|Series| Feedback Techniques
What is Augmented Feedback?
Influenced by the work of Schmidt & Wrisberg (2008) and Utley & Astill (2008), Augmented Feedback (AF; also known as extrinsic feedback) has been defined as,
“Information that cannot be elaborated without an external source; thus, it is provided by a trainer or a display” (Sigrist, Rauter, Riener, & Wolf, 2013).
Within this definition, the only time that feedback is not augmented is when it is generated by the organism itself. To help unpack this definition, I will introduce an example inspired by the world of sport.
I want to learn how to drive in golf. To simplify things, in this instance, I just want to aim for maximum distance in a straight line. To do that I have to move my body in a particular fashion to hit the ball in a particular way. At first, I learn this task at a driving range in solitude. Under the above definition; any feedback that isn’t generated from my own repetition of the task is augmented. If we were to add another golfer into this scenario, I can now attend to their technique and the distance that they are able to achieve with that technique. We could also add in a trainer. The trainer watches my technique and provides me with advice on how to change my technique in order to maximise distance. Finally, we could add in further feedback in the form of an augmented display. This clever bit of imaginary feedback takes the positional data of my club (containing velocity, maximum extension and posture) and uses it to provide me with continuous feedback. This data is projected into a special pair of glasses that I am wearing (think google glass) so not to interfere with my vision and thus restrict my ability to hit the ball.
According to the definition introduced at the start of this blog; the other golfer, the trainer and the augmented display are all different types of AF. The first form of AF (the other golfer) is perceptual by nature; I can see the other golfer throughout the whole process of the swing. I can hear the sound of the club connecting with the ball. Finally, I can see the result of the process with the flight of the ball and the distance completed. The second form of AF (the trainer) is perceptual only by the delivery of the information; I hear the advice given to me by the trainer. The task here is to use this information to adjust my posture to achieve maximum distance. I may respond to information such as, “stand with your legs wider” and “hold the club higher up”. In comparison to the other two forms of augmentation, the latter form is fundamentally different in how it achieves maximising the distance that the ball travels. What I have yet to mention about the augmented display is how it takes the data about the golf club’s position, velocity and maximum extension point and delivers it to the golfer in a continuously useful fashion.
In order for this information to be useful, it is transformed into a single image much like Lissajous Feedback. We now have several variables mashed together into a single image. The golfer uses this to adjust their posture and swing, in order to maximise distance.
Much like the augmented display in our golf example, using the previous definition in the context of Coordinated Rhythmic Movement; LF can also be defined as augmented feedback, as it is by Wenderoth, Bock, & Krohn (2002). I will argue that this definition is problematic for the following reason. When people are trained solely with the use of LF in new coordination patterns (such as 90°); they become completely dependent on the FB (Ronsse et al., 2011). Meaning, removal of the FB equates to the removal of any gains that the FB provides.
Feedback Dependence: Why does this happen?
As far as motor production (the movement of the limbs), the task is no different with or without feedback. So why is there such a difference in production? In short, it’s about what the feedback does to the visual information of the task. Without any special feedback (like LF), the complexity of Coordinated Rhythmic Movement is dependent on the phase of the movement. Different phases indicate the phase difference between limbs. I think of this as the lag between two oscillating limbs, with 0° being perfect unison, 180° in perfect opposition and 90° being at that awkward point in between. It turns out that without feedback, 90° is hard to learn! How would this task be defined within the literature? Fortunately, a lot of thought has gone into how to define whether a task is simple or complex.
According to Wulf & Shea (2002), “a complex task is one that cannot be mastered in a single session, has several degrees of freedom and tends to be ecologically valid.” On the other hand, “a simple task can be mastered in a single session, has only one degree of freedom and tends to be artificial.”
Is 90° a simple or complex task?
Without any special feedback (like LF), producing Coordinated Rhythmic Movements at 90° is by definition, a complex task. It takes multiple training sessions to gain a level of stability in a new pattern, such as 90° (Snapp-Childs, Wilson, & Bingham, 2015). What happens to this complexity when we introduce LF? Results from a recent coordination paper show that by using LF, a high level of stability (indicating mastery) can be gained across multiple phases (including 90°) after just 9 minutes of practice (Kennedy, Wang, Panzer, & Shea, 2016). Thus, the introduction of LF fundamentally changes the definition of the task, from complex; to simple.
How does Lissajous Feedback change the task complexity?
The key component behind this change in task complexity is the transformative nature of LF. What was initially required of the task, has now fundamentally changed. Initially, the task required the ability to perceive two oscillators simultaneously. This is no longer required; the two separate oscillators have become one dot. The transformed task is to trace the onscreen shape (which is a circle for 90°) with the single dot that you control. To do that, the limbs must produce oscillating movements and while that is true, nothing on screen oscillates. Lissajous feedback has fundamentally transformed the visuo-perceptual information from a multi-oscillator tracking task to a single-dot tracing task.
Redefining Augmented Feedback
Part of this issue is that our current definition of augmented feedback is too general. If we were to define augmented feedback as, ‘to use what’s there (the information in its default state) but to add something useful to it’. Augmented feedback stands to be something that is additional rather than transformative. Thus, true augmentation lies with the ability to guide the learning of the task at hand, and for the learning of such to be retained when the augmentation is removed. If this is not the case, the feedback cannot be categorised as augmentation.
Just like LF, under the new definition where the other golfer and the trainer provide augmented feedback in various forms (visual information and knowledge of results); the augmented display should not be defined as augmented feedback but more accurately as transformative feedback. In both our golf example and LF, the nature of the feedback has fundamentally altered the task so that the default information is no longer dominant in regards to task performance. The new information provided by the visual display has transformed a perceptually complex task into a perceptually simple one.
Lissajous Feedback is commonplace in CRM, but it is not the only form of feedback used. Next, I will focus my review on a different form of feedback, which takes into account the default information of the task and avoids transforming this information by maintaining informational uniformity in the augmentation it presents.
Kennedy, D. M., Wang, C., Panzer, S., & Shea, C. H. (2016). Continuous scanning trials:Transitioning through the attractor landscape. Neuroscience Letters, 610, 66–72. http://doi.org/10.1016/j.neulet.2015.10.073
Ronsse, R., Puttemans, V., Coxon, J. P., Goble, D. J., Wagemans, J., Wenderoth, N., & Swinnen, S. P. (2011). Motor Learning with Augmented Feedback: Modality-Dependent Behavioral and Neural Consequences. Cerebral Cortex , 21(6), 1283–1294. JOUR. http://doi.org/10.1093/cercor/bhq209
Schmidt, R., & Wrisberg, C. (2008). Motor learning and performance: A situation-based learning appraoch. Human Kinetics Publishers.
Sigrist, R., Rauter, G., Riener, R., & Wolf, P. (2013). Augmented visual , auditory , haptic , and multimodal feedback in motor learning : A review. Psychonomic Bulletin and Review, 20:21(53), 21–53. http://doi.org/10.3758/s13423-012-0333-8
Snapp-Childs, W., Wilson, A. D., & Bingham, G. P. (2015). Transfer of learning between unimanual and bimanual rhythmic movement coordination: transfer is a function of the task dynamic. Experimental Brain Research, 233, 2225–2238. http://doi.org/10.1007/s00221-015-4292-y
Utley, A., & Astill, S. (2008). Motor control, learning and development. Bios Instant Notes. Taylor & Francis.
Wenderoth, N., Bock, O., & Krohn, R. (2002). Learning a new bimanual coordination pattern is influenced by existing attractors. Motor Control, 6, 166–182. Coordinated Rhythmic Movement.
Wulf, G., & Shea, C. H. (2002). Principles derived from the study of simple skills do not generalize to complex skill learning. Psychonomic Bulletin & Review, 9(2), 185–211. JOUR.