Learning when to learn: Context-specific instruction encoding

Humans are efficient at learning new behaviors through verbal instructions, bypassing the need for trial-and-error learning, but requiring a certain degree of cognitive flexibility. Recent theories suggest that abstract functions of cognition, like cognitive flexibility, can be (up)regulated by contextual features in our environment with which they are associated. Here, we tested whether people show a more flexible control mode in contexts where one frequently learns new task instructions. Specifically, participants had to encode and execute a series of new task instructions, and the likelihood of encountering a new (relative to old) task instruction varied in function of whether the instruction was presented above versus below the center of the screen (Experiment 1, n = 40), or in a blue versus green font (Experiment 2, n = 40). We measured the level of interference from retained new task instructions on secondary task performance to index cognitive flexibility, and show enhanced flexibility (i.e., enhanced receptibility for new task instructions) for contexts in which the likelihood of encountering new task instructions was relatively high. Similarly, participants were also faster to encode new instructions in these contexts. A third, online study (n = 77) replicated the main finding of Experiment 2, whose context manipulation (i.e., font color) was most effective. Together, these findings suggest that cognitive flexibility can bind to, and be signaled by, contextual features in our environment. Such ‘learning when to learn’ provides a (self-)regulating mechanism for the cognitive processes that support the encoding (and execution) of new task instructions.