Film “Amadeus”: Mozart plays piano upside down.
The problems of Differential Learning is not unusual exercises. Instead, it is the purpose of their usage.
Written by Peter Joffe
Take a look at this formula. It seems pretty complicated, right?
This is the math behind the “Differential Learning”, a very fashionable concept in sports science. The formula describes the behaviour of dynamic systems (Frank et al., 2008).
Well, that looks quite scientific, so what practical advice follows? “Chest the ball with your right eye closed, left arm straight up, right arm straight lateral” (Schollhorn et al.,2012).
Hm… It is a bit strange. But unusualness never scared me. Honesty, I prefer creative and even weird exercises in my sessions to kids queuing up to do nice-structured drills. So, let’s don’t be scared of math and try to understand the state of the art in the field of motor learning and practicality of the new pedagogical concept.
Differential Learning claims to have its roots in Ecological Dynamics and Nonlinear Pedagogy; thus, it will be useful to know more about these ideas.
Therefore, firstly we will try to understand the main theories of learning motor skills in sports.
Secondly, we will get an idea of nonlinear learning.
Finally, we will discuss whether Differential Learning is a good way to teach sports.
The development of motor skills in sports.
In this chapter, I will present the motor skills acquisition theories, which, in my opinion, contain the core concepts.
I should note that these theories have developed rather complex ideas over the years, and I, perhaps, do not follow them in detail. Therefore, I will try to present the topic in a simplified way.
The main competing groups in the field of motor learning are Motor Programmes theories and Ecological Dynamics theories.
Learning as creating motor programmes.
For quite a long time, people are interested in how sports experts can react quickly to situational changes.
It looks like they know what to do in advance. That led to the idea of the existence of some sort of prepared responses.
So in Motor Programmes theory view, there is a library of motor programmes in the brain. That is a coded representation of the real world stored somewhere in the “control centre”. These are very detailed instructions which do not need feedback and conscious control.
Therefore the motor programme is:
“… a set of muscle commands that are structured before a movement sequence begins and that allows the entire sequence to be carried out uninfluenced by peripheral feedback” (Keele, 1968, citation taken from Summers et al.,2009).
When an expert meets some environmental situation, he/she just automatically choose the corresponding response or set of responses. It is why the expert’s actions are so fast and precise.
In that case, learning is: creating a new item in the library or enrich old articles with further details.
The obvious problem with such an interpretation of motor learning is storage.
Indeed, there are infinite numbers of real-world situations; hence, each response should be, in essence, unique. If all responses have to be stored somewhere in the brain beforehand, that demands huge capacities.
Moreover, if you need to make a new set of commands for every situational change, how is it possible to do so quickly?
Still, in frames of motor programme theory, “Schema” concept was developed to address these questions.
It suggests that learning is not about developing a detailed programme but rather a generalised plan for a class of movements—”Recall” schema. This schema defines relations between the initial conditions, response parameters, sensory information, and action’ outcome (Schmidt R, 1975).
For dealing with concrete situations, this theory suggests a “Recognition” schema. It updates the main plan with concrete details of speed, force and effectors. For example, you may execute the same schema, say, a football kick, stronger or softer, with right or left foot.
The analogy that Richard Schmidt gave in his seminal article is meeting a dog. When you meet a dog, you Recognise that it is a dog, even if you have never seen that dog before. That is because you learned about dogs as a class of animals (Recall schema). So you don’t need to spend your time again on every dog you meet to understand that it is a dog. You just need to update your behaviour according to a particular dog (whether it is big or small, aggressive or not, etc.). That is much faster.
So learning, in that case, is a creating the general plan. At some point, enrich it with different variants of timing, force and effectors.
In my opinion, that makes sense, and this is quite a traditional approach.
Not surprisingly, some studies found that learning Recall schema needs more stable conditions and stays with us for a long time. In contrast, Recognition benefits from variable training and can be forgotten easier (Shea et al., 2001).
Prior knowledge about a response.
In essence, Motor Programmes theories, including Schema, requires one thing—planning. Although planning is definitely present in many actions, there are plenty of unexpected situations in sports games where experts still can find original technical solutions. Often these solutions are spontaneous and not planned in advance.
That can be not easy to explain, remaining exclusively within the framework of the theory of motor programmes.
So let’s have a look at other ideas.
In this theory, skilful performance may arise from direct interaction between learner and environment without the mediation of some entity in mind (control centre).
In its critique of motor theory Ecological Dynamics argues:
How can movement be rigidly programmed if, by its nature, every movement is different, even in standard actions? More than eighty years ago, pioneering Bernstein’s experiment of filming blacksmith hammering proved that subtle corrections of repetitive movements happen online and cannot be preprogrammed. Additionally, how every movement can be planned if the environment itself is continually changing? And, finally, who makes these programmes and where they are stored?
Hence, Ecological Dynamic suggests that it is possible to act without any programmes by assembling self-organised brain and body structures (Kelso, J., & Haken, H.,1995)
At a micro-level, self-organisation rules are simple; however, at the macro-level system can produce immensely complex behaviour.
We can find some examples in nature.
Have a look at this picture. Amazing, isn’t it? This is starlings murmuration.
Like the individual bird in the flock has no idea why she needs to turn, individual neurones in the brain have no “knowledge” of why it needs to fire. Moreover, there is no specific entity in the brain with better knowledge of the situation. However, despite that, self-organised systems can produce amazingly coordinated behaviour in reaction to environmental stimuli.
Therefore for in Ecological Dynamics learning is:
“… skilled performance derives from the increasingly improved (functional) fit of an individual and an environment, rather than from an increased complexity of acquired knowledge and associated computational and memorial processes.” (Araújo et al.,2011).
What exactly means “fit to the environment?” Perhaps that is:
Decreasing sensory threshold. Improving sensitivity to environmental cues and practical knowledge of environment which allows to see more possibilities. Developing body to afford new solutions. Improving coordination between all elements of the response. Making fast, precise and sensitive online corrections of the actions based on information from body and environment.
So, no programmes?
Well, an argument contra may be the experiment with spinal cord injured patients. They were unable to perform real foot movements (similar to amputees) however had distinctly different brain activation patterns when they imagined various foot movements. This means that they had some specific information about the actions of the feet in the brain, and they could activate it without actual movement.
Perhaps the murmuration maybe not an exactly correct analogy. Birds flock needs external input for changing its shape. Even more importantly, it has no memories of previous shapes and cannot repeat them. In contrast, the patient just needed to imagine movement, and he could reproduce it in the brain. So it looks like we have some mental representations of the actions we learned, at least of some of them.
Bernstein theory of the construction of motor skills.
Bernstein is deservedly regarded as one of the fathers of motor skills science. His ideas are still relevant today.
Both theories discussed above consider his views as their support.
What are his thoughts?
According to Bernstein, movements are constructed at different neurophysiological levels (Bernstein N.A.,1947).
These levels were developed evolutionary for dealing with the increasing complexity of motor tasks. For example, the lowest level A is responsible for regulating basic muscle tonus. It appeared when there was no movement in space, and muscles were needed for moving food inside the body. Level B —the synergies level that coordinated actions of different muscles and muscle groups was developed when limbs appear to grab and catch but still without whole-body movement in space. Level C— spatial, responsible for acting and moving in space. And so on.
Almost every movement is a mix of contributions from different levels where every level adds its colour to create a unique spectre for a particular motor skill. For example, for running, you need to keep posture (level A), muscles synergy between limbs (level B), interaction with a surface and keeping balance (level C) .
In the organisation of movement, Bernstein distinguishes background levels and leading level. His analogy is a pilot and engineers in the plane. Engineers are responsible for ensuring that everything works properly whereas the pilot can fly the plane.
Different levels may take a leading role depending on the stage of learning and the movement’s purpose.
For example, when we learn the running technique, a leading role may be assigned to level C. However, when we run to score a goal, level D (goal orientation) takes the lead (picture 1).
Picture 1. Construction of motor skills according to Bernstein.
Like the Ecological Dynamic, Bernstein thought that realisation of movement is possible due to the constant online flow of feedback and feedforward corrections in muscles and the brain. That is because everything, including automatic and repetitive actions, is not the same. There are no two absolutely similar movements. He coined the expression “Repetitions without repetition”, which was cited thousands of times.
Bernstein was strongly against the existence of rigid motor scripts somewhere in the brain.
He said: “Learning is not creating a formula of a movement somewhere in mind but rather creating the ability to solve the problem” (Bernstein N.A. 1991). For him, skill acquisition is not a finding and clearing some hidden ideal way that already exists but instead repeating the process of solving repeatedly, finally creating optimal for a particular person “dictionary” (his expression) of possible solutions.
Still, it looks like he had no problem with the existence of some sort of motor programmes, (otherwise what the “dictionary” mean?). He, however, made them flexible and gave them “manoeuvrability” (my translation from Russian) and the ability for instantaneous “recoding”—changing code (my translation). That is needed for active interaction with a constantly changing environment.
Learning, according to Bernstein:
Creating basic patterns: assigning leading level, necessary backgrounds, and developing energy-efficient synergies between muscle groups.
Developing sensitive and fast goal-orientated corrections.
Creating automatisms in background levels. Giving them more control of the action.
Hardening learned skills by repeating them in variable situations.
Summarising all above.
It looks like we still do not know for sure how exactly motor skills are built.
At one extreme of the learning concepts is the notion that we theoretically can acquire movement skills without actually doing it. All is needed just to implant some algorithm into the brain. The body is just a tool that follows commands.
Well, it looks like it cannot be done this way.
Learning motor skills is not only about the brain. It is unseparated from body and environment. The body is not a passive tool with which the brain operates. Instead, it is a part of the cognition which is directly in the environment. Body influences brain not less than vice-versa. All this is one dynamic system.
At an opposite end is the idea that denies the existence of any memory about learned movement. All motor actions are self-organised and spontaneous. Like the flock has no memories about its previous movement.
That also looks unlikely because this concept denies learning altogether.
How it may work.
Skills acquisition indeed based on creating and strengthening neurone’s connections in different brain areas.
It is not, strictly speaking, stored information but rather dynamic highways that facilitate flow of signals. The neurophysiological basis for that may be axon’s myelination and synapsis strengthening (Kato et al., 2020).
We unlikely have rigidly coded detailed information about the specific movement. Instead, these highways form some flexible skeletons or scaffoldings, whatever you call it, upon which variable constructions can be rapidly built.
Environmental situations can cause experts to make unplanned and spontaneous decision. Nevertheless, this “discovery” is always based on previous learning, experience, and body readiness. Johan Cruyff intuitively got it right describing how he did his famous “turn” for the first time against Swedish defender in the World Cup 1974 :
“The idea came to me in a flash, because at that particular moment it was the best solution for situation I was in. There are impulses that arise because your technical and tactical knowledge has become so great that your legs are able to respond immediately to what your head wants them to do. Even if that‘s nothing more than a flash in the brain.” (Cruyff, J. , 2016).
The unskillful player cannot find an original technical solution. Like novice musician cannot improvise. You need to have a repertoire of well-learned skills.
Armed with at least some understanding of what it is about, we can move on to new sports pedagogy theories.
Nonlinear methods are based on Ecological Dynamics.
Linear learning concept suggests the existence of some sort of ideal technique and ideal way of teaching. Thus there is a Teacher armed with a perfect and suitable-for-all methodology who leads the process: “Do as I do, or as I tell you to do”.
“Traditional teacher-centred methods of teaching and coaching assume a gradual, linear process of learning, with teaching methods often characterised by blocked practice drills with augmented teacher instruction and feedback designed to help students develop sound technique or idealised motor patterns” (Renshaw,et al.,2018).
In contrast, Nonlinear Pedagogy takes its ideas from Nature and puts learner-environment interaction in the centre.
Nature provides the learning environment and constraints that guide the acquisition process.
In sports, the coaches take the role of Nature. They need to identify short-term and long-term goals, create a training environment and exercises-constraints, that channel the search for technical solutions needed to achieve goals.
These exercises and training games should be “ecological”. It means that they represent game situations and make tactical sense.
Nonlinear learning implies that learner should get skills by the exploration of possible ways to solve problems. This is the individual way. There is no ideal technique suitable for all (Correia et al.,2019)
Nevertheless, there is even a more radical approach that claims to take Ecological Dynamics and Nonlinear Pedagogy ideas even further. It suggests that not only learning structured around the teacher is ineffective, but even an ecological approach based on constraints is not individualised enough. So, let us talk about Differential Learning.
Some consider this method as an extreme of Nonlinear Pedagogy.
Its proponents argue that there is no need for guided constraints, coach’s feedback and corrections , and repetitive practice.
The coach must ensure that the exercise does not repeat itself by continually adding new random variations to each rep. As I said at the beginning, sometimes these variations look strange.
The idea is that by always adding distracting, “noisy” factors to the target action, the coach creates destabilisation of the learner-environment system. This destabilisation pushes the system towards self-organisation at a new level. In simple terms, that means learning the motor skills more robustly.
The main proponent behind this theory is Wolfgang Schöllhorn.
It was his formula that was presented at the beginning of the article.
This formula was derived from inanimate dynamic systems behaviour, and it implies that the nature of noise does not matter.
Simply put, it does not matter how you change the target exercise; the main thing is to change it.
Thus the exercises suggested by Differential Learning is a mix of different, sometimes strange variations.
Let’s take a shooting drill in football.
Among ecological variations of exercise like, for example, “shooting bouncing ball” or ball “moving from different directions at different velocities” are variations “to shoot with eyes blinking” or “shooting with a tennis ball”. As a practitioner, I feel that something is wrong here.
Of course, I support the diversity and variability of practice; however, Differential Learning, in my opinion, lost touch with reality.
The problem of applying math in practical science is not unusual. Math is a useful tool, but one should have a comprehensive understanding of the field where math is applied. Richard Feynman explained that:
“Mathematicians are only dealing with the structure of reasoning and they don’t really care about what they are talking about… But, you have to have some understanding of the connection of the words with the real world.”
In my view, Differential Learning could not avoid this problem also.
I have a doubt that rules of inanimate or even simple animate dynamic system (e.g. birds flock) applicable to human’ learning. Acquisitions of motor skills is quite complicated process which depends on multiple neurophysiological, psychological, biomechanical, and, perhaps, even sociological factors. It is unlikely can be described even by 4-th order polynomial function.
Conducted in support of Differential Learning studies have not convinced me. They are often small (sometimes 4 subjects in a group), short (e.g. 8 sessions) and have dubious testing procedure (Schöllhorn et al., 2012).
Even more importantly, they mostly deal with adult players who were many years in the sport. Thus claiming that you have made a study where:” the players of the 5-th German division learn shooting skills” or “semi-professional badminton players learn to serve”(Henz et al., 2016) is, in my opinion, incorrect.They learned these skills somewhat twenty years ago when they were kids.
It is completely different to give a “strange” exercises to the already experienced player and the kid who just started to learn.
Why it is nonsense to learn playing the piano standing on one leg?
In my opinion, non-ecological constraints during the initial stages of acquisition interrupt the formation of basic movement patterns. The importance of the creation of these patterns is clear in Schema and Bernstein theories. It may look not so obvious in Ecological Dynamics, but, in my view, this theory supports ecological practice even stronger. The very logic of Ecological Dynamics is against unnatural exercises. Learners need all their senses and abilities to explore all spectre of possibilities. They have to form a harmonious learner-environment system.
Why do we need to diminish their abilities by creating unnatural restrictions? Why on earth kid needs to learn chest control with one eye closed?
Kicking a tennis ball does not help your child adapt to the football environment. When 6 y.o. kid plays with a size 3 ball; it is good because he has a small foot, and in this instance, a smaller ball preserves a foot-ball ratio. Additionally, this ball is lighter. That makes work with this ball more natural for this kid.
On the contrary, the tennis ball is far too small and difficult to control. That limits many explorational opportunities.
If I stand on one leg when I learn to play the piano, I will be busy not to fall, instead of music.
Variability of practice.
I am strongly supportive of the variable practice. It is crucial for acquiring motor skills.
Variability helps the learner to adapt to different circumstances. It indeed pushes player away from stagnation.
So perhaps the most debatable questions about variability in motor learning are:
When and what kind.
The coach should add variations only when students learn basic form robustly?
Not really, in my opinion. Instead, you can do it earlier when a template of motor skill is formed.
When the goal-orientation level, according to Bernstein theory, takes the lead.
For instance, in shooting, the reference point for me is:
I start to add variations to a shooting drill when, instead of thinking about how to hit the ball, the child starts thinking about how to score.
In this way, when the basic patterns begin to emerge, you can continue with the representative exercises where these patterns will be refined and enriched with ecological variation many times over.
I want to emphasise that variations should be representative—taken from the real game. Of course, they don’t have to be exactly like a game. Simplify exercises if necessary. Ultimately, they must be affordable and facilitate the acquisition of certain skills.
“Non-ecological” variations at the stage of initial learning may be detrimental.
Cacophony and improvisation.
The problems of Differential Learning is not unusual exercises. Instead, it is the purpose of their usage.
Different and unusual exercises can be implemented if you clearly understand the purpose. For example, I gave my students one-leg tennis strokes to accentuate core muscles’ work and keep balance. My boxers had sparring sitting in the chairs and one-against-three. Perhaps, you will find the purpose of “one-eye-closed chest control” as well. Nothing is wrong with it.
Differential Learning, however, is a mixture of ecological and non-ecological variations, where exercises are used inconsistently, without a specific purpose, only for the sake of destabilisation.
I think this method has limited usefulness.
Wolfgang Schöllhorn gave the analogy of Differential Learning as jazz trio improvisation without the main melody. No rules, it is complete harmony and understanding between jazzmen.
In reality, however, improvisation without the main tune in music is quite unusual and requires much skill. If you try to give this task to beginners, you will get cacophony.
Moreover, if we stick to jazz’s analogy, Differential Learning for me is to get musicians to play while standing on one leg, one eye closed, and turning upside down. This is not a real freedom. This is not a musical improvisation. Perhaps some great pianists can play even while upside down. However, they are not great because of that, and they did not learn this way.
So, let’s learn sports with both eyes open.
Araújo, Duarte, and Keith Davids. “What exactly is acquired during skill acquisition?.” Journal of Consciousness Studies 18, no. 3-4 (2011): 7-23.
Bernstein N.A. (in Russian) “О построении движений “ Медгиз-1947
Bernstein N.A. (in Russian) “О ловкости и ее развитии “ Физкультура и спорт-1991
Correia, Vanda, João Carvalho, Duarte Araújo, Elsa Pereira, and Keith Davids. “Principles of nonlinear pedagogy in sport practice.” Physical education and sport pedagogy 24, no. 2 (2019): 117-132.
Cruyff, J. a. (2016). My turn : the autobiography. London: Macmillan
Frank, T. D., M. Michelbrink, Hendrik Beckmann, and Wolfgang I. Schöllhorn. “A quantitative dynamical systems approach to differential learning: self-organization principle and order parameter equations.” Biological cybernetics 98, no. 1 (2008): 19-31.
Henz, Diana, and Wolfgang I. Schöllhorn. “Differential training facilitates early consolidation in motor learning.” Frontiers in behavioral neuroscience 10 (2016): 199.
Kato, D., Wake, H., Lee, P.R., Tachibana, Y., Ono, R., Sugio, S., Tsuji, Y., Tanaka, Y.H., Tanaka, Y.R., Masamizu, Y. and Hira, R., 2020. Motor learning requires myelination to reduce asynchrony and spontaneity in neural activity. Glia, 68(1), pp.193-210.
Kelso, J., & Haken, H. (1995). New laws to be expected in the organism: Synergetics of brain and behaviour. In M. Murphy & L. O’Neill (Eds.), What is Life? The Next Fifty Years: Speculations on the Future of Biology (pp. 137-160). Cambridge: Cambridge University Press. doi:10.1017/CBO9780511623295.012
Renshaw, Ian, and Brendan Moy. “A Constraint-Led Approach to Coaching and Teaching Games: Can going back to the future solve the «they need the basics before they can play a game» argument?.” Ágora para la Educación Física y el Deporte 20, no. 1 (2018): 1-26.
Schollhorn, W., Patrick Hegen, and Keith Davids. “The nonlinear nature of learning-A differential learning approach.” The Open Sports Sciences Journal 5, no. 1 (2012).
Schmidt, Richard A. “A schema theory of discrete motor skill learning.” Psychological review 82, no. 4 (1975): 225.
Shea, Charles H., Qin Lai, David L. Wright, Maarten Immink, and Charles Black. “Consistent and variable practice conditions: Effects on relative and absolute timing.” Journal of motor behavior 33, no. 2 (2001): 139-152.
Summers JJ, Anson JG. Current status of the motor program: revisited. Hum Mov Sci. 2009 Oct;28(5):566-77. doi: 10.1016/j.humov.2009.01.002. Epub 2009 Feb 23. PMID: 19230995.