| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Reports |
LA Boyd, PT, PhD, is Assistant Professor and Canada Research Chair in Neurobiology of Motor Learning, Department of Physical Therapy, Faculty of Medicine, University of British Columbia, T325-2211 Westbrook Mall, Vancouver, British Columbia V6T 2B5 Canada
ED Vidoni, PT, MSPT, is a graduate research assistant and PhD candidate, Department of Physical Therapy and Rehabilitation Science, University of Kansas Medical Center, Kansas City, Kan
CF Siengsukon, PT, MPT, is a graduate research assistant and PhD candidate, Department of Physical Therapy and Rehabilitation Science, University of Kansas Medical Center
Address all correspondence to Dr Boyd at: lara.boyd{at}ubc.ca
Submitted April 30, 2007;
Accepted October 12, 2007
 | Abstract |
|---|
Subjects: Younger (n=11, mean age=26.0 years), middle-aged (n=13, mean age=50.7 years), and older (n=11, mean age=77.5 years) adults who were neurologically intact participated in the study.
Methods: Participants practiced a sequencing task with repeated motor, spatial, and temporal dimensions for 2 days; on a separate third day, participants completed retention and interference tests designed to assess sequence learning and which elements of the sequence were learned. The mean median response time for each block of responses was used to assess motor sequence learning.
Results: Younger and middle-aged adults demonstrated sequence-specific motor learning at retention testing via faster response times for repeated sequences than random sequences; both of these groups showed interference for the motor dimension. In contrast, older adults demonstrated nonspecific learning (ie, similar improvements in response time for both random and repeated sequences). These findings were shown by a lack of difference between random and repeated sequence performance in the older adult group both at retention testing and during interference tests.
Conclusion and Discussion: Our data suggest that, when younger and middle-aged adults practice sequences containing multiple dimensions of movement, the motor element is most important for motor learning. The absence of sequence-specific change demonstrated by an older adult group that was healthy suggests an age-related impairment in motor learning that may have profound implications for rehabilitation.
| Introduction |
|---|
 TopAbstract Introduction Multidimensional Sequence...MethodResultsDiscussion and ConclusionsReferences |
|---|
| Multidimensional Sequence Learning |
|---|
|
TopAbstractIntroductionMultidimensional Sequence...MethodResultsDiscussion and ConclusionsReferences |
|---|
No other past work has examined the unique or combined contributions of the motor, spatial, and temporal dimensions of movement during sequence learning. This omission makes it unclear what dimension of movement physical therapists should focus on when constructing interventions. Often, motor learning principles,8,10–14 drawn from a serial response paradigm,15 have been used to guide rehabilitation. In the serial reaction time (SRT) task,15 participants respond to a repeated pattern of stimuli. Traditionally, only one dimension of movement must be learned, typically the motor map between stimulus and response.15,16 Although an experimental paradigm, the SRT-type sequence task provides an ideal means to truly isolate each dimension's role in overall sequence learning in a controlled environment.
In the present study, we adapted the SRT to contain interwoven motor, spatial, and temporal patterns or dimensions. After 2 days of practice, we serially altered one sequence dimension at a time (eg, a novel motor sequence performed in conjunction with the repeated spatial and temporal sequences that were practiced) on a separate third retention test day to determine the unique contribution of each dimension to overall learning of the repeating sequence. These "interference" tests allowed us to infer the contributions of each specific element of the repeated sequence to overall motor learning.
The present study represents an important first step toward understanding what dimensions of movement might be most important for stimulating motor learning; findings from this study may be extended to the clinic. For example, gait training contains motor, spatial, and temporal components—motor activity to move the limbs, spatial positioning of the limbs, and the temporal ability to coordinate limb and trunk activity. Therefore, the present work may provide physical therapists with an indication of which dimension of movement might be most important to emphasize while clients learn complex motor skills in the clinical setting.
Age and Multidimensional Motor Learning
Older adults have been shown to be able to learn simple (one-element) sequences.17–19 However, as higher-order (probabilistic instead of deterministic relationships) sequences are practiced14,20 or more "complex" movements are required,12 impaired learning sometimes has been demonstrated in older individuals. Age-related deficits have been reported in learning sequences of spoken words,21 spatial sequence learning,22 and nonspatial visual letter sequence learning,23 but not in simple visuomotor learning (ie, key-press sequences such as those used in the SRT task).13,18,24 In addition, little research has been conducted to examine sequence learning in the middle-aged adult. Often, this age range has been examined only as the control group for a sample with the disorder13,18,25,26 or during life-span studies.27–29
Several problems limit the generalizability of past work investigating the impact of aging on sequence-specific motor learning. Some work14,20 has shown age-related deficits in motor sequence learning; however, it is unclear why these findings have differed from those of other researchers17,19,24 who have not discovered impaired learning with advancing age. One problem with past work is the failure to use delayed retention tests to dissociate temporary practice or performance effects, which are noted within a single session, from longer-term changes in behavior that are related to learning.30 This is a critical distinction, as performance and learning processes are now considered to be both behaviorally30 and neuroanatomically31 distinct.
Purpose
The present study, therefore, considered 2 questions. First, we used interference tests to determine "what" was learned during multidimensional sequence acquisition. Second, we tested the effect of age on multidimensional motor learning of a task with motor, spatial, and temporal requirements. Past work20 investigating aging using unidimensional (ie, motor) sequences has not shown a consistent effect of age on sequence learning. Because we used a task that tested learning for all 3 dimensions of movement, however, we anticipated that our results might differ.
| Method |
|---|
|
TopAbstractIntroductionMultidimensional Sequence...MethodResultsDiscussion and ConclusionsReferences |
|---|
All participants in the older adult (mean age=77.5 years, range=71–83), middle-aged adult (mean age=50.7 years, range=38–64), and younger adult groups (mean age=26.0 years, range=23–35) were free from neurological damage (Tab. 1). The category of middle age is ill-defined and highly subjective.32 We chose to divide younger adults from middle-aged adults at age 36 years and middle-aged adults from older adults at age 70 years. Each individual provided institutionally approved informed consent; all participants also were screened for dementia with the Mini-Mental State Examination (MMSE) and for hand dominance with the Edinburgh Inventory (EI). Participants were not enrolled if: (1) they scored below the 25th percentile on the MMSE using age-adjusted norms,33 (2) they exhibited any frank or clinically evident signs of neurological impairment or disease,34 or (3) they had any orthopedic condition or color blindness that would impair response ability.
|
Stimuli were presented one at a time within 6 circles arranged equidistant from each other (spatial sequence) and from a centered fixation cross (Fig. 1). At predetermined intervals, one circle would fill with a color (red, yellow, blue, or green), prompting the participant to respond with the appropriate key press (motor sequence). Stimuli appeared between 400 and 1,500 milliseconds after the previous response or, if no response occurred, after 2,500 milliseconds. Stimulus onset times were pseudo-randomly assigned so that, in each 12-element set, 4 timed stimuli onsets occurred in less than 500 milliseconds, 4 onsets occurred between 500 and 1,000 milliseconds, and 4 onsets occurred more than 1,000 milliseconds after the previous stimulus (temporal sequence). The majority of participants (n=27) used their dominant hand as determined by the EI, although some participants used their nondominant hand (n=8).
Those participants who used their nondominant hand were equally distributed across age groups. Following completion of a 12-stimuli set, a fixation cross flashed (250 milliseconds).
|
On the third day, all participants completed sequence retention tests and temporal, spatial, and motor interference tests. Performance order was counterbalanced across participants. The retention block repeated the same sequence that was practiced on days 1 and 2. The interference tests manipulated 1 of the 3 learned sequence dimensions (Tab. 2): color (motor sequence), position (spatial sequence), or stimulus onset (temporal sequence). For example, in the temporal interference test, colors in the 12-element trial appeared in the same position and same order as during repeated sequence practice. However, the timing of stimulus onset was randomly altered (ie, identical motor and spatial sequence, but novel timing of interstimulus intervals). The amount of interference (slowing) for each transfer condition indicated the amount of learning of this distinct dimension. Thus, retention test data were used to demonstrate motor sequence learning and interference tests were used to determine what specific elements of the sequence were learned. If individuals learned a particular dimension of the sequence, then we expected that altering this dimension should disrupt performance (ie, slow response times).
|
) controlled the presentation of stimuli and recorded response time (RT) and key-press accuracy. Because all participants were highly accurate in their responses (average of 97% correct on day 1 of practice), key-press accuracy was not included as a dependent measure.
Outcome Measures
Response time (reaction time + movement time) is the time between stimulus onset and key press; it was measured and stored for each trial. As is standard procedure in SRT task data analyses,1,6,8,35 we calculated the median RT for each 12-element sequence trial. Calculation of median RT values for each sequence trial reduces the sensitivity of this measure to very large or very small values. Because RT data can be highly variable, the use of median RT as an outcome measure reflects a more conservative approach to data management.36 Response times were summarized by calculating the mean median for each block of 10 trials.15,37 This procedure was performed for both random and repeated sequences. Premature and absent responses were not included in performance measures.
Early in practice, RT is often very long but decreases rapidly as participants become familiar with the task. Faster RTs can be due to nonspecific learning of the relationship between stimulus and response (ie, learning to associate the color with a key press) and do not necessarily relate to learning the practiced, repeating sequence (ie, sequence-specific learning). To account for this and to allow for between-subject comparison, change scores for each block were calculated on a subject-wise basis using the random sequence block on practice day 2 as a baseline.13,38,39 This block was chosen because individual participants were already familiar with the task and, thus, nonspecific learning effects were reduced. Therefore, RT change scores were calculated by subtracting the mean median response of the random block from day 2 from the mean median response in all other repeating sequence blocks for each participant. The same transformation was used to calculate a learning score for retention and interference tests.
Statistical Analyses
Motor performance: acquisition practice.
Acquisition practice performance was assessed using a 2-factor repeated measures analysis of variance (ANOVA) (group x block [1–10]), with RT change score as the dependent variable. Additionally, a 2-factor repeated-measures ANOVA (group x block [1–10]), with absolute RT as the dependent variable, was performed to assess raw performance differences between groups.
Motor learning: retention tests.
Motor sequence learning was evaluated in 2 ways. First, to determine whether sequence-specific learning took place in each group, paired t tests determined whether absolute retention test RTs were significantly different from random performance. Next, to assess between-group differences in motor sequence learning, a 1-way ANOVA was performed using retention test RT change scores. Because of our sample size, we also performed effect size (ES) calculations with a pooled standard deviation to verify the existence of meaningful differences between the day 2 random and the sequence retention blocks.40 Thomas et al40 estimate that an ES of .8 or greater represents a large effect, .5–.8 represents a moderate effect, and below .5 represents a small effect.
Transfer interference of motor learning.
To evaluate which specific elements (ie, motor, spatial or temporal) of our 3-way sequence were learned, we considered the effect of altering one element of the repeating sequence via a paired t test and calculated the ES using a pooled standard deviation40 between retention and each interference condition.
Response variability.
To additionally characterize age-related differences in sequence acquisition, we examined response variability by practice block for each group. The coefficient of variation of RT for each 12-element trial was calculated. The mean coefficient of variation was used to summarize variability in block. Change in variability from the first sequence block to the sequence retention block was determined and correlated with age, providing a measure of overall change in variability.
To account for repeated comparisons in our analysis of the interference tests we used a significance level of 
=.016 to protect against a type I error. Where assumptions of sphericity were violated, a Greenhouse-Geisser correction was used. All statistical analyses were performed using SPSS 13.0.
| Results |
|---|
|
TopAbstractIntroductionMultidimensional Sequence...MethodResultsDiscussion and ConclusionsReferences |
|---|
|
Motor Learning: Retention Tests
Consistent with our data from the acquisition practice phase, the older adult group failed to learn the repeated sequence; absolute RTs for the older adult group did not significantly differ between performance of day 2 random sequences and repeated sequences measured at the delayed retention test (P=.589, small ES=.1006; Fig. 2B). In contrast, the younger adult and middle-aged adult groups showed significantly improved RTs on the retention test compared with the random sequence performance (both P=.001, younger adults: large ES=1.75, middle-aged adults: large ES=1.54). Larger decreases indicate faster RTs for the repeating sequence in the retention test compared with random performance for the younger adult and middle-aged adult groups but not the older adult group (Fig. 3). A main effect of group (F=8.81; df=2,32; P=.001) confirmed this age-related difference in sequence-specific motor learning. The Tukey Honestly Significant Difference revealed that the older adult group displayed significantly less change in RT than the middle-aged adult and younger adult groups.
|
Younger adult group.
Paired t tests revealed that the younger adult group learned the motor portion of the repeated sequence as illustrated by a significant difference between the retention and motor interference tests (Fig. 4; P=.001, large ES=1.04). This was not the case for the spatial (P=.17, small ES=0.265) or temporal conditions (P=.521, small ES=0.152).
|
Older adult group.
For the older adult group, no transfer learning was evident from the interference testing for any of the elements of the repeated sequence (Fig. 4; motor: P=.440, small ES=0.106, spatial: P=.231, small ES=0.107, temporal: P=.940, small ES=0.008).
Relationship Between Age and Sequence-Specific Learning
Linear regression analyses were performed to illustrate the relationship between age and sequence-specific motor learning. These analyses demonstrated a moderate relationship between age and motor sequence learning, as evidenced by change at retention (Fig. 5; r2=.324).
|
|
| Discussion and Conclusions |
|---|
|
TopAbstractIntroductionMultidimensional Sequence...MethodResultsDiscussion and ConclusionsReferences |
|---|
What Was Learned?
Our finding that the motor pattern was the most salient dimension used by younger and middle-aged adults during sequence learning is consistent with others who have reported interference when introducing a novel motor sequence.5,6 Across the literature at large, however, little consensus exists regarding what the brain finds to be salient during motor sequence learning. In part, the lack of consensus concerning what aspect of movement is most helpful for learning may result from varied experimental designs. In past work,7,41 learning of only 2 dimensions of movement—spatial and motor—has been examined. These studies have largely suggested that spatial information is more useful for performance,7,8,42 with the motor element playing a lesser, but by no means underrepresented, role.35,36 However, investigations of what is learned during sequence practice have not consistently used a retention test design30 in order to separate temporary performance effects from longer-term learning. In addition, we know relatively little about the contribution of temporal elements of movement to learning. Previous work has suggested that temporal context may help create expectations about the next response in a sequence43 and direct attention to serial events.44 At present, however, no other work has considered the integrated learning of motor, spatial, and temporal sequences.
Our findings may be at least partially explained by a learning model recently proposed by Keele et al.45 In this model, the authors posit that individual dimensions of complex sequences may be learned independently or as an integrated unit or both. In unidimensional learning, participants learn only one specific characteristic of a sequence of movements. More complex relationships such as those between movement dimensions (eg, motor, spatial, and temporal elements) are encoded by a multidimensional learning system.
The proposed model of Keele and colleagues raises the interesting question of which system each of our groups used to encode the repeating sequence. The older adult group did not perform repeated sequences better than random ones during the retention test. This finding suggests that older individuals could not take advantage of the multidimensional sequence regularities to improve performance; they treated each element in the repeating sequence as isolated and independent from previous or subsequent stimuli. We believe that result reflects a fundamental failure of the multidimensional learning system in older adults. This may have large implications for clinical motor learning situations. Although our data still need to be verified in a therapeutic setting, they suggest that even older adults who are neurologically healthy may not be able to learn complex, integrated sequences of movement such as those that may be necessary for activities of daily living.
In contrast, the younger adult and middle-aged adult groups did learn the multidimensional sequence and were able to transfer this knowledge in the motor domain. We propose that this reflects the salience of the motor sequence during multidimensional learning. According to Keele et al,45 even when multidimensional relationships are no longer readily apparent, such as during interference or transfer situations, if the sequence has been learned, tasks can be performed using unidimensional pathways. This system appeared to be available to the younger adult and middle-aged adult groups. For example, despite a novel spatial presentation in the interference transfer condition, the young adult and middle-aged adult participants were able to rely on the motor dimension of the sequence that they learned during practice to produce faster responses.
An Age-Related Sequence Learning Deficit
The absence of a difference during performance of repeated and random sequences in the older adult group suggests a failure to learn anything beyond general, nonspecific task characteristics. Variability data support and extend these findings. Recently, it has been proposed that variability is not only helpful, but essential, for motor learning and performance. Stergiou et al46 have suggested that the motor system strives for an optimal level of variability both in learning and in performance. These authors posit that, during learning, variability results in the development of a flexible motor plan for movement, a motor plan that, in turn, facilitates motor skill performance. In the present study, participants demonstrated an increase in response variability with practice that was inversely related to age; that is, older individuals became less variable with sequence practice. These data suggest that aging may result in more rote and inflexible behaviors that are less helpful for motor skill acquisition.
Past work has not uniformly noted impaired sequence-specific learning in older adults who are healthy. For example, compared with younger people, older adults demonstrate deficits in using past experience to improve visual search times47 and in sequence learning of a simple (one-element visuomotor) task.48 We speculate that the increased difficulty imposed by our multidimensional task was at least partially responsible for age-related differences in sequence-specific learning. An interaction between task difficulty and age has previously been noted,49 demonstrating that older adults who are healthy may have difficulty learning complex sequences.20,21 Skill complexity is notoriously difficult to quantify and often relies on qualitative and subjective judgment.50
Our finding that older adults were unable to benefit from practice of the repeating sequence to improve performance is certainly not the first data to imply that older adults who are healthy might struggle with sensorimotor learning. Harrington and Haaland12 found a comparable lack of sequence-specific learning in elderly participants who were healthy and concluded that diminished implicit motor memory was responsible. Recently, Shea and colleagues51 reported similar age-related findings using a continuous tracking task. Indeed, a theme is beginning to emerge suggesting that a deficit in sequence-specific learning might be a general feature of aging. Our findings appear to confirm and extend this earlier work.
Another, but not mutually exclusive, explanation for why our older adult group failed to demonstrate sequence-specific motor learning centers on the relative complexity of our multidimensional task. Information Theory is a useful tool that may assist both researchers and clinicians in comparing the relative complexity of various tasks.52,53 According to Information Theory,|| each presented element of a sequence or "signal" reduces response uncertainty by an amount that depends on all of the other possible responses. Thus, the sequence-related information that has to be learned can be calculated in bits that represent the average minimum number of yes-no questions necessary to determine each possible response. The relative complexity of various experimental tasks may explain the wide range of findings in the literature concerning sequence learning in healthy older individuals.50
If we consider the degrees of freedom available in the 12-element repeating sequence used in the present study (4 motor responses, 6 spatial positions, 3 onset times), we can conclude that, to learn our experimental task, participants encoded 57.37 bits of information. In contrast, a common motor-only sequencing task with 10 elements (the SRT task) that Boyd and Winstein13,26 and other authors15,17have reported requires that the learner encode only 12.17 bits of information. Previous work by Howard and Howard17 suggested that older adults can learn novel motor sequences. However, this work considered the acquisition of a short (16.26-bit) versus long (25.77-bit) sequence and found no age-related differences.
Most importantly, it is likely that none of the aforementioned experimental tasks approximates the complexity encountered in daily life. Dialing an unfamiliar 10-digit phone number, which has no spatial uncertainty, requires learning 33.22 bits of information. The interface and utilization of an automatic teller machine (ATM) are even more complex. There are commonly 19 buttons divided into 2 separate pads. Four ordered steps can approach 49 bits of information or more: (1) PIN entry (4 buttons selected from 1 of 2 keypads), (2) transaction selection (1 button from 1 of 2 keypads), (3) account selection (1 button from 1 of 2 keypads) with uncertainty regarding how choices will be arranged on the screen, and (4) withdrawal amount (2–3 buttons from 1 of 2 keypads).54,55 There is also a potential time limit for key selection (sufficient speed, too long). It may be critical to consider the issue of task complexity carefully when conceiving therapeutic interventions; we suggest that use of Information Theory may be one method with which to accomplish this task.
An alternate interpretation for our findings of diminished sequence-specific learning in a group of older adults who are healthy centers on the possibility that less change in RT with practice reflects peripheral physiologic changes associated with aging. We find this explanation unlikely. Because we indexed RT for repeated sequences to that seen during random sequences through our calculation of a change score, we insulated our data from large variations between the older and younger groups in overall speed of responding. Thus, we indexed capability of motor learning–related change in the capability for responding, rather than just examining general speed of responding that is influenced by age-related physiologic alterations. Similarly, it might be postulated that a set-shifting deficit rather than an impairment in sequence-specific learning was the cause of the limited change noted in our older adult group. However, because both random and repeated sequence performance improved similarly across day 1 of practice, we believe that a set-shifting deficit is an unlikely explanation of our findings.
Another possible explanation for the differences noted across our age groups might be a relative difference in computer keyboarding experience. It is plausible that our younger and middle-aged groups had much more experience with computer keyboards than did the older group prior to participating in this study. To avoid this possibility, however, we used several experimental controls, the most important one being the use of a change score as the primary dependent measure for between-group comparisons. Because the older adult group probably had less keyboarding experience, it would be most likely that they would have the slowest RTs initially; indeed, this was the case. However, if the older adult group actually had learned the repeating sequence, their initially slow RTs should have inflated their change scores, making them larger. Instead, the opposite was the case; the older adult group showed the smallest amount of RT change across practice.
Clinical Implications
We discovered that, for younger and middle-aged adults, the motor dimension of movements was most salient for sequence learning. Perhaps more clinically relevant, we also found that older, but healthy, adults were unable to learn our multidimensional repeating sequence; the net result was a failure to improve performance beyond that seen with nonspecific learning or general task familiarization. The data presented here have strong implications for physical therapy practice. As the population at large ages and increasingly accesses rehabilitation services, it will be important to consider age-related deficits in sequence-specific learning when planning and implementing therapeutic interventions. Currently, it is unclear whether increasing practice or adjusting the structure of practice sessions through instructions, feedback, or focus of attention can ameliorate or lessen age-related learning deficits; future work will have to consider these possibilities and communicate the results to enhance the efficacy of physical therapy interventions. It also is possible that neuropsychological testing may have revealed cognitive deficits that would help to explain the sequence-specific learning deficit we noted56; future work should consider this possibility. Finally, we suggest that Information Theory may provide a useful tool for both the clinical and research communities as a method to more precisely capture, consider, and communicate task demands when assessing motor sequence learning.
| Footnotes |
|---|
Funding for portions of this work was provided by awards from the North Family Foundation and Vancouver Costal Health Research Institute and Foundation to LAB. The authors thank Joan McDowd, PhD, for facilitating access to the Grayhawk Healthy Elder database.
* Dell Computer Corp, One Dell Way, Round Rock, TX 78682. 
These participants also were serving as hand-matched controls in a separate investigation.
Neurobehavioral Systems Inc, 828 San Pablo Ave, Suite 216, Albany, CA 94706.
SPSS Inc, 233 S Wacker Dr, Chicago, IL 60606.
|| Signal information: Hs=log2N, where N=number of possible outcomes.
| References |
|---|
|
TopAbstractIntroductionMultidimensional Sequence...MethodResultsDiscussion and ConclusionsReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
