Retraining Automaticity: Recovering the Procedural Memory of Walking after Stroke

Special Article – Gait Rehabilitation

Phys Med Rehabil Int. 2017; 4(4): 1125.

Retraining Automaticity: Recovering the Procedural Memory of Walking after Stroke

Studer MT¹* and Winningham R²

¹Healthy Aging and Neurology, Northwest Rehabilitation Associates, Salem, Oregon, USA

²Behavioral Sciences Division, Western Oregon University, Monmouth, OR, USA

*Corresponding author: Studer MT, Healthy Aging and Neurology, Northwest Rehabilitation Associates, 3270 Liberty Rd S. Salem, Oregon, 97302, USA

Received: September 19, 2017; Accepted: October 13, 2017; Published: October 20, 2017


Walking upright, on two feet, without a cane or other assistive device - is perhaps the most ubiquitous and arguably the most human of all procedural memories. Yet, walking did not start out as a procedural memory that we were innately born able to do. How did walking become “automatic&”, or procedural? Is that capacity lost after stroke? The motor control for walking, while it can be adjusted for environment and task, is a relatively individualized pattern superimposed on many basic similarities from person to person. The pathway by which we develop “our&” recognizable pattern of walking includes two essential ingredients. First, we need an environment that allows for locomotion as a primary means of transit (i.e., repetitions). Second, we need sufficient demand (in the form of dual task distractions) to push the organization of the motor control of gait into the more primitive structures of the basal ganglia and cerebellum. Our recognizable “walk&”, evolves over time and, barring injury or change in somatotype (e.g., significant weight gain or loss), remains relatively constant through our adult years. In this paper, we will discuss both what is known in regard to creating this procedural memory, and what is known about re-creating this level of automaticity, particularly after stroke.

Keywords: Automaticity; Stroke; Gait


Approximately 450,000 people survive their first stroke each year in the United States [1,2]. A majority of cerebrovascular accidents (CVA) or strokes lead to limitations in the ability to walk, as stroke is a leading cause of disability in the United States [1]. Clearly, a stroke in the main motor centers of the cortex and internal capsule will directly impair walking. As walking is a complex task, strokes that affect vision, sensation, coordination, without direct lesion to the motor centers in the cortex or the cerebellum, will additionally result in observable gait impairments, be it expressed in safety or independence. Additionally, lacunar strokes affecting subcortical structures of the basal ganglia and thalamus affect gait, through the operating systems and interconnected nature of gait transitions and adaptability – turning, slowing, accelerating, adjusting to environmental demands (e.g., friction or uneven surfaces). Our main means of functional mobility, ambulation or “gait&” is nearly inescapably affected somehow, after a stroke, as more than half of stroke survivors over 65 years old experience reduced mobility.1 Because of the many means by which even one category of diagnosis – stroke – can affect gait, the potential and methods for recovery are quite diverse. However, the goal for all stroke rehabilitation would remain the same: restore patient’s automatic gait function, so it can be carried out reliably without increased cognitive burden.

Relearning to walk after sustaining a stroke may involve compensation, using a new area of the brain to substitute for the lesioned areas (e.g. primary motor cortex and conscious control of each step in gait), and early recovery strategies (assistive device, preferential weight shift, bracing) after stroke do encourage this behavioral adaptation [3-5]. While this functional recovery, toward meaningful and in some cases independent gait is occurring, the system may be charged to undergo a corollary process of neuroplasticity. Already well-cited and detailed elsewhere, neuroplasticity includes making and growing new connections as well as reorganization of responsibilities. Neuroplasticity is a viable option in most every stroke survivor and may not have the burden of a timeframe since stroke to be a viable option for improvement [3-5]. Partially because of the distributed responsibilities and complexities of gait, it is challenging for a person to truly make a complete recovery of ambulation after a stroke. Even when a person returns to an assemblance of normal in observation of gait, have they truly regained the act of walking in the same manner that they had prior to the event? One of the reasons that this is challenging to answer is because the act of walking is a procedural memory, stored in a normal central and peripheral nervous system. The task remains flexible and responsive to environmental demands (adaptability), yet can be operated largely without conscious control of the frontal lobe in steady-state and predictable environments. Can a person regain the procedural memory of walking after a stroke? If so, how? If the recovery is incomplete, is striving for the development of a new procedural memory with consideration for current capabilities (hemiplegia, sensory impairments, abnormal tone) still desirable? What are the mechanisms by which we can facilitate a return of the old procedural memory, or the creation of a new, reliable program that, albeit changed, can also operate without conscious control?

What is a Procedural Memory?

Memory is not a single entity or process, we have multiple memory systems [6] that can work independently or in concert with each other [7]. It is widely accepted that we have both a declarative and nondeclarative memory systems. The declarative memory system allows us to recall, think about and discuss (or declare) our recollections. The non-declarative memory system includes procedural memory, which supports motor behaviors. The declarative system relies on medial temporal brain structure, such as the hippocampus [8], while procedural memory relies more on frontal-striatal systems [7]. The split between declarative memory and procedural memory was made clear with a patient known as H.M. After part of his medial temporal lobes were removed, H.M. couldn’t make new declarative memories but he could make new procedural memories, indicating an anatomical and functional dissociation between these memory systems [9]. Procedural memories are important for many of our daily activities, we use procedural memories to help us walk, sit, stand, use eating utensils, drink from a cup, use a phone, type, and brush teeth. Therapists often work on these tasks, in order to maximize independence and quality of life. If someone has severe amnesia, like we would see in the later stages of most types of dementia, repeats a motor behavior many times, in the same sequence, without error, then they can develop a new procedural memory [10]. The ability to create new procedural memories in the absence of declarative memories is an important intervention tool for therapists.

Another feature of procedural memories is their automaticity. Orrel and colleagues [11] reported that people who had experienced strokes could learn to do a balancing task, even if they lacked explicit or declarative knowledge of that task [11]. In other words they learned how to do the task but they lacked any knowledge of the task that they could think or talk about. Similar effects are possible for walking among people who have experienced a stroke. Kal and colleagues conducted a meta-analysis looking at the effects of procedural learning in people who experienced strokes [12]. The authors concluded that there is good theoretical support for the hypothesis that procedural learning can lead to improvements after a stroke, they asserted that there is a need for additional empirical evidence in this area. It is at this very point that we propose a seminal shift in the approach toward regaining automaticity. In their discussion, Kal and colleagues noted limitations in the studies that were included in the meta-analysis. We are adding to this list of limitations and proposing practical methodological change that can be implemented in the clinic and laboratory alike, described below.

Gait Features, Storage Centers, Neural Control

Every person’s walking is uniquely “theirs&” due to anthropometrics, personality, fitness, nurture (observation and imitation of those around them in development), and impairments (arthritis, restriction after chronic ankle sprains, etc). The commonality that we all share is more so in where we store walking and the motor programming therein, rather than the result of what “it&” looks like, the effects of the motor program [13].

The act of walking has a general “set&” and organization of efferent neural control that has diverse location in the primary motor area (M1), the supplementary motor area (M2), the Basal Ganglia, and the cerebellum as “procedural memory centers&” [13]. Reinforcing this concept are recent studies including Lee and colleagues’ 2017 study that cites the loss the procedural pattern of gait in subjects with caudate lesions [14]. Procedural memory centers have output through the internal capsule, brainstem, spinal cord, and peripheral nervous system. Most certainly, the parietal lobe and cerebellum share sensory responsibilities to monitor the map and expectations of the act of walking, while the cerebellum receives a copy of the intended act – and serves as an online or real-time comparator. We will not belabor the sensory system’s map of walking here, but rather refer the reader to Takakusaki’s (2017) [14] review for a more detailed explanation.

The importance of neural control centers for walking, in this paper, becomes how. How did walking come to be stored in these locations? Are we born with the organization of walking as an a priori procedural memory with the motor control established? If we have “centers of gait control&” or a central program generator – are we born with this…or is it developed? While neural centers are “in place&” for a maturation of walking and mobility, the transition to get to walking most often goes through the developmental stages, including activities that only gradually begin to approximate walking. Arguably, the act of walking itself, once achieved at +/-12 months, has benefited along the way from rolling, creeping, crawling, and plantigrade stages more so from the resources (muscular strength and endurance in the core and extremities), than from a central neural standpoint (motor program of walking). So, if the procedural memory of walking has not truly been created/formed until we walk – how, then, does walking become a procedural memory? Just like any other procedural memory, walking takes repetitions in isolation and repetitions with dual task demands, to become a well-formed, resilient procedural memory.

Developing Automaticity: Repetitions and Dual Tasking

As an adult, with 6-10,000 steps per day, gait should be wellformed as a procedural memory [15-17] given – as long as they are free of other influences (e.g., clothing or footwear changes, pain, etc). Clearly, the more skilled a person is, the harder it is to distract them enough to affect their performance [15-20]. Examples include the elderly patient that is not acclimated to new footwear, is ambulating with a new assistive device, is experiencing ankle pain, or is 5 days post surgical from total knee replacement. These individuals are not operating under full automaticity and can be more easily affected with distractions.

In a related consideration, patients with mild cognitive impairment (MCI) may have fewer cognitive resources to devote to the “equation&” of dual task tolerance [21,22]. In contrast, redundancy or reserve in intelligence and cognition can afford enough shared resources to divide (simultaneous), or decide (allocate appropriate resources through filtering) when faced with dual task demands.

How many repetitions of a task are needed, to make it “automatic&”, though? What are the testable parameters to define “arrived&” for automaticity [23,24]? It is difficult to estimate the number of trials needed to make a new procedural memory, as many factors are likely to influence the outcome. The timing of the rehearsal trials matters, for example, we know that spaced retrieval can lead to new skill development much quicker than other rehearsal schedules, even for people with cognitive impairment [25,26]. Another factor is whether the learner can use declarative memory to support the development of the new procedural learning [21]. If the learner needs to inhibit an existing procedural memory, in order to perform a new behavior that could also reduce the efficiency of learning.

Popularized in the 1990s, body-weight supported treadmill training (BWSTT) provided a stimulus for the research and clinical practice of gait rehabilitation after stroke. This environment of care provides safety, graded weight-bearing, avenues for task-specific overtraining, higher intensity, and reduced demand on therapists to afford more repetitions and more mobilization of the moderately and severely impaired patients [27-29]. The rehabilitation of gait after stroke continues to evolve through technology, including feedback platforms, body-worn sensors, virtual reality, and by many other means [30-32]. Through all of these advances in technology and evidence, a consistent element of retraining gait, is repetitions. As noted above and additionally supported in normal motor control acquisition, repetitions are one necessary element that must be included as humans build procedural memories and gain automaticity. Can automaticity be acquired by repetitions alone, or does the normal, much less impaired nervous system need a “push&”, a stimulus that requires the motor act be processed without conscious control? Here lies the essential question behind the novel application of dual task training that we are suggesting, provided not merely to reduce fall risk in distracting environments, but for the identified purpose of improving the very task from which we lure attention from. This is accomplished by a very familiar strategy in rehabilitation that is perhaps not so apparent under the veil of dual task training - that strategy being forced use. Dual task training forces the mechanics of gait to be processed and eventually stored at a procedural level, a stimulus that can lead to more reliable, immutable, and permanent motor control - resistant to interactions with the environment and affording a more adaptable survivor.

While dual-task demands are everywhere, no two people should be affected by the same environmental and task demands in the same manner. This variability is due to: 1) their past experiences with the proposed task or related tasks; 2) their relative automaticity in a primary motor task (influencing the cognitive reserve available); and 3) each person’s tolerance for a specific type of secondary task. Everyone has an individual skill set with both biological and experiential (nature and nurture) influence. As will be discussed later in this paper, research has clearly identified that gait can be altered through distractions. This is quantifiable. Dual task cost, the measure of dual task tolerance as noted earlier, has been shown to be higher with age, certain disease states, and with given levels of experience [33-37]. Variables from person-person in DT cost or tolerance is thought likely due to resources of conduction speed and reaction time, among other variables.

While there is general consensus that divided attention negatively impacts our ability to make new declarative memories [38]. However, it is less clear whether and how divided attention affects the development of procedural memories and their automaticity. Dividing attention might affect features of procedural memories and motor learning by impairing declarative memories that might support effective behavioral expressions [38]. For example, one might be learning to use a (4 legged) “quad&” cane, which would be motor learning, but the use and initial practice with the quad cane might be supported by the declarative memory that the cane should be placed in the opposite hand, from the injured or weak leg. Performance will suffer without that declarative knowledge. But other researchers have found that divided attention might aid certain aspects of motor learning. Divided attention has been shown to help learning for a visual - motor task relative to undivided attention or no training at all [39]. But it could be that for easier tasks the greater attention given under divided attention or dual task conditions could lead to better performance [40,41].

Methodologies of Testing Distraction Influence on Gait

Rehabilitation professionals continue to recognize but not yet fill the void that we have in both understanding and testing the effects of attention, and more specifically distractions, on mobility. Quite simply, we cannot reliably measure dual task tolerance in a valid and sensitive manner without contrived test batteries – yet. While there is often recognition of dual task intolerance, we do not have evidencebasis behind our practice to measure, prognosticate a response or justify further care.

There is increasing evidence to support the disabling effects of distractions on gait in people that have sustained a stroke. There is disagreement in exact methodologies for how to study this effect. Additionally, the predictive value across dual task testing to fall risk, remains elusive. Most typically, studies do operationalize the primary task as ambulation. The variables combined with walking have a broad range, and have included: 1) various types of secondary task distracter, or “modes&” of distraction: cognitive, visual, auditory, or manual [42,43]; 2) different methodologies (testing each task as a single-task prior to combining) [43,44-47]; 3) considerations of task complexity/novelty and reality (meaning how contrived or realitybased the task should or can be) [42,47], and 4) different forms of walking (direction, or speed) including backwards, self-selected, and maximal efforts [42,47]. In all, the most common investigations in dual task literature are dual task cost (DTC) or dual task effect (DTE), being the statistical reflection of change in performance in the primary task from single-task to dual task conditions. In many recent studies, scientists have measured cost for walking and the distracter, in an attempt to investigate the relative cognitive participation for each stimulus [42,44-47]. Well-established tests that are intended to reflect DT capacity include versions of the Timed Up and Go test, specifically the cognitive and manual dual task versions (CTUG and TUG-m). Each version retains the physical TUG test and was developed by the original TUG author, Shumway-Cook and colleagues [48-50]. Respectively, these versions superimpose serial subtraction and holding a cup of water, while conducting the timed walk test. Each test has inherent limitations in that the secondary tasks are not measured for accuracy or participation and, arguably (using McIsaac’s definition [21]), may not truly be a secondary task, but rather just a more complex version of the TUG.

Yang and colleagues (2016) [51] studied three different modes of dual task interference for chronic stroke survivors (verbal fluency, serial 3 subtractions or carrying a cup of water), across four different conditions of walking (walking forward at self-selected and maximal speed, walking backward at self-selected speed, and crossing over obstacles). In each condition, the participants’ performances in gait, as well as their cognitive participation/accuracy, were measured. The authors found a strong correlation with the cognitive version of the Timed Up and Go (CTUG) [51]. This study provides insight and contributes to the literature with some novel considerations in regard to the gait parameters for dual task testing.

Emerging DT testing strategies include overlaying distractions during sensory-organization testing (SOT), clinical test of sensory integration in balance (CTSIB), and timed [44,46,52] or sensorbased measurements during an ADL or in gait. Still, the most common applications of DT testing in the clinic include the CTUG and TUG-M, or an application that includes an overlay of common psychological tests such as the Auditory Stroop or Trails B tests [53- 54] in function. The future of dual tasking is likely to soon include more wireless gait analysis, as this is presently being explored on a clinical basis.

As noted above, a more recent consideration of dual task testing and examination includes the recognition of the types of distractions, or dual task modalities. These distinct modalities include: cognitive, visual, auditory, and manual distractions [42,43]. Table 1 describes the four modalities and provides some examples of each in daily function.

Citation: Studer MT and Winningham R. Retraining Automaticity: Recovering the Procedural Memory of Walking after Stroke. Phys Med Rehabil Int. 2017; 4(4): 1125. ISSN:2471-0377