Brain activity reveals multiple motor-learning mechanisms in a real-world task

Abstract

21Many recent studies found signatures of motor learning in neural beta oscillations (13)(14)(15)(16)(17)(18)(19)(20)(21)(22) 30Hz), and specifically in the post-movement beta rebound (PMBR). All these studies were in 23 simplified laboratory-tasks in which learning was either error-based or reward-based. Interestingly, 24 these studies reported opposing dynamics of the PMBR magnitude over learning for the error-based 25 and reward-based tasks (increase verses decrease, respectively). Here we explored the PMBR 26 dynamics during real-world motor-skill-learning in a billiards task using mobile-brain-imaging. 27Our EEG recordings highlight opposing dynamics of PMBR magnitudes between different subjects 28 performing the same task. The groups of subjects, defined by their neural-dynamics, also showed 29 behavioral differences expected for error-based verses reward-based learning. Our results suggest 30 that when faced with the complexity of the real-world different subjects might use different learning 31 mechanisms for the same complex task. We speculate that all subjects combine multi-modal 32 mechanisms of learning, but different subjects have different predominant learning mechanisms. 33 recorded with the DSI-24 there were 6 PMBR Increasers and 4 PMBR Decreasers. 126Correspondingly, there was no correlation between the system and the PMBR-Error correlation 127 (Spearman rank correlation r=0.01, p=0.97). 128Based on the EEG data, which suggests two groups of subjects with different PMBR 129 dynamics, we looked for behavioral signatures in the task performance of different learning 130 between these groups. The lag-1 autocorrelation (ACF(1)) of the performance measure (in our case, 131 the directional error of the target-ball relative to the pocket) was suggested as an index of 132 performance in motor-skill learning, where close to zero values corresponds to high skill (van Beers 133 et al., 2013). We calculated for each subject the ACF(1) in the first and the second half of the 134 training session (blocks 1-6 and 7-12, respectively). The ACF(1) values of both groups were 135 significantly greater than zero during both halves of the session (t test p<0.01), as expected for 136 naïve participance ( Figure 3A). But, the decay in the ACF(1) from the first half of the training 137 session to the second was significantly higher for the PMBR Decreasers (t test p<0.01, Figure 3B). 138Additionally, the decay in the intertrial variability (measured from the first block (trials 1-25) to 139 the learning plateau (trials 201-300)) was also significantly larger in the PMBR Decreasers (t test

Embodied virtual reality for the study of real-world motor learning

“…The decay of intertrial variability over trials is a prominent feature of skill learning [64][65][66][67][68][69][70][71], but not found in motor adaptation experiments. In the physical pool paradigm, we found two types of subjects that differ in their intertrial variability decay as well as other behavioural and neural markers [22]. Altogether, this suggested the contribution of two different learning mechanisms to the task: error-based adaptation and reward-based reinforcement learning, where the predominant learning mechanism is different between subjects.…”

“…The movement of the real balls on the pool table were tracked with a computer vision system mounted from the ceiling (Genie Nano C1280 Color Camera, Teledyne Dalsa, Waterloo, Canada), with a resolution of 752x444 pixels and a frequency of 200Hz. Image videos were recorded and analysed with our custom software written for the real-world paradigm [21,22].…”

“…In contrast to lab-based tasks, real-world neuroscience approaches study neurobehavioral processes in natural behavioural settings [17][18][19][20]. We recently presented a naturalistic real-world motor learning paradigm, using wearables for full-body motion tracking and EEG for mobile brain imaging, while making people perform actual real-world tasks, such as playing the competitive sport of pool-table billiards [21,22]. We showed that motor learning is a full-body process that involves multiple learning mechanisms, and different subjects might prefer one over the other.…”

Altered Dynamics of Cortical Beta-Oscillations during Motor Learning in Cerebellar Ataxia

“…Understanding these interactions, and in particular their relevance to the pathophysiology of cerebellar ataxia, likely requires a network perspective that includes extracerebellar regions, in particular cerebral cortex. While previous studies have established links between motor learning and cortical beta-oscillations (Alayrangues et al, 2019;Darch, Cerminara, Gilchrist, & Apps, 2020;Espenhahn et al, 2019;Haar & Faisal, 2020;Jahani, Schwey, Bernier, & Malfait, 2020;Palmer et al, 2019;Pollok, Latz, Krause, Butz, & Schnitzler, 2014;Tan et al, 2016;Tan et al, 2014;Torrecillos et al, 2015), the involvement of cortical beta-oscillations in specific learning mechanisms has remained largely unclear, leave alone their relevance to motor learning dysfunction in cerebellar ataxia.…”

Playing the piano with a robotic third thumb: Assessing constraints of human augmentation

“…Haar et al, 2019) and the learning mechanism can differ between users (e.g. Haar and Faisal, 2020). In the case of augmentation technology, the relevant features can be either those related to performing the task itself without the augmented device, or features related to the control interface of the augmented device.…”