Over the last decades, many researchers have investigated sequential motor skills. In order to do so, I many years ago developed the Discrete Sequence Production (DSP) task (Verwey, 1999) and more recently proposed the Cognitive framework for Sequential Motor Behavior (C-SMB; Verwey et al., 2015). Since the publication of C-SMB in 2015 various labs have reported new findings using the DSP task. In this presentation I will present a review of the main new findings and propose an update, C-SMB 2.0. I will end my talk speculating about the neural substrate of serial motor skills and outline a neural version, N-SMB.

Introduction: Older adults report that restoration (or maintenance) of mobility is among their chief concerns. A key component of mobility is the ability to adapt locomotion to changing environmental demands. A split-belt treadmill paradigm, in which a treadmill belt under each foot moves at differing speeds, has been used to study adaptation of spatial and temporal gait parameters. Historically, studies using this task have focused on anterior-posterior (AP) spatiotemporal gait parameters because this paradigm is primarily a perturbation in the AP direction, but it is important to also understand whether and how medial-lateral (ML) control adapts in this scenario. The ML control of balance must be actively controlled and adapted in different walking environments.

Methods: Younger and older adults adapted walking to a split-belt treadmill paradigm. We computed step length, step width, and several ML balance parameters. We also acquired a structural (T1) MRI scan and a diffusion weighted MRI scan for each participant.

Results: We found that younger adults showed sustained, asymmetric changes in ML balance parameters during split-belt walking whereas older adults did not. This suggests that younger adults are exploiting passive dynamics during the task. In order to better understand this asymmetric adaptation, we examined whether individual differences in ML balance parameters were correlated with regional brain gray matter volumes, cortical thickness, or white matter diffusion indices. We found that, regardless of age, more asymmetrical adaptation of ML balance parameters was associated with greater cerebellar gray matter volume, more gyrification in the pre and post central gyri, and more fractional anisotropy in the corticospinal tract. This is particularly interesting given that our recent work and that of others has shown that the largest age effects on gray matter volume and cortical thickness occur in the primary motor and somatosensory cortices and the cerebellum.

Conclusions: Younger adults adapt to a split-belt walking paradigm via asymmetric changes in ML balance parameters, whereas older adults show less asymmetry. Both younger and older adults show associations between brain structural metrics and asymmetric adaptation, suggesting that older adults who maintain brain structure the most exhibit adaptive changes that are more similar to those of younger adults.

Since the seminal findings of Patient H.M., sensorimotor learning has been thought to reflect the outcome of a unitary implicit process. More recently, it has become clear that explicit, cognitive strategies play an important role in sensorimotor learning; however, they remain poorly understood. Here, I will discuss our recent efforts to pin down the computational and neural underpinnings of these strategies. We find that strategies account for the lion’s share of performance improvements in sensorimotor learning tasks. They appear to rely on various processes associated with executive function, such as mental imagery and simulation, which are constrained by working memory capacity and computational processing limitations. What’s more, the process that we thought of as being emblematic of implicit learning, which gives rise to sensorimotor aftereffects, appears incapable of explaining sensorimotor learning even after long periods of training. As a consequence, we hypothesize that the strategies themselves are likely to become proceduralized over time or serve as the input to another implicit process, such as model-free reinforcement learning. This work suggests that we may need to rethink current theories of motor control and broaden the scope of the putative neural substrates thought to underlie sensorimotor skill acquisition to accommodate the important contribution of cognitive strategies.

Memory consolidation is central to the transformation of new knowledge about skill from a labile state to a more stable one. The importance of consolidation is highlighted by its recognition as one of four pillars of learning (Daheane, 2020). In the case of motor skill acquisition, successful consolidation is often manifest as preservation of a motor skill, even in the face of interference, or as a gain in skilled performance across significant time periods despite no additional motor practice. A brief review of my early behavioral and more recent neurophysiological efforts addressing motor adaptation and motor sequence learning will attempt to: (a) describe how post-practice consolidation can be mediated to optimize learning, and (b) to delineate some of the key neural players central to consolidation of procedural skill.

Previous theoretical accounts of attention in motor skill learning have largely concentrated on task-specific processes. As a result, there is little understanding of how momentary and dispositional influences on practice-related attention states are mediated by task-extrinsic sources. To progress towards more comprehensive address of task-extrinsic and intrinsic attention interactions for motor learning, an attention mode transfer conceptual framework is introduced. This framework draws on motor learning theory describing distinct forms of motor behavior driven by stimulus-oriented and sequence-oriented execution modes. From cognitive control theory, the framework incorporates the metacontrol state concept, which describes persistence versus flexible control policies that establish distinct attention modes for goal-oriented behavior. Cognitive and neural correlates of attention control are derived from research on attention training techniques collectively referred to as mindfulness meditation. This includes converging evidence that attention modes established by mindfulness states remain relatively inert to proactively influence subsequent goal-oriented behavior. The present conceptual framework proposes that modes of attention established by mindfulness techniques transfer into subsequent skill practice to regulate the motor execution mode implemented during learning. The attention mode transfer framework explains findings from a series of studies investigating short-term effects of mindfulness meditation on motor sequence learning based on the serial reaction time (SRT) task paradigm. For example, in meditation naïve adults, single-session focused attention (FA) mindfulness meditation enhances SRT performance improvement relative to control conditions. These improvements have been associated with reliance on stimulus-oriented execution processes even though sequence-oriented processing is available based on implicitly embedded sequences in the SRT. A propensity for stimulus-oriented responding is consistent with heightened selective attention, which has transferred from a state of increased cognitive control established by the preceding FA. Increased cognitive control in the SRT following FA has been confirmed based on heightened N2 amplitude using EEG event-related potential methodology. Preceding the SRT with single-session open monitoring (OM) mindfulness meditation, which is associated with weakened cognitive control and less selective attention modes, also enhances performance improvements. In this case though, attention modes transferred from OM engendered sequence-oriented execution. However, this transfer only occurred in individuals reporting low effort levels for completing OM, a more advanced mindfulness technique than FA. This latter finding suggests that the degree of attention mode transfer from mindfulness states to motor learning might depend on the level of proficiency with the mindfulness technique. Furthermore, research on mindfulness state influences on stimulus versus sequence-oriented performance of sport skills suggests that mindfulness attention modes might not transfer to or can interfere with established attention modes in well-learned skills.

At a global level, elder adults aged >65 make up more than 50% of the population across developed countries and management of their health and wellbeing is an increasing societal issue. To keep an active quality of life, the ability to learn and retain motor functions becomes of key importance. I consider a generic approach for learning and managing motor functions less effective for wide-ranging differences in motor capacities of the ageing population. Instead, hypothesise an increased benefit from an individualized learning approach. To support this claim, I outline neurobiologically plausible mechanisms to formulate predictions aimed at this approach, coined as Individualised COgnitive and Motor learning for the Elderly (ICOME). Particularly important is to prime cognitive control by inducing optimal learning states with cognitive training. Combining these with the monitoring of sensorimotor rhythms of synchronisation and desynchronisation during motor sequence learning - a framework is formed to guide modelling Ongoing work performed at the Faculty of Behavioural, Management and Social Sciences Lab in University of Twente aimed to tackle this problem will also be covered.

The execution of a sequential motor task is often characterized by the emergence of hierarchical structures, commonly referred to as motor chunks, which facilitate the accurate execution of sequences at increasing speeds. Comparing motor skill acquisition in more than 100 healthy adults, we found that optimal motor skill acquisition relies on prioritizing the optimization of the accuracy over the speed, which enables the generation of efficient chunking patterns at the early stages of training. Finding this process to be diminished in older adults, we applied noninvasive brain stimulation in an attempt to support the aging brain to compensate for these deficits. Our results showed that anodal transcranial direct current stimulation applied over the motor cortex restored the mechanisms involved in the storage of spatial features, without directly affecting the speed of execution of the sequence. This led older adults to sharply improve their accuracy at the early stages of training, resulting in an accelerated emergence of motor chunks. Therefore, our results suggest atDCS can partially restore motor skill acquisition in individuals with diminished learning capabilities by aiding the integration and storage of task-relevant information upstream of the primary motor cortex in the motor network, without directly influencing the motor execution of the sequence.

Aging is associated with alterations in brain activation as well as in structural and functional brain connectivity that have consequences for motor behavior. Here, I will address the effects of aging on motor learning capacity and discuss the associated differences in brain function between young and older adults during the initial and final stages of motor training. I will reflect on over-activation in the aging brain during planning and execution of bimanual skills and discuss the changes in brain activity as a result of practice. The obtained results will be discussed from the perspective of the compensation versus dedifferentiation hypothesis. Compensation refers to altered brain activation patterns in older adults with positive consequences for motor behavior. Dedifferentiation refers to a decrease in the distinctiveness of neural representations, possibly reflecting reduced inhibitory control mechanisms in aging adults. In the first part, I will provide evidence for retained training-induced neuroplasticity in older adults during bimanual skill learning and will characterize the alterations in brain activation associated with aging. In the second part, I will discuss learning capacity in older adults in relation to the structural integrity of the brain with a specific focus on grey matter integrity of subcortical structures. In the third part, I will present a global and local network view on age-related functional connectivity alterations that appear to be consistent with the notion of brain dedifferentiation. On one hand, functional brain connectivity within the resting-state ‘’motor’’ network (intra-network connectivity) is increased in older age and negatively related to performance on bimanual coordination tasks. On the other hand, increased functional connectivity between the different networks (inter-network connectivity) as a function of aging is negatively related to motor performance. The latter findings suggest that reduced segregation (increased inter-network connectivity) in the older brain is associated with decreased levels of motor performance. Interestingly, practice of motor skills in older adults may lead to increased segregation as shown by a decrease of activity in areas constituting the default mode network. This is positively related to motor retention. Finally, I will elaborate on ongoing research about the role of neurochemicals in motor learning. In summary, our work makes a case for the benefits of a complementary multimodal imaging approach to investigate age-related changes in the interactions between brain and motor behavior.

Key words: bimanual movement, motor neuroscience, brain imaging, skill learning

Introduction: Studying people with clinical conditions in which sensorimotor and cognitive processes are affected may allow for a better understanding of the involvement of these processes in motor learning and control. Here, I will address insights from studies on motor abilities in Parkinson’s and Alzheimer’s disease. While Parkinson’s disease is primarily associated with motor symptoms, people also experience non-motor symptoms including impairments in cognition. Conversely, while Alzheimer’s disease is primarily associated with cognitive symptoms, studies have demonstrated that individuals may also experience sensorimotor impairments. I will discuss how studying motor learning and control from a clinical perspective can elucidate the underlying cognitive and neurobiological mechanisms. I will also address to what extent motor performance abilities may have clinical utility, by evaluating whether they can serve as an early disease marker.

Methods: I will present a series of experiments in which we use a combination of pharmacological, behavioral, genetic, and neuroimaging approaches to study motor sequence learning and motor control in people with neurodegenerative disorders. First, we examined the effects of dopaminergic medication on motor learning and control in people with Parkinson’s disease. Second, we evaluated differences in motor learning in individuals with mild cognitive impairment (MCI) and Alzheimer’s disease compared to cognitively normal individuals. We also examined associations between motor learning abilities and existing Alzheimer’s biomarkers.

Results: We found that dopaminergic medication selectively affected sensorimotor and cognitive processes underlying learning and control in people with Parkinson’s disease. While motor responses of individuals with MCI and Alzheimer’s were generally slower than those of controls, learning occurred regardless of cognitive status and did not significantly differ among the three groups. Results of correlational analyses showed no significant associations between motor learning scores and Alzheimer’s biomarkers.

Conclusions: These findings show that dopamine can affect specific cognitive processes underlying motor learning and control in Parkinson’s disease. I will discuss the effects in light of the dopamine overdose hypothesis, which suggests that processes relying on dorsal versus ventral striatal circuitries may be differently sensitive to dopamine modulations. Furthermore, the relatively preserved motor learning abilities in MCI and Alzheimer’s disease suggests that the cognitive impairments seen in these conditions may not concern motor control processes, which could have implications for the design of rehabilitation programs. However, the observed independence of existing biomarkers indicates that sequence learning may not be suitable as an early disease marker.

Key words; motor sequence learning, dopamine, cognition, neurodegenerative disorders

Introduction: Goal-directed movements can occur in different contexts, e.g. while operating different tools such as a tennis or a squash racket. Our brain therefore needs to learn both the specific sensorimotor transformations of each context and which context (including possible novel contexts) it observes at any given time(Oh und Schweighofer 2019). As the exact solution is intractable, our brain has to rely on approximations. We propose a model whereby human learners approximate this problem by pursue the computational goal of minimizing free energy(Friston 2010),constrained by a generative model of context.

Methods: Our model simulates a learner who, on each trial, experiences a noisy observation of a sensorimotor perturbation context (e.g. a visuomotor rotation) and infersthe size of the perturbation and context responsibility. The perturbation observed is either the same as on the previous trial or is drawn from a large pool of possible perturbations that can also be previously unobserved. Our algorithm minimizes free energy under this generative model, with the additional approximation that it only begins to represent a potential novel context in memory if this reduces the free energy for the current observation, thus accounting for limited memory.

Results: This model qualitatively capturescanonical phenomena, such as savings, spontaneous and evoked recovery, and slower decay after learning smaller perturbations, under a single set of hyperparameters.It does so via inferred context rather than memory updating, in line with recent findings(Heald et al. 2021; Oh &Schweighofer2019).

Conclusions: It has been suggested that the brain may solve the contextual inference problem through a sampling-based algorithm (Heald et al. 2021). However, sampling tends to be computationally intensive or, if done overtly would not produce the relatively consistentmoment-to-moment motor output typically observed in humans. Our model obviates the need for samplings, rendering ita plausible starting point for eventually modelling implementation in biological hardware

Keywords: context inference: motor learning, free energy, Bayesian modelling

Augmented feedback is used to facilitate motor learning. By providing objective information on movement success and/or discrepancies between the actual and the desired movement outcome, it supplements the ever present subjective (intrinsic) feedback. While there is substantial research on how different variables of augmented feedback design affect motor accuracy and consistency in early stages of learning, empirical data on how these variables affect long-term learning in terms of automatization is scarce. Motor automatization, however, is of high significance, as it enables a high level of motor performance even in dual-tasking situations that challenge performers limited cognitive/attentional resources. Over the last decade, our group addressed this issue from a behavioral, a neurophysiological, and a genetic perspective.

Drawing on basic assumptions from the Parallel-Neural-Network-Model, the Feedback-Intervention- Theory, and the Explicit-Hypothesis-Testing-Approach, our group conducted a series of experiments on extensive feedback-based motor sequence learning with a dual-task paradigm to measure motor automaticity. The effects of feedback frequency, bandwidth-feedback, and normative feedback were assessed in between-subject designs. Moreover, the implications of the Reward-Prediction Error Hypothesis of Dopamine in reinforcement learning, and the concept of Hierarchical Error Processing, were considered with an inner-subject design. In order to examine the neural processing of valence-dependent feedback, event-related potentials (ERPs) in the electroencephalogram (feedback-related negativity [FRN], late fronto-central positivity [LFCP], P300) were assessed. In a post-hoc study, we also ventured a genetic variation of the dopamine metabolism and its potential effect on dopamine-mediated long-term-potentiation (as induced by positive reward-prediction errors) to trigger motor automatization.

The empirical data from these studies revealed that motor automaticity can be facilitated by a reduced frequency of augmented error feedback, and by events that enable the perception of success (e.g., bandwidth feedback or normative positive feedback). Also, valence-dependent neural feedback processing changed with extensive practice of a novel motor task. ERPs were differentially predictive for different categories of short-term behavioral changes and long-term motor automatization. Concerning long-term learning and motor automatization, a positive correlation was found for the reduction of dual-task costs and LFCP-amplitudes in the early practice. The assumption that the ERP-components are related to distinct mechanisms (reinforcement learning, supervised learning) of feedback-dependent learning could be supported based on their predictive value for different trial-to-trial adaptations (FRN: goal-independent; LFCP: goal-directed), and further by dissociating changes in their latencies. Moreover, a genetic variation of a dopaminergic enzyme called catechol-O-methyltransferase (COMT) predicted individual differences in motor automatization.

Key words: augmented feedback, automaticity, motor learning, event-related brain potentials

Motor imagery has been defined as the cognitive process during which motor actions are internally simulated without producing an overt action. This process might be beneficial for recovery after neurological damage but it may also improve motor skills in healthy individuals. In a series of studies with a Go/NoGo version of the discrete sequence production (DSP) paradigm we explored several issues by focusing on both behavioral and electrophysiological measures (the electroencephalogram [EEG], the electromyogram [EMG], and the electrooculogram [EOG]). First, we examined whether learning a fine hand motor skill with motor imagery induces sequence-specific learning effects, which was shown to be the case. In a second study, we explored to what extent learning by motor imagery could replace learning by motor execution and in a third study, we examined whether the learning effect by motor imagery is effector-specific or not. In a fourth study, we explored whether transcranial direct current stimulation (tDCS) above the primary motor cortex combined with motor imagery influences sequence-specific learning. In a fifth study, the question was addressed whether pianists have additional benefits of motor imagery as compared to controls. More generally, we also examined in this study whether pianists display improved sequence leaning, which could be related to structural changes in the connectivity between different cortical motor areas. Finally, we explored whether motor imagery can be considered as more effortful than motor execution, possibly due to the lack of sensorial feedback and the need for inhibitory control processes, and whether motor imagery is actually comparable to motor preparation. Several implications of our findings for motor sequence learning will be discussed.

Key words: motor imagery, motor sequence learning, motor preparation, EEG

Motor planning is a key element of skilled motor control, particularly for sequences of movements and its impairment can disrupt fluent goal-directed behaviour. Until recently, our knowledge of covert motor sequence planning in humans has been indirect, primarily derived from invasive recordings in animal models, computer simulations and behavioural markers during sequence execution. I will show how we can use pattern classification of non-invasive neurophysiological (MEG/EEG) and fMRI signals to uncover the rapid pre-planning of sequence elements and high-level features such as sequence order and timing from memory. Our results suggest that sequence planning entails an automatic separation of order and timing prior to motor execution, including a covert parallel pre-ordering of sequence elements and a serial rehearsal of planned sequence speed. Together, our findings favour a hierarchical model of sequence control with potential relevance to clinical interventions.

Previous studies linked enlarged beta desynchronization during motor preparation and execution (movement-related beta desynchronization; MRBD) and enlarged beta synchronization during postmovement (post-movement beta rebound; PMBR) to motor control. However, what remains unclear is how both beta synchronization and desynchronization during motor are linked to MSL. Therefore, this thesis aimed to test whether beta activity over M1 reflects motor sequence learning. Furthermore, it aimed to test whether RT correlates with beta activity over M1. If there is a relationship between RT and Beta activity over M1, this would mean that it would be possible to predict when motor learning expertise would be gained. In this thesis, participants carried out a go/nogo DSP task. Event-related desynchronization and synchronization (ERD/S) values were extracted from three separate bands: β1 (12– 17Hz), β2 (18–23Hz) and β3 (24–29Hz) in 100ms time windows for the motor preparation, motor execution and post-movement phase. The results revealed a larger β2 ERD post-training compared to pretraining during motor preparation. It was also revealed that there were no changes observed in MRBD. Based on the literature it was suggested that beta activity during motor execution may reflect task difficulty. Additionally, an enlarged post-movement beta rebound (PMBR) over M1 was observed for both the left and right hand in ß1 and the left hand in ß2 and ß3. This supports previous findings in which enlarged PMBR was linked to error-based motor sequence learning. Lastly, a positive relationship was observed between ß2 ERD/S over M1 for left-hand sequences in Block 5 and a negative relationship in Block 1. The relationship between ERD/S and RT should be further investigated in future research to confirm these findings.

While non-invasive Brain-to-Computer interface can accurately classify the lateralization of hand moments, the distinction of fingers activation in the same hand is limited by their local and overlapping representation in the motor cortex. In particular, the low signal-to-noise ratio restrains the opportunity to identify meaningful patterns in a supervised fashion. Here we combined Magnetoencephalography (MEG) recordings with advanced decoding strategy to classify finger movements at single trial level.

We recorded eight subjects performing a serial reaction time task, where they pressed four buttons with left and right index and middle fingers. We evaluated the classification performance of hand and finger movements with increasingly complex approaches: supervised common spatial patterns and support vector machine (CSP + SVM); unsupervised linear finite convolutional neural network (LF-CNN) and linear finite recursive neural network LF-RNN; available deep learning algorithms as EEGNet, FBCSP-ShallowNet, Deep 4.

The right vs left fingers classification performance was accurate above 90% for all methods. However, the classification of the single finger provided the following accuracy: CSP+SVM : –0.68 ± 7%, LF-CNN : 71 ± 10%, LF-RNN : 76 ± 5%, EEGNet : 72 ± 7%, FBCSP-ShallowNet : 66 ± 9%, Deep4 : 69 ± 8%. CNN methods allowed the inspection of spatial and spectral patterns, which reflected activity in the motor cortex in the theta and alpha ranges.

Thus, we have shown that the use of CNN in decoding MEG single trials with low signal to noise ratio is a promising approach that, in turn, could be extended to a manifold of problems in clinical and cognitive neuroscience.

Humans excel in instruction following to boost performance in unfamiliar situations. We can do so through so-called prepared reflexes: Abstract instructions are instantly translated into appropriate task rules in procedural working memory, after which imperative stimuli directly trigger their corresponding responses in a ballistic, reflex-like manner. But how much control do we have over these instructed task rules when their reflexes suddenly lose their relevance? Inspired by the phenomenon of directed forgetting in declarative working memory, we here tested across four experiments whether the presentation of (implicit or explicit) task cancellation cues results in the directed dismantling of recently instructed task rules. Our findings suggest that—even when cancelation cues are actively processed—such dismantling does not occur (Experiment 1–3) unless the no-longer relevant task rules are replaced by a new set of rules (Experiment 4). These findings and their implications are discussed in the broader context of action control and working memory.

Introduction: Learning a motor skill generally requires physical practice (PP) of the task. However, motor skills can be learned in the absence of overt movement either through motor imagery (MI) which involves mentally rehearsing the movement, or action observation (AO) which consists in observing others performing the skill. Sleep has been shown to benefit the consolidation of motor skills learned by PP, mainly through thalamocortical spindle activity (11-16 Hz) during non-rapid-eye-movement stage2 (NREM2) sleep. Recent work has further outlined the essential contribution of the clustering and rhythmic occurrence of sleep spindles in the consolidation process. However, it remains to be determined whether sleep spindle activity also supports the consolidation of non-physically learned movements, and how it affects the transfer of the consolidated motor skill to another effector. Hence, we aimed at determining whether motor skill learning through PP, MI and AO share similar sleep-dependent consolidation mechanisms.

Methods: Forty-five young adults learned a motor sequence task either through PP, MI, or AO (N=15 per group). Electroencephalographic sleep recordings were collected during a 90-min daytime nap following practice. Post-sleep behavioural retention and inter-manual transfer tests were performed to evaluate skill consolidation and effector transfer capacity, respectively. The amount of sleep spindles was extracted and related to behavioural performance. Time-frequency (TF) analyses were conducted to reveal the clustering and temporal organisation of NREM2 sleep spindles.

Results: Significant positive correlations were observed between the amount of spindles and the magnitude of skill consolidation following PP (r=.60, p=.019) and MI (r=.62, p=.014). However, the magnitude of skill transfer correlated positively with the amount of spindles following AO only (r=.59, p=.021). Additionally, TF maps revealed a predominant cluster-based organisation of NREM2 sleep spindles for all groups, with significant power increases in the spindle frequency band every 3-4 seconds.

Conclusion: Our findings reveal that a temporal cluster-based organisation of NREM2 sleep spindles underlies motor skill consolidation in all groups, albeit with different behavioural outcomes. We show that a daytime nap offers an early sleep window that promotes the consolidation and retention of the learned motor skill following PP and MI practice, while an effector-unspecific sequence representation favouring skill transfer is consolidated after AO practice. Altogether, we demonstrate that PP per se is not a pre-requisite for sleep-related consolidation of motor skills, and that the clustering of sleep spindles in trains may be a general mechanism for effective skill consolidation and transfer, depending on the initial practice mode.

Keywords: Sleep, memory consolidation, motor imagery, action observation

The striatum is a key structure in the motor learning network. However, the causal role of it can so far only be evaluated in animal models, patients with lesions or invasive brain stimulation methods. Novel innovative non-invasive invasive transcranial electrical stimulation methods allow now for the first time to modulate striatal activity and determine its impact on motor sequence learning. We show that theta-burst-patterned transcranial electric temporal interference stimulation (tTIS) can increase the activity in the striatum and the connected motor network with respective impact on behavioral performance in cohorts of healthy young and healthy old subjects. This proof-of-concept study opens exciting possibilities to target deep brain structures, such as the striatum, non-invasively in humans allowing to better understand their functional role.

Enhancement and optimization of motor learning through dyad practice A future challenge is to investigate training procedures and the underlying processes for enhancing and optimizing motor learning. The majority of the experiments on motor learning focuses on the effectiveness of the training procedures as indicated by delayed retention tests or transfer tests. Effectiveness is evaluated by faster response times, better transfer performance to movement variations, or resistance to forgetting following extensive practice. From an applied and theoretical perspective training efficiency is another important factor that has to be considered in designing and evaluating training protocols. Efficiency is evaluated by the time, the costs, risks for injury, or other resources that have to be taken into account to conduct the training session. Therefore, it seems that optimal training protocols should be both, effective and efficient.

An attractive training format that reduces potential risks of injury at individuals, makes lesser demands on instructors and coaches and cuts costs without sacrificing the amount of learning are cooperative and/or dyad training protocols. In dyad training protocols performers acquire a new skill in groups of pairs where individuals alternate between physical and observational practice on consecutive trials and are often permitted an inter-trial dialogue to exchange ideas. Even though observational practice is considered not as effective as physical practice combining both practice formats in a training session as in dyad training protocols can be an effective and efficient way to enhance motor learning. However, while conditions of practice, such as providing feedback, observation, or distribution of practice have received a good bit of experimental attention, little attention has been directed to combined practice regimen to increase the amount of learning.

The purpose of the present presentation is to tackle the difficult question of delineating the nature of the contribution made through observation and the inter-trial dialogue when it is combined with physical practice in motor learning in dyad training protocols to determine training effectiveness and efficiency. Besides, some theoretical considerations, recent experimental data will be presented that identify the conditions (cognitive and motor factors) and the process under which and by which combined training enhances motor learning in a dyad training protocol. Thus, the central hypothesis is that important new theoretical insights to enhance and optimize the learning of movements can be gained by forcing the perceptual-motor system to change continuously between observational and physical practice and to permit an inter-trial dialogue.