Instruction-based learning (IBL) is an essential human skill, enabling flexible and efficient application of novel rules. In most studies, instructions specify affirmative (or positive) stimulus-response (S-R) rules (i.e., if condition A, then execute action X). In this talk, however, I will focus on so-called negative instructions (i.e., if condition A, then do not execute action X but execute alternative action Y or Z) and present some insights from recent behavioral and functional magnetic resonance imaging (fMRI) studies.

Behaviorally, relative to ad-hoc negation of previously established neutral S–R associations, actual negative instructions seemed to induce proactive preparatory processes, resulting in greater implementation efficiency. However, when compared to standard positive instructions, rule implementation still came with considerable costs. I argue that this is mainly due to the interplay of two dissociable representations in negative instructions: The originally negated, instruction-related S-R association (i.e., A-X) persistently influences instruction implementation while an implementation-related S-R association (i.e., A-Y or A-Z) gains increasingly stable control across repeated implementation trials.

At the neural level, we attempted to capitalize on this property by applying multivariate pattern analysis sensitive to individual rule identities. Irrespective of instruction type, rule identities were represented in a set of lateral prefrontal and parietal cortical areas. However, there was no evidence for a dynamic representational change in any of these regions.

I will discuss the implications of these findings and how further examination of negative instructions could advance the general understanding of initial instruction-induced learning processes.