Examining the Modulating Effects of Reinforcement Learning and Explicit Instructions on Cognitive Stability and Flexibility

Context-sensitive adaptations of cognitive stability and flexibility, two key functions of cognitive control, are a prerequisite for goal-directed behavior (Dreisbach & Fröber, 2019; Dreisbach & Wenke, 2011; Egner, 2023; Goschke, 2003). Cognitive stability and flexibility are assumed to be associated with complementary benefits and costs in performance (Dreisbach & Fröber, 2019; Goschke, 2013; but see e.g., Egner, 2023). The flexibility to switch between tasks is traded off against the cost of distractibility (Dreisbach et al., 2005; Tharp & Pickering, 2011), while a stable task focus comes at the concurrent cost of rigidity (Hefer & Dreisbach, 2017). In the planned study, we aim to investigate potential modulators of these cognitive control states. Being able to reliably modulate or induce meta-control processes that promote stability and flexibility is not only of theoretical relevance but may potentially enable us to predict, reduce, or even optimize the costs of control configurations in various contexts. Various studies have demonstrated a modulating influence of reward on mobilization of cognitive control as well as on the mode of cognitive control (Dreisbach & Fischer, 2012; Dreisbach & Fröber, 2019; Goschke & Bolte, 2014). Extending previous research on reward effects, some studies have further demonstrated that not only specific behaviors but also higher-order control processes can be modulated via reinforcement learning by linking reward to varying abstract control demands. For example, disproportionally reinforcing task switches compared to repetitions led to more voluntary switching behavior and smaller interference from response-incongruent trials (Braem, 2017). Making reward more contingent on incongruent than congruent trials led to reduced congruency effects (Held et al., 2024). Reinforcement learning thus operates on the level of abstract meta-control processes, modulating future control states. Building on these findings, in the planned study, we will investigate whether selectively rewarding cognitively stable or flexible behavior increases goal shielding or shifting respectively. Control states are suggested to be subject to deliberate meta-control strategies. It has been shown that individuals adjust their level of cognitive flexibility based on information about to-be-expected control demands (Liu & Yeung, 2020). Previous studies have further reported instructed adjustments in participants’ degree of parallel processing (Lehle et al., 2009; Lehle & Hübner, 2009). Similarly, we observed corresponding adaptations in participants’ cognitive flexibility in terms of altered task-switch costs when we explicitly instructed them to switch flexibly between tasks compared to when we instructed them to shield their current task against distractions (Bartossek et al., 2024; unpublished manuscript). Because evidence for instruction-based adjustments of meta-control strategies is scarce, another aim of the present study is to replicate our previous findings and test if cognitive control modes can be regulated strategically in a top-down manner. Furthermore, a current debate revolves around the mechanisms underlying stability and flexibility. While control-dilemma theory suggests a stability-flexibility tradeoff (Dreisbach & Fröber, 2019; Goschke, 2003, 2013; Goschke & Bolte, 2014), recent studies suggest that stability and flexibility can be independently regulated (Geddert & Egner, 2022; Bejjani et al., 2022; Egner, 2023; Nack & Yu-Chin, 2023; Geddert & Egner, 2024). The present study thus additionally aims to test the hypothesis of an antagonistic relationship between cognitive stability and flexibility. To these ends, we will conduct two online task-switching experiments. In Experiment (Exp.) 1, we will selectively reward participants for either stable or flexible behavior throughout experimental blocks. Thereby, we aim to investigate whether abstract control states can be induced via reinforcement learning. In Exp. 2, we will present strategy instructions targeted at promoting either increased cognitive stability or flexibility. By testing whether these explicit instructions will induce corresponding strategic adjustments of control states, we aim to replicate our previous finding of volitional control regulations. Additionally, fluctuations in cognitive stability and flexibility might arise independently of the manipulations in Exp. 1 and 2 at individual subject levels. As stability and flexibility are hypothesized to involve a tradeoff (Dreisbach & Fröber, 2019; Goschke, 2003, 2013; Goschke & Bolte, 2014), we will test if individual adjustments in cognitive stability between conditions negatively correlate with concurrent changes in cognitive flexibility in both experiments.