Facilitating Implicit Learning to Improve Neurorehabilitation in Stroke

Someone has a stroke in the US every 45 seconds, resulting in over 700,000 new strokes every year and stroke is the leading cause of disability in Veterans (American Heart Association Statistics Committee and Stroke Statistics Sub-Committee). The vast majority of these cases result in motor impairments, which frequently cause individuals to become dependent on others for daily functioning (modified Rankin Scale 3-5, see Lees et al., 2006). Specifically, upper extremity hemiparesis is the leading cause of functional disability after stroke and upper arm function explains about 50% of the variability in reported quality of life (Wyller et al, 1997). As such optimizing upper arm neurorehabilitation is a critical problem to address in the aging Veteran population.

"Rehabilitation, for patients, is fundamentally a process of relearning how to move to carry out their needs successfully" (Carr \& Shepherd, 1987). This statement posits that at its core neurorehabilitation is motor learning, but despite this principle, research in motor learning has had little impact on stroke rehabilitation (Krakauer, 2006). Recently there has been an interest in developing and testing new methods to optimize upper extremity rehabilitation. Investigators at the Baltimore VAMC have pioneered task oriented training paradigms to improve mobility (Macko et al., 2005) in those with chronic stroke. As part of this programmatic approach novel upper extremity robotics training programs have been developed to improve reaching, and limb coordination. However, the majority of these interventions rely on error-based learning strategies during rehabilitation, which foster task-related explicit knowledge. However, a corpus of motor learning research indicates that this may not be the best strategy to optimize motor learning, and thus neurorehabilitation.

Error-based learning involves receiving continual feedback of movement with the intent that the learner will make corrections to the movement in real time. Thus learning occurs through a series of repetitions in which the learner continually reduces the discrepancy between the ideal behavior and the observation of their own behavior. In other words, error-based learning fosters an adaptation to achieve the desired behavior. In contrast, operant conditioning learning strategies consists of the learner only receiving feedback about the quality of their movement at the end of the behavior. Thus, learning occurs through a series of reinforcement of the desired behavior in its entirety, which is more model-free than the adaptation incurred during error-based learning. A primary distinction between these two learning strategies is that error-based learning fosters explicit knowledge of the task, whereas operant conditioning fosters implicit knowledge (Krakauer \& Mazzoni, 2011). These two types of knowledge have drastic implications for functional outcomes (i.e. motor performance, cognitive workload, and retention).

Prior to stroke, upper arm functions such as reaching and grasping were largely done without the use of explicit knowledge. In other words, healthy individuals devote little conscious effort about how they are controlling their limbs, they just 'do it'. Although, using explicit strategies during learning can facilitate the rate of learning, if given enough time, individuals who have limited explicit knowledge will perform equally well (Maxwell et al, 1999). Despite a slower rate of learning, the payoff of reducing explicit knowledge of the task can be very advantageous during motor performance. Notably, retention of the learned behavior is greater in individuals who learned under conditions that inhibit explicit knowledge. For example Malone and Bastian (2010) had individuals learn a novel walking task (split belt treadmill where the belts move at different rates) and in those in which explicit knowledge was limited exhibited learning that persisted longer than those who relied on explicit knowledge during learning. In addition, limiting explicit knowledge during motor learning may result in reduced cognitive workload and maintained performance under conditions of challenge (Zhu et al., 2011). In conclusion, promoting explicit knowledge during rehabilitation rather than unconscious control (limiting explicit knowledge) reduces sustainability of the newly acquired motor skill, and consumes cognitive resources, which need to be available for other demands. As such, automatic control of these behaviors is critical to perform daily activities, suggesting operant conditioning (which limits explicit knowledge) as superior to error-based learning.

Those with stroke are able to learn tasks implicitly, although the rate of learning may be delayed as compared to healthy controls (Pohl et al., 2001) and delayed further as a function of stroke severity (Boyd et al., 2007). Further, simply providing explicit information about an implicit task has been shown to reduce the learning rate and retention in those with basal ganglia stroke (Boyd et al., 2004; Boyd et al., 2006) and damage to sensorimotor areas (Boyd et al., 2003; Boyd et al., 2006; Winstein et al., 2003). While these studies highlight the importance limiting explicit knowledge during learning they were done in the context of learning implicit sequences rather than the development of skill, which while related, rely on different aspects of motor learning (Krakauer \& Mazonni, 2011, Yarrow et al., 2009). In the context of functional skill learning, the timing/ type of feedback have been robustly shown to affect the learning rate as well as retention and have been implicated to affect knowledge type (Levin et al., 2010). Specifically, providing feedback about task performance less frequently and after performance rather than during (i.e. delayed) have been shown to increase learning retention and likely facilitate implicit learning (Cirstea et al., 2006; Winstein et al., 1996). Additionally, feedback about the results (knowledge of results) rather than the performance (knowledge of performance) has shown to increase retention and limit explicit knowledge (Cirstea el al., 2006; Sidaway et al., 2008; Winstein, 1991). Accordingly, the current proposal will attempt to foster implicit knowledge during the development of motor skill by manipulating when feedback is given and type of feedback.

The aim of the current study is to determine the effect of error-based learning versus operant conditioning learning on critical outcomes of neurorehabilitation (i.e. performance after learning, generalizability, cognitive workload imposed by the task, and retention).