hypothesized that improvements in performance would be specific to the first position in the sequence, and the initiation costs that are indicative of the cognitive control engaged to perform abstract task sequences would be reduced. This hypothesis is based on the observation that other control costs, such as task switching, are reduced, but not eliminated with practice (Berryhill & Hughes, 2009;Sabah et al, 2019;Strobach et al, 2012). Alternatively, performance improvements that were consistent across the abstract task sequences would indicate a potential strengthening of the associations between the sequence and task level associations at each serial position in the sequence such that performance was facilitated throughout.…”