Frontoparietal action-oriented codes support novel instruction implementation

Abstract

12A key aspect of human cognitive flexibility concerns the ability to rapidly convert 13 complex symbolic instructions into novel behaviors. Previous research proposes 14 that this fast configuration is supported by two differentiated neurocognitive states, 15 namely, an initial declarative maintenance of task knowledge, and a progressive 16 transformation into a pragmatic, action-oriented state necessary for optimal task 17 execution. Furthermore, current models predict a crucial role of frontal and parietal 18 brain regions in this transformation. However, direct evidence for such 19 frontoparietal formatting of novel task representations is still lacking. Here, we 20 report the results of an fMRI experiment in which participants had to execute novel 21 instructed stimulus-response associations. We then used a multivariate pattern-22 tracking procedure to quantify the degree of neural activation of instructions in 23 declarative and procedural representational formats. This analysis revealed, for the 24 first time, format-unique representations of relevant task sets in frontoparietal 25 areas, prior to execution. Critically, the degree of procedural (but not declarative) 26 activation predicted subsequent behavioral performance. Our results shed light on 27 current debates on the architecture of cognitive control and working memory 28 systems, suggesting a contribution of frontoparietal regions to output gating 29 mechanisms that drive behavior. 30 31 32Some of the most advanced collaborative human achievements rely on our ability 33 to rapidly learn novel tasks. Instruction following constitutes a powerful instance of 34 this ability as it combines the flexibility to specify complex abstract relationships 35 with an efficiency far superior to other forms of task learning such as trial and error, 36 or reinforcement learning. These unique characteristics make it a distinctive skill 37 that separates humans from other species 1 . While recent years have witnessed 38 substantial progress in our understanding of instruction following, the neural and 39 cognitive mechanisms underlying this rapid transformation of complex symbolic 40 information into effective behavior are still poorly understood. Specifically, a critical 41 question that remains unresolved is whether a declarative representation of task 42 information is sufficient or whether an additional representational state, closely 43 linked to action, precedes optimal performance. 44Previous behavioral studies have consistently reported an intriguing signature of 45 instruction processing, namely, a reflexive activation of responses on the basis of 46 merely instructed stimulus-response (S-R) associations (defined as "intention-47 based reflexivity", or IBR). IBR occurs even when instructions are task-irrelevant 48 and have not been overtly executed before 2-7 , which suggests a rapid 49 configuration of instructed content predominantly towards action. Instruction 50 implementation also has a profound impact on brain activity, as shown by 5...

Neural oscillations track the maintenance and proceduralization of novel instructions

“…This might be due to the response set used in the specific task, or to other idiosyncratic factors. Despite these inconsistencies in the behavioral measures, the dissociation between Implementation and Memorization is well established at the neural level (Demanet et al, 2016;González-García et al, 2019;Muhle-Karbe et al, 2017).…”

“…A growing number of fMRI studies has primarily focused on revealing which brain regions support the implementation of novel task sets (Demanet et al, 2016;González-García, Arco, Palenciano, Ramírez, & Ruz, 2017;Hartstra, Kühn, Verguts, & Brass, 2011;Hartstra, Waszak, & Brass, 2012;Ruge & Wolfensteller, 2010) and, more recently, on characterizing implementation-specific neural representations (Bourguignon, Braem, Hartstra, De Houwer, & Brass, 2018; González-García, Formica, Wisniewski, & Brass, 2019;Muhle-Karbe, Duncan, De Baene, Mitchell, & Brass, 2017;Palenciano, González-García, Arco, Pessoa, & Ruz, 2019;Ruge, Schäfer, Zwosta, Mohr, & Wolfensteller, 2019). These results consistently point towards a crucial role of frontoparietal regions, and in particular of the prefrontal cortex (PFC).…”