Learning as change of coordination dynamics: theory and experiment

Learning of coordination patterns was investigated theoretically from the point of view of a dynamic theory of biological coordination and with reference to recent experiments on the learning of relative timing patterns. The theory is based on theoretical and experimental work showing that coordinated movement is characterized not only by the actually performed pattern of coordination but by an entire dynamics of coordination. Theoretically, such dynamics are captured as equations of motion of relevant collective variables. Experimentally, signatures of these underlying dynamics can be found in the temporal stability of coordination patterns, which can be assessed through various stability measures as well as through processes of pattern change. We argue that not only intrinsic coordination tendencies, but also specific behavioral requirements, be they perceived, memorized, or intended, must be expressed in terms of such dynamics. The concept of behavioral information captures such requirements as part of the coordination dynamics. We expound two hypotheses on the nature of learning in this framework. First, we assume that at each point during the learning process the system is governed by a well-defined coordination dynamics. This equation of motion evolves with learning so as to acquire an attractor solution near the to-be-learned pattern. Second, we hypothesize that this change of the coordination dynamics, captured by the time course of memorized behavioral information, can itself be ascribed to an additional layer of dynamics, the slower learning dynamics. Testable consequences of these views are discussed in the light of recent experimental findings on the learning of a relative phase in rhythmic movement: (a) Learning affects dynamic properties of performed coordination patterns, in particular, their stability; (b) the change of the coordination dynamics due to learning leads to specific changes of behavior also under conditions other than the learned condition, namely, to systematic deviation toward the learned patterns; (c) learning may lead to instabilities in the coordination behavior if initial and learned performance differ sufficiently; and (d) the dynamic properties of the performed coordination patterns are distinct on the two time scales of learning and of performance.     

Evolution of behavioral attractors with learning: nonequilibrium phase transitions.

Learning a bimanual coordination task (synchronization to a visually specified phasing relation) was studied as a dynamical process over 5 days of practicing a required phasing pattern. Systematic probes of the attractor layout of the 5 Ss' coordination dynamics (expressed through a collective variable, relative phase) were conducted before, during, and after practice. Depending on the relationship between the initial coordination dynamics (so-called intrinsic dynamics) and the pattern to be learned (termed behavioral information, which acts as an attractor of the coordination dynamics toward the required phasing), qualitative changes in the phase diagram occurred with learning, accompanied by quantitative evidence for loss of stability (phase transitions). Such effects persisted beyond 1 week. The nature of change due to learning (e.g., abrupt vs. gradual) is shown to arise from the cooperative or competitive interplay between behavioral information and the intrinsic dynamics.     

Strength Recovery Patterns Following Isometric and Isotonic Exercise

In 3 experiments, the author tested the hypothesis that coordination dynamics is the content of a generalized motor program (GMP) for rhythmic interlimb coordination. In Experiment 1, learners (N = 14) practiced a ?90° movement with either identically timed or differently timed limbs. Both acquisition and transfer to novel (effector and pattern) timings were unaffected by the learning condition and were suggestive of the intrinsic dynamics for in-phase and antiphase. In Experiment 2, learners' (N = 13) acquisition of 2 different phase relations (?90° and ?45°) was qualitatively identical. Attractor reconstruction revealed an increase in the predictability of individual movement trajectories and a decrease in attractor dimensionality over learning. Transfer for both ?90° and ?45° was again suggestive of the intrinsic dynamics. In Experiment 3, learning altered participants' (N = 8) performance of in-phase and antiphase relations. Together, the results suggested a single continuum of phase relations, called an attractor landscape, that produces similar patterns of CE and VE for both previously stable and learned coordinations.     

Early learning differences between intra- and interpersonal interlimb coordination

Intrinsic coordination patterns exist between limbs such that 1) coordination at these states is inherently stable, 2) any other pattern requires learning to produce, and 3) this learning is subject to interference from a systemic bias towards intrinsic patterns. The dynamics that govern intrapersonal interlimb coordination also govern interpersonal coordination. However, intrapersonal coordination exhibits greater coupling strength and thus more stable intrinsic dynamics than interpersonal coordination. Because the strength of intrinsic coordination tendencies has consequences for learning coordination patterns, the differences in coupling strength between intra- and interpersonal coordination should impact the ability to perform new coordination patterns via greater or less interference from intrinsic dynamics. This was investigated by measuring participants’ performance as they learned a new coordination pattern alone (intrapersonal) or in pairs (interpersonal). Participants were implicitly tasked with learning the pattern as they separately controlled the vertical and horizontal position of an on-screen cursor to trace a circling target. We observed better performance of dyads on first trial and steeper learning trajectories for individuals. Overall, these results indicate that individuals experienced greater interference from stronger intrinsic coordination dynamics during early learning but could overcome this interference and achieve similar performance to that of dyads with very little practice.     

Learning and transfer of a relative phase pattern and a joint amplitude ratio in a rhythmic multijoint arm movement

According to the coordination dynamics perspective, one can characterize the learning of novel relative phase patterns as the formation of a stable attractor in the coordination landscape of the order parameter relative phase. The authors examined 18 participants' learning and transfer of a 90 degrees relative phase pattern and a 0.6-joint-amplitude ratio between the elbow and wrist. Variability in the relative phasing and the joint amplitude ratio between the elbow and wrist decreased with practice. Positive transfer of the 90 degrees relative phase pattern was not dependent on the learning arm (dominant or nondominant). Positive transfer of the joint amplitude ratio was dependent on the learning arm and the direction of transfer. The results demonstrated that relative phase is an order parameter that characterizes the coordination dynamics of learning and transferring multijoint arm movements, and they provide preliminary evidence that joint amplitude ratios act as order parameters in the learning and transfer of multijoint arm movements.     

Learning With Real Machines or Diagrams: Application of Knowledge to Real-World Problems

In this study, we investigated how learning from different media, either from real pulley systems or from simple line diagrams, affected mechanical learning and problem solving. Novice subjects learned about pulley systems by comparing the efficiency of different systems and receiving feedback on their accuracy. The main outcome measures were subjects' ability to compare pulley system efficiency, their level of mechanical reasoning, and their ability to apply knowledge of system efficiency and construction details. Experiment 1 showed that (a) subjects learning with the two types of media made equal improvement on the learning task, and (b) all subjects showed an increase in quantitative understanding as they learned, but (c) subjects who learned hands-on, by manipulating real pulley systems, solved application problems more accurately than those who learned from diagrams. Experiment 2 showed that both the realism of the stimuli and the opportunity to manipulate systems contributed to the improved performance on application problems.     

Qualitative and quantitative change in the dynamics of motor learning

The experiments examined qualitative and quantitative changes in the dynamics of learning a novel motor skill (roller ball task) as a function of the manipulation of a control parameter (initial ball speed). The focus was on the relation between the rates of change in performance over practice time and the changing time scales of the evolving attractor dynamic. Results showed 3 different learning patterns to the changes in the dynamics as a function of practice that were mediated by the initial ball speed. Only participants who learned the task showed a bifurcation in coordination mode that was preceded by enhanced performance variability. The observed multiple time scales to motor learning are interpreted as the products of the dynamical stability and instability realized from (a) the continually evolving landscape dynamics due to bifurcations between attractor organization and (b) the transient phenomena associated with moving toward and away from fixed-point dynamics.     

The influence of augmented feedback and prior learning on the acquisition of a new bimanual coordination pattern

The present research examined two variables regarding the acquisition of a new bimanual coordination pattern: the role of previous experience and the nature of augmented feedback. Two groups of participants acquired a new coordination pattern (135 degrees relative phase) following two sessions of practice of another novel pattern (90 degrees relative phase). Transfer of learning in these groups was compared to two groups that had not previously learned a new pattern, but were nevertheless influenced by coordination patterns that are intrinsic to the task of bimanual relative timing (in-phase, 0 degrees, and anti-phase, 180 degrees). The findings revealed that new learning overshadowed the influence of the intrinsic patterns. Learning was also greatly affected by augmented feedback: dynamic, on-line pursuit tracking information was more effective in transfer than static, terminal feedback. Implications of these findings regarding theoretical constructs in motor learning are discussed.     

Self-organization of coordinative movement patterns

We present a strictly operational approach — in which theoretical tools and experimental data are developed together — to the problem of understanding the coordination of movement patterns. The empirical aspects are guided by synergetics, a theory of spontaneous pattern formation in open systems. Following an outline of our theoretical strategy, recent experimental results are reviewed that demonstrate the validity of the approach. From these studies, it is possible to establish the linkage between order-order transitions in movement behavior and other nonequilibrium phase transitions in nature. Behavioral patterns (corresponding to low-dimensional attractors for collective variables) and their dynamics are shown to arise in a purely self-organized fashion from cooperative coupling among individual components. This step has been implemented analytically and computationally. The insights gained from the present research allow a generalization in the form of seven theoretical propositions that aim at characterizing pattern formation, stability and change in complex, biological systems. In turn, a number of new research directions emerge, including studies of the collective dynamics of the environment-movement system, learning, and multiple-limb coordination.     

Effector system mechanisms of coordination.

Current interest in motor learning has prompted researchers to develop models that explain the events or mechanisms underlying skill acquisition. For the most part these models have focused on sensory and perceptual mechanisms. This commentary presents a theoretical explanation of effector system mechanisms of coordination. Options are available in the central control centers and their peripheral components, tonic and phasic motor units. Implications for the methodology of motor learning are discussed as they relate to a paramount factor responsible for prompting individuals to select the most efficacious effector system mechanism option.     

Coordination Patterns in Ball Bouncing as a Function of Skill

By observing the coordination patterns of people of different ages and skill levels bouncing a ball, the authors addressed hypotheses regarding (a) the relative increase and decrease of degrees of freedom with learning and (b) the order of progression in the changing organization of those degrees of freedom with development and learning. The movement patterns of the dominant arm and hand of subjects at various skill levels were contrasted in a cross-sectional design so that the way movement organization changes as a function of practice could be examined. Nine subjects aged between 4 and 22 years bounced a basketball at a preferred rhythm while standing still. Each performance trial was videotaped, and motions of arm and ball were digitized. Statistical analyses were made of linear and angular body motions: pairwise correlations and multiple regressions, amplitude and period variability, phase relations, and spectral and coherency measures. The coordination patterns revealed that (a) the movement patterns of less skilled subjects were more variable than those of more skilled subjects and (b) the motions of the articulators showed directional changes as a function of skill level that began both proximally and distally and moved toward the center of the effector chain with practice. The findings point up the relevance of studying change of system organization for understanding control and coordination.     

虚拟场景中不同客体朝向条件对内在参照系建立的影响

通过眼动记录和部分场景再认两种方法,探讨了虚拟建筑物对称场景中物体朝向统一、凸显两种条件对内在参照系建立的影响。结果发现:(1)场景中均为有朝向建筑物且朝向统一时,被试选择物体朝向与对称轴建立内在参照系的可能性没有差异;(2)场景中只有一个有朝向建筑物,其余均为无朝向物体时,即朝向凸显条件下,被试倾向于选择对称轴来建立内在参照系。物体朝向对内在参照系建立的影响作用具有局限性和不稳定性。     

How motor practice shapes memory: retrieval but not extra study can cause forgetting

We investigated the retrieval specificity of retrieval-induced forgetting (RIF) of motor sequences. In two experiments, participants learned sequential finger movements, each consisting of the movement of two fingers of either the left or the right hand. In the learning phase, these motor sequences were graphically presented and were to be learned as responses to simultaneously presented letter stimuli. Subsequently, participants selectively practiced half the items of one hand. A final recall test then assessed memory for all initially learned items. We contrasted different kinds of selective practice with each other. Whereas retrieval practice required retrieving motor sequences in response to letter stimuli from the learning phase, extra study was an extension of the learning phase, that is, participants performed motor sequences in response to the same animation graphic display as in the learning phase again accompanied by the letter stimulus. All practice conditions strengthened the practiced items, but only retrieval practice resulted in RIF. Thus, the strengthening of items through practice did not suffice to induce forgetting of related motor sequences. Retrieval was a necessary component for practice to shape memory for body movements by impairing the subsequent recall of motor sequences that were related to the practiced motor sequences.     

Feedback interference and dissociations of classification: Evidence against the multiple-learning-systems hypothesis

Researchers have argued that different categorization problems are learned by separate and distinct cognitive systems. They propose that an explicit system is responsible for learning rule-based categories and that a separate implicit system learns information-integration categories. One source of supporting evidence involves experiments in which observers perform a concurrent memory-scanning task that interferes with the processing of feedback. Researchers have reported a dissociation in which this manipulation impairs learning of a rule-based category but not an information-integration category. In the present research, we test the hypothesis that the dissociation was the result of lowered perceptual discriminability in the rule-based structure in comparison with the information-integration one. We demonstrate an example of an alternate rule-based category with easy-to-discriminate stimuli in which performance is unaffected by the interfering memory-scanning task. Furthermore, we demonstrate that learning of an information-integration category with low perceptual discriminability is impaired by the memory-scanning task. These demonstrations of the reverse dissociation challenge the interpretation that rule-based and information-integration category structures are learned by separate cognitive systems.     

Explanatory limitations of the HKB model: incentives for a two-tiered model of rhythmic interlimb coordination

The HKB model for rhythmic interlimb coordination has highlighted the importance of coordinative stability and loss of stability, and introduced, with this focus, a new set of explanatory constructs. However, the phenomenological character of both parts of this model (i.e., the potential and the associated system of coupled oscillators) precludes an understanding of how the observed stability characteristics are related to more specific (e.g., biomechanical and neurophysiological) aspects of the movement system. A two-tiered model (involving a distinction between 'neural' and 'effector' dynamics) is discussed that offers handles for addressing such underpinnings of the identified coordination dynamics. The promise of the model in this regard is illustrated by two recent studies showing how explicit accounts of the effector dynamics may help disclose why (and how) particular properties of the peripheral system affect the overall coordination dynamics.     

Continuous executive function disruption interferes with application of an information integration categorization strategy

Category learning is often characterized as being supported by two separate learning systems. A verbal system learns rule-defined (RD) categories that can be described using a verbal rule and relies on executive functions (EFs) to learn via hypothesis testing. A nonverbal system learns non-rule-defined (NRD) categories that cannot be described by a verbal rule and uses automatic, procedural learning. The verbal system is dominant in that adults tend to use it during initial learning but may switch to the nonverbal system when the verbal system is unsuccessful. The nonverbal system has traditionally been thought to operate independently of EFs, but recent studies suggest that EFs may play a role in the nonverbal system—specifically, to facilitate the transition away from the verbal system. Accordingly, continuously interfering with EFs during the categorization process, so that EFs are never fully available to facilitate the transition, may be more detrimental to the nonverbal system than is temporary EF interference. Participants learned an NRD or an RD category while EFs were untaxed, taxed temporarily, or taxed continuously. When EFs were continuously taxed during NRD categorization, participants were less likely to use a nonverbal categorization strategy than when EFs were temporarily taxed, suggesting that when EFs were unavailable, the transition to the nonverbal system was hindered. For the verbal system, temporary and continuous interference had similar effects on categorization performance and on strategy use, illustrating that EFs play an important but different role in each of the category-learning systems.     

Learning to learn causal models

Learning to understand a single causal system can be an achievement, but humans must learn about multiple causal systems over the course of a lifetime. We present a hierarchical Bayesian framework that helps to explain how learning about several causal systems can accelerate learning about systems that are subsequently encountered. Given experience with a set of objects, our framework learns a causal model for each object and a causal schema that captures commonalities among these causal models. The schema organizes the objects into categories and specifies the causal powers and characteristic features of these categories and the characteristic causal interactions between categories. A schema of this kind allows causal models for subsequent objects to be rapidly learned, and we explore this accelerated learning in four experiments. Our results confirm that humans learn rapidly about the causal powers of novel objects, and we show that our framework accounts better for our data than alternative models of causal learning.     

When kinesthetic information is neglected in learning a Novel bimanual rhythmic coordination

Many studies have shown that rhythmic interlimb coordination involves perception of the coupled limb movements, and different sensory modalities can be used. Using visual displays to inform the coupled bimanual movement, novel bimanual coordination patterns can be learned with practice. A recent study showed that similar learning occurred without vision when a coach provided manual guidance during practice. The information provided via the two different modalities may be same (amodal) or different (modality specific). If it is different, then learning with both is a dual task, and one source of information might be used in preference to the other in performing the task when both are available. In the current study, participants learned a novel 90° bimanual coordination pattern without or with visual information in addition to kinesthesis. In posttest, all participants were tested without and with visual information in addition to kinesthesis. When tested with visual information, all participants exhibited performance that was significantly improved by practice. When tested without visual information, participants who practiced using only kinesthetic information showed improvement, but those who practiced with visual information in addition showed remarkably less improvement. The results indicate that (1) the information is not amodal, (2) use of a single type of information was preferred, and (3) the preferred information was visual. We also hypothesized that older participants might be more likely to acquire dual task performance given their greater experience of the two sensory modes in combination, but results were replicated with both 20- and 50-year-olds.     

Learning new names for new objects: cortical effects as measured by magnetoencephalography

We tracked the evolvement of naming-related cortical dynamics with magnetoencephalography when five normal adults successfully learned names and/or meanings of unfamiliar objects. In all subjects, the learning of new names was associated with pronounced cortical effects. The learning effect was of long latency and emerged as a change of activation in the same cortical network that was active during naming of familiar items. In four out of five subjects, the cortical learning effect occurred in the inferior parietal lobe. In three of these subjects, the cortical effect was left-sided. These results suggest that the inferior parietal lobe plays an important role in the acquisition of novel words, presumably as a part of working memory systems.