Learning Under Conditions of Action Information Feedback

Methodological and empirical issues in research on action information feedback (AIF) are discussed, with particular reference to the procedures and results of Annett (1970). Positioning responses practiced under conditions of AIF training are learned, though generally less well than under terminal IF conditions. Effects obtained as a result of AIF training may be subject to variations in movement extent required, gain and other transformations, frequency, temporal locus, and mode of augmentation, many aspects of which remain to be systematically studied.     

The locus of adaptation to altered gain in aimed movements

Altered gain settings cause a mismatch between the actual movement amplitude across the surface and the distance covered on a real time visual display. The present study pursued three objectives of how adaptation to altered gain affects aimed motor behavior. First, we replicated findings of an earlier study reporting a negative linear relation between gain and both target acquisition time and end-point variability. This means that our data do not agree with the classic U-shaped relation between gain and acquisition time. Second, our results proved to be robust when we manipulated movement difficulty by varying target distance. And third, dividing a movement into four successive sections on the basis of key kinematic events revealed the locus of the adaptation to altered gain within movement execution. Time differences between gain conditions proved to start at a very early part of the movement, but appeared to be absent in the final movement section. In contrast, differences between gain conditions regarding the use of online feedback were also present in the last part of the movement during which the final target approach takes place.     

Trajectory evolution and changes in the structure of movement amplitude time series

With increases in the index of difficulty [ID = log₂(2A/W)], the time-series structure of movement amplitude values shift from pink to white noise. The appearance of pink noise at low-ID levels may be attributed to the dominance of feedforward control processes, while the appearance of white noise at high-ID levels may be attributed to increased reliance on visuomotor feedback processes needed to guide movement into the target region. Such within-movement corrections may disrupt the pink-noise time-series correlations that exist in the absence of feedback processing. In our prior work, movement amplitude was defined as the distance moved from movement start until its end. In contrast, in the current study we examined the time-series structure of movement amplitude values at each of 10 different percentages of time into the movement trajectory—ranging between 10 and 100% of the movement time (%MT)—at a low (2 bits) and a high (5 bits) ID level. We hypothesized that at both ID levels a pink-noise time-series structure would be seen during the early portions of the movement trajectory when feedforward control should dominate, but during later portions of the trajectory, increased whitening of time-series structure would emerge only under ID 5 as there would be an increased need to engage visuomotor feedback processes. Under ID 2, the same level of pink noise should be maintained across all %MT levels as movement should be under the same level of feedforward control throughout the trajectory. The only unpredicted result occurred at ID 2 where the pink-noise level increased with increases in %MT. We hypothesize that such strengthening of pink noise as a function of %MT reflects the engagement of early trajectory corrections superimposed on the initial feedforward signal, but, once such initial adjustments were made, feedforward processes increasingly took over as the trajectory neared its goal.     

Adaptation to a Nonlinear Visuomotor Amplitude Transformation With Continuous and Terminal Visual Feedback

The control of a cursor on a computer monitor offers a simple means of exploring the limits of the plasticity of human visuomotor coordination. The authors explored the boundary conditions for adaptation to nonlinear visuomotor amplitude transformations. The authors hypothesized that only with terminal visual feedback during practice, but not with continuous visual feedback, humans might develop an internal model of the nonlinear visuomotor amplitude transformation. Thus, 2 groups were engaged in a sensorimotor adaptation task receiving either continuous or terminal visual feedback during the practice phase. In contrast to expectations, adaptive shifts and aftereffects observed in visual open-loop tests were linearly related to target amplitudes for both groups. Although the 2 feedback groups did not differ with respect to adaptive shifts and aftereffects, terminal visual feedback resulted in stable visual open-loop performance for an extended period, whereas movement errors increased after continuous visual feedback during practice. The benefit of continuous visual feedback, on the other hand, was faster closed-loop performance, indicating an optimization of visual closed-loop control.     

Adaptation to a nonlinear visuomotor amplitude transformation with continuous and terminal visual feedback

The control of a cursor on a computer monitor offers a simple means of exploring the limits of the plasticity of human visuomotor coordination. The authors explored the boundary conditions for adaptation to nonlinear visuomotor amplitude transformations. The authors hypothesized that only with terminal visual feedback during practice, but not with continuous visual feedback, humans might develop an internal model of the nonlinear visuomotor amplitude transformation. Thus, 2 groups were engaged in a sensorimotor adaptation task receiving either continuous or terminal visual feedback during the practice phase. In contrast to expectations, adaptive shifts and aftereffects observed in visual open-loop tests were linearly related to target amplitudes for both groups. Although the 2 feedback groups did not differ with respect to adaptive shifts and aftereffects, terminal visual feedback resulted in stable visual open-loop performance for an extended period, whereas movement errors increased after continuous visual feedback during practice. The benefit of continuous visual feedback, on the other hand, was faster closed-loop performance, indicating an optimization of visual closed-loop control.     

The effect of visual guidance on the acquisition of a simple motor task

Subjects were required to learn to depress a bar using concurrent visual feedback, which was removed for test. The overshooting found in test is attributed to visual dominance of the feedback traces; the way in which it occurs is not clear, since the reduction of the ratio of movement to display gain from 1:15 to 1:7.5 did not reduce overshooting. Performance improved after extended practice, but the direction of error remained positive. The results are discussed in terms of motor programming and feedback control of movement and are interpreted as evidence against Adams' (1971) two-trace theory.     

A Comparison of Two References for Using Knowledge of Performance in a Motor Task

In the present study, the learning of a task in which the goal of the movement was not isomorphic with a specific movement pattern was examined. The subjects' (N = 48) goal in the task was to be both spatially and temporally accurate in reaching 4 targets with a right arm lever movement. After each acquisition trial, the displacement profile of the movement just produced was provided to all subjects as knowledge of performance (KP). The relative effectiveness of 2 possible references, with which subjects could compare the KP, was examined. One of the references examined was knowledge of results (KR), which was provided by reporting the total absolute timing and amplitude errors from the 4 targets. The other reference examined was a criterion template (CT), which was defined as the most efficient movement pattern for reaching the 4 targets. In the feedback display, CT was superimposed on the displacement profile of the movement just produced. A factorial design, in which 2 levels of KR (KR, no KR) were crossed with 2 levels of CT (CT, no CT), produced 4 feedback conditions. After 120 acquisition trials with feedback, immediate and delayed retention tests without feedback and a reacquisition test with KR (20 trials per test) were conducted. Acquisition results indicated that KR was a better reference than CT for per-forming the timing aspect of the movement and for producing the generalized motor program (GMP) associated with the most efficient movement pattern. Delayed retention results showed that KR was also a better reference than CT for learning the most efficient GMP. The calibration strategy undertaken by subjects who were provided with KR during acquisition explains the superiority of the KR reference. The calibration strategy is compared with the pattern-matching activity that was probably undertaken by subjects who had received CT as a reference.     

The relationship between types of feedback, gain of a display and feedback precision in acquisition of a simple motor task

Three experiments are reported which investigate the role of concurrent and terminal feedback in the acquisition of a discrete positioning task. Experiments I and II compare the efficiency of concurrent visual feedback (CVF) and terminal visual feedback (TVF) as training methods when the gain of the visual display is varied from 1:1 to 4:1. There is a consistent interaction between feedback method and gain of the display over the recall trials. Concurrent visual feedback is inferior to terminal visual feedback at a gain of 4:1 in Experiment I and when the displayed and actual movement directions differ (Experiment II). Experiment III explores the relationship between concurrent and terminal feedback when feedback is of a digital form and its precision is varied. Concurrent feedback is worse as a training method although there is no interaction between feedback method and precision of feedback. These findings are discussed in the light of a variety of factors which could contribute to the inferiority of concurrent feedback as a training method.     

Studies on sensory feedback: III the effects of display gain on tracking performance

Instrumentation is described which permits study of the effects of different forms of visual feedback display on the patterns of fine movement obtained from the extended human index finger when the subject is attempting to keep his finger at a fixed point in space. The task is a compensatory tracking task in which the only source of input to the system is the subject's own finger movement. The effects of increasing the gain (or amplification) of a proportional error signal on the pattern of finger movement was studied. Gains of 1, 2, 4, 10, 20 and 40 were studied with a group of 24 subjects. Increasing the gain of a proportional error signal resulted in a marked improvement in the ability of subjects to maintain their extended finger at a fixed point in space. As the gain of the error signal was increased, the subject's high-amplitude, low frequency errors were reduced, and there was a progressive appearance of high-frequency activity of low-amplitude, more accurately centred about the reference position in space. A total off-target area measure (integrated absolute error) showed marked decrease in scores as the amplification of the error signal was increased from 1 through 10. Beyond this gain there was no appreciable additional improvement in motor control, however no degradation of control was noted to characterize the group performance.

Exploratory studies were undertaken to permit comparison of the effects of increasing the gain of a proportional visual display with the effects of increasing the gain of non-proportional visual and auditory displays. An increase in the dominant-energy frequency was noted as the error signal gain was increased, independent of whether a proportional visual, or non-proportional visual or auditory display was used. This observation suggests that common mechanisms mediate the processing of the gain parameters of feedback displays, in some measure independent of the display form or the sensory modality used for presentation.     

Electrophysiological measures reveal the role of anterior cingulate cortex in learning from unreliable feedback

Although a growing number of studies have investigated the neural mechanisms of reinforcement learning, it remains unclear how the brain responds to feedback that is unreliable. A recent theory proposes that the reward positivity (RewP) component of the event-related brain potential (ERP) and frontal midline theta (FMT) power reflect separate feedback-related processing functions of anterior cingulate cortex (ACC). In the present study, the electroencephalogram (EEG) was recorded from participants as they engaged in a time estimation task in which feedback reliability was manipulated across conditions. After each response, they received a cue that indicated that the following feedback stimulus was 100%, 75%, or 50% reliable. The results showed that participants’ time estimates adjusted linearly according to the feedback reliability. Moreover, presentation of the cue indicating 100% reliability elicited a larger RewP-like ERP component than the other cues did, and feedback presentation elicited a RewP of approximately equal amplitude for all of the three reliability conditions. By contrast, FMT power elicited by negative feedback decreased linearly from the 100% condition to 75% and 50% condition, and only FMT power predicted behavioral adjustments on the following trials. In addition, an analysis of Beta power and cross-frequency coupling (CFC) of Beta power with FMT phase suggested that Beta-FMT communication modulated motor areas for the purpose of adjusting behavior. We interpreted these findings in terms of the hierarchical reinforcement learning account of ACC, in which the RewP and FMT are proposed to reflect reward processing and control functions of ACC, respectively.     

Watching a cursor distorts haptically guided reproduction of mouse movement

Participants moved a mouse along a force-feedback-defined linear path, either without vision or while watching a cursor set to 1 of 3 levels of visual:haptic gain (all >1:1). They attempted to haptically reproduce the movement without visual feedback. Errors increased with gain, reaching 70% overestimation at the highest gain. Forewarning participants about gain variability did not eliminate this effect. The gain level was potentially cued during the movement by the mismatch between visual feedback and kinesthetic feedback. Moreover, because participants did not achieve cursor-speed constancy across gain levels, visual speed was another cue to gain. Collectively, these cues failed to prevent visual distortion of movement reproduction.     

Timing and accuracy of visually directed movements in children: control of direction and amplitude components

The reaction times (RTs), movement times (MTs), and final accuracy of hand movements directed towards visual goals were measured in 6-, 8-, and 10-year-old children, using tasks in which direction and amplitude components of movement were distinctly required. The tasks were performed with and without visual feedback of the limb. RTs decreased with age, and were shorter in directional than in amplitude task, in all ages. MTs were the longest at age 8 in both tasks, equally short at ages 6 and 10 in the directional task, the shortest at age 10, and intermediate at age 6, when amplitude had to be regulated. In the amplitude task, the target distance generally affected MTs under both visual conditions, but to a lower degree at age 10 than in the two younger groups. Movement accuracy, which was in all cases higher with visual feedback, showed different developmental trends among the two spatial components: directional accuracy was not different among the three groups of age, whereas amplitude accuracy showed a nonmonotonic development in the nonvisual condition, with an increase between age 6 and age 10, and the lowest level at age 8. In the visual condition, amplitude accuracy did not change with age. The specification of direction seems therefore to predominantly load the preparatory stage of the response. Amplitude specification seems to be more dependent on on-going regulations and to undergo a longer and more complex development, with a critical period around age 8 when a greater propensity for a feedback-based control appears on the two components. With increasing age, amplitude tends to be specific to a greater extent by a feedforward process.     

Wrist and arm movements of varying difficulties

Although a great deal of experimental attention has been directed at understanding Fitts' law, only a limited number of experiments have attempted to determine if performance differs across effectors for a given movement difficulty. In three experiments reciprocal wrist and arm movements were compared at IDs of 1.5, 3, 4.5 and 6. When absolute movement requirements and visual display were constant, participants' movement times and response characteristics for the arm and wrist were remarkably similar (Experiment 1). However, when amplitude for wrist movements was reduced to 8° and the gain (4×) for the visual display increased participants' movement time, defined on the basis of kinematic markers (movement onset − movement termination), was increasingly shorter relative to arm movements as movement difficulty was increased (Experiment 2). Experiment 3 where the arm was tested at 32° and 8° with the 8° movements provided the same gain (4×) that was used for the 8° wrist movements in Experiment 2, no advantage was observed for the arm at the shorter amplitude. The results are interpreted in terms of the advantages afforded by the increased gain of the visual display, which permitted the wrist, but not the arm, to more effectively preplan and/or correct ongoing movements to achieve the required accuracy demands. It was also noted that while the wrist was more effective during the actual movement production this was accompanied by an offsetting increase in dwell time which presumably is utilized to dissipate the forces accrued during movement production and plan the subsequent movement segment.     

Adaptive Inferential Feedback Partner Training for Depression: A Pilot Study

Adaptive inferential feedback (AIF) partner training is a cognitive technique that teaches the friends and family members of depressed patients to respond to the patients’ dysfunctional thoughts in a targeted manner. These dysfunctional attributions, which AIF addresses, are a common residual feature of depression amongst remitted patients, and are associated with poor long-term consequences. Thus, an AIF partner training intervention, as a supplement to individual cognitive-behavioral therapy (CBT), may help to improve clinical outcomes through the continuing reinforcement of the cognitive restructuring that takes place in the context of the patient’s individual treatment. This 10-patient pilot study examines the feasibility and outcome of the augmentation of standard CBT with 4 sessions of AIF partner training. The patients’ depression, anxiety, negative inferences, and perception of social support, and their partners’ knowledge and provision of AIF significantly improved over the course of the study, with gains maintained at 2-month follow-up. Further research is needed to investigate any incremental value of this intervention beyond standard CBT. A detailed case example, illustrating the application of AIF partner training techniques, is included.     

Visual correction of a rapid goal-directed response

The purpose of this study was to investigate the role of dynamic and static visual cues in improvement of accuracy during a pointing movement. In the experiment, subjects were required to point finger rapidly at visual targets as accurately as possible. Movement amplitude was 15 cm, and movement times ranged from 100 to 190 msec. Three visual feedback conditions were applied: no feedback, dynamic ongoing feedback on the complete hand trajectory, and static error feedback on the movement end-point. Two spatial movement outcomes were considered, mean constant error and intraindividual dispersion of pointings. Data were analyzed with regard to effects of feedback and speed. Under the no-feedback condition, accuracy was lowest; constant error was not speed-dependent, whereas dispersion increased with speed of movement. Accuracy was highest under the complete feedback condition and was speed-dependent, as shown by both constant error and dispersion. Under error feedback, accuracy was intermediate and was also speed-dependent. The results are discussed in terms of the interchange between correcting mechanisms vs delayed control within the motor regulatory processes.     

Methodology for motor learning: a paradigm for kinematic feedback

Knowledge of results (KR)--information feedback about goal achievement--has been one of the most extensively examined variables in motor learning. In most natural movement learning situations, however, instructors more common]y provide augmented information regarding various kinematic or kinetic aspects of the movement pattern itself (sometimes termed knowledge of performance, KP). But despite the inherent interest in kinematic feedback, several factors reviewed here have operated to inhibit its study, the most important of which has been the lack of a suitable laboratory task and paradigm. The limitations of earlier paradigms have concerned (a) the use of overly simple motor behaviors, probably to minimize the problems in kinematic measurement, (b) the tendency for the environmental goal or the task to be isomorphic with the kinematic pattern, and (c) thc failure to use transfer or retention tests as measures of learning effects of the feedback manipulations. In this article, we describe our efforts to create a new paradigm for kinematic feedback, the rationale for its development, and the details of its operation. Finally, we provide evidence that the task and paradigm are sensitive to manipulations of kinematic feedback, providing some assurance that the paradigm can potentially answer future research questions about the role of kinematic feedback for learning.     

Role of proprioceptive information in movement programming and control in 5 to 11-year old children

The role of proprioceptive inputs in the control of goal-directed movements was examined, by means of the tendon vibration technique, in 5 to 11-year old children performing a serial pointing task. Children pointed, with movements of various amplitudes and at various positions, by alternating wrist flexions and extensions. Tendon vibration was applied to both agonist and antagonist muscles to perturb relevant muscular proprioceptive inputs during the static or dynamic phase of the task, i.e., during stops on targets or during movement execution. Constant and variable amplitude errors as well as constant position error were evaluated. Vibratory perturbation applied during movement execution resulted in a similar reduction in movement amplitude, yielding an increased constant error in all age groups and a systematic position error in the direction of the movement starting point. Perturbing proprioception during static phases preceding movement resulted in an age-related increase in the variable amplitude error, which was maximal in 5-year old children performing extension movements. The results were interpreted in terms of the use of proprioceptive information in the feedforward and feedback based components of movement control in children. In particular, the results indicated (1) developmental changes in the relative weighting of each component, (2) an increased capacity to move from one strategy to the other, depending on the availability of information, and (3) developmental changes from an alternated to an integrated control of amplitude and position in serial pointing.     

Direct learning in dynamic touch

A dynamic touch paradigm in which participants judged the lengths of rods and pipes was used to test the D. M. Jacobs and C. F. Michaels (2007) theory of perceptual learning. The theory portrays perception as the exploitation of a locus on an information manifold and learning as continuous movement across that manifold to a new locus, as guided by information available in feedback. The information manifold was defined as a 1-dimensional space of inertial variables. To encourage maximal learning, a 2-step procedure was used in each of 2 experiments. Each step comprised a pretest to identify the starting locus on the information manifold, a practice phase in which feedback specifying the optimal locus was given, and a posttest in which the ending locus on the manifold was identified. In the 2nd step, a different feedback variable specified a different optimum. In both experiments, participants, who sometimes began at different loci, showed the predicted movement toward the optimum in each phase. Whereas previous applications of the theory posit the existence of information-for-learning without identifying a candidate variable, such a candidate is identified.     

Effects of feedback from active and passive body parts on spatial and temporal parameters in sensorimotor synchronization

Previous research on sensorimotor synchronization has manipulated the somatosensory information received from the tapping finger to investigate how feedback from an active effector affects temporal coordination. The current study explored the role of feedback from passive body parts in the regulation of spatiotemporal motor control parameters by employing a task that required finger tapping on one’s own skin at anatomical locations of varying tactile sensitivity. A motion capture system recorded participants’ movements as they synchronized with an auditory pacing signal by tapping with the right index finger on either their left index fingertip (Finger/Finger) or forearm (Finger/Forearm). Results indicated that tap timing was more variable, and movement amplitude was larger and more variable, when tapping on the finger than when tapping on the less sensitive forearm. Finger/Finger tapping may be impaired relative to Finger/Forearm tapping due to ambiguity arising through overlap in neural activity associated with tactile feedback from the active and the passive limb in the former. To compensate, the control system may strengthen the assignment of tap-related feedback to the active finger by generating correlated noise in movement kinematics and tap dynamics.     

海洛因戒断者对金钱奖赏的反馈加工异常：来自ERP的证据

本研究采用金钱奖赏延迟任务考察了海洛因戒断者对金钱奖赏和损失的预期和反馈加工及其电生理机制。结果发现,预期阶段两组被试由不同线索诱发的P3波幅不存在显著性差异。在金钱奖赏加工的反馈阶段,海洛因戒断组由金钱收益条件诱发的RewP波幅显著小于对照组由金钱收益条件诱发的RewP波幅,而在无收益无损失条件下两组被试诱发的RewP波幅不存在显著性差异。此外,海洛因戒断组由金钱收益条件诱发的P3波幅与无收益无损失条件诱发的P3波幅不存在显著性差异,而对照组由金钱收益条件诱发的P3波幅显著大于其由无收益无损失条件诱发的P3波幅。在金钱损失加工的反馈阶段,由金钱损失和无收益无损失条件诱发的RewP波幅在两组被试之间不存在显著性差异,且两组被试由金钱损失条件诱发的P3波幅均显著大于其由无收益无损失条件诱发的P3波幅。以上结果说明海洛因戒断者的金钱奖赏反馈加工存在异常,具体表现为其对金钱奖赏的敏感性降低或对金钱奖赏的趋近动机减弱。