The Nonlinear Nature of Learning - A Differential Learning Approach

Picture a football coach on a training pitch, with a whistle around the neck, stopping a player mid-motion. "Watch the footwork. Strike it again. Same position, same angle, same follow-through. And again." The logic feels unassailable — find the correct movement and repeat it until it sticks. Wolfgang Schöllhorn and his colleagues would tell you that this logic is not just incomplete; it is built on a physical assumption about causality that turns out to be wrong. The assumption is this: in a linear system, the same cause produces the same effect. Apply this to movement training and you get the standard pedagogical package — prescribe a single ideal technique, derived from studying elite athletes, then have learners approximate it through repetition and corrective feedback. Break complex skills into phases, train each phase separately, and reassemble them later. Schöllhorn describes these as "simplified projections of reality," useful enough as shortcuts but actively misleading when applied to bodies in motion. The fixed-target model treats the learner as a machine converging on a set point. But the body is not a machine. It is a complex system, and complex systems do not behave linearly. That distinction — between linear and nonlinear causality — is the conceptual engine of the differential learning approach. In a nonlinear system, small causes can produce big effects, and vice versa. A system in this state is sensitive to fluctuations, and those fluctuations matter. Skilled movement, Schöllhorn points out, is not the same action repeated identically. Even the world's best athletes never execute a movement the same way twice. Variability is not the enemy of skill. It is present in skilled performance at every level. The question is what you do with it. Traditional training tries to suppress variability. The differential approach does the opposite. It amplifies fluctuations deliberately, through two operating principles: no repetition and constantly changing movement tasks. In practice, learners perform no precise movement repetitions and receive no corrective interventions during training. Instead, the practice environment introduces what Schöllhorn calls stochastic perturbations — random or unpredictable variations layered around a central movement pattern. The learner's system is forced to adapt continuously, and in adapting, it self-organizes. The coordination solutions that emerge are the learner's own, not copies of an external archetype. Schöllhorn calls these functional movement patterns — stable solutions the body finds for itself when pushed into unstable territory. To test whether this principle actually works on a training pitch, Schöllhorn, Hegen, and Davids ran a pilot study with 24 semi-professional footballers from an eighth-division German club, randomly assigned to three groups. The groups were matched for age and experience — participants averaged roughly 23 to 24 years old, with around 18 to 20 years of football behind them. The intervention targeted two techniques simultaneously within each session: ball control and shooting at goal. Over four weeks of training, twice a week, there were eight sessions total, each lasting about 25 minutes. Before the intervention began, everyone took a pre-test. After the four weeks, there was a post-test, followed by a two-week break with no training. Finally, participants took a retention test. That final test is the one that matters most because performance that fades during a rest period reflects temporary acquisition, not real learning. The three groups trained differently. The classical group worked with ideal-movement archetypes in blocked practice — goal shooting in the first half of each session, ball control in the second — with repetition and corrective feedback delivered every third trial. The two differential groups trained with deliberately amplified variability: large instructed fluctuations in stance, kicking leg, arms, trunk, head position, and ball contact. No repetitions and no corrections. The differential blocked group practiced the two techniques in sequence, as the classical group did, but with the variability amplified. The differential random group practiced the same varied exercises but in a randomized order within sessions, switching between the two techniques unpredictably. The results landed clearly in the differential groups' favor. For goal shooting, the post-test means were 29.5 points for the classical group, 39.8 for the differential blocked group, and 42.8 for the differential random group. The comparison statistic — called D-emp, the empirical value from a nonparametric single-comparison test — was 24, well above the critical threshold D-crit of 5.27 needed for significance at an alpha level of 0.05. For ball reception, lower scores are better, as the measure is the summed distance between foot contact and the ball's final resting position. Post-test means were 465 centimeters for the classical group, 424 for the differential blocked group, and 416 for the differential random group. Again, the classical group was significantly apart from both differential conditions. The two differential groups, by contrast, were not reliably different from each other at the post-test. For reception, D-emp was 2, well below the threshold of 4.13. For goal shooting, it was 0. Both differential schedules beat the traditional program, and they did it comparably at acquisition. Retention is where the story gets more interesting. For goal shooting, the retention means were 30.8 for the classical group, 41 for the differential blocked group, and 45 for the differential random group. The classical group had regressed almost to its pre-test level, while the differential groups held their gains. For goal shooting specifically, the random differential group now pulled ahead of the blocked differential group — D-emp of 7, above the threshold of 4.13. The group that practiced in the most unpredictable, never-repeating sequence produced the most durable skill. That single result is worth sitting with. Never repeating the same movement, never receiving a correction, switching randomly between two techniques — and the outcome is better retention than drilling carefully structured repetitions of a prescribed form. Schöllhorn is careful to note the study's limits. Only 12 of the 24 participants completed every test and training session, which means the core learning trajectories rest on four participants per group. One outlier in the random differential group complicates the picture. The study has pilot status, explicitly. However, it fits within a broader body of prior work Schöllhorn cites, including a cycling study where a differential protocol, over four weeks, reduced maximum heart rate in a stress test from 178 beats per minute to 164, and lowered peak lactate from 7.6 millimoles per liter to 4.8. This indicates physiological efficiency, not just performance scores. The theoretical explanation Schöllhorn offers is that perturbation works by purposefully destabilizing the learner's movement system. A destabilized system requires less energy to transition to a new stable state. When the learner is never allowed to settle into a groove — because the groove keeps shifting — the system is forced to explore its own coordination space. The solutions it finds are better calibrated to the individual's actual dynamics than a movement pattern borrowed from someone else's body. That is what "functional" means in Schöllhorn's framing: not aesthetically correct, not textbook-ideal, but stable and effective for this person in these conditions. It also changes what training looks like at a practical level. A differential coach does not stop play to demonstrate the right way. The differential coach changes the constraint — different stance, different starting position, different pressure, different sequencing — and lets the body figure out the rest. Two different techniques can be trained within a single session without negative interference. Random switching between them, if anything, produces better retention than blocked practice. The detour through variation is not a detour at all. The linear instinct — find the target and repeat until you hit it — is intuitive because it maps onto how we think about most skill-building. Practice makes perfect. Ten thousand hours. Deliberate repetition. But Schöllhorn's argument, backed by this pilot data and the broader differential learning literature, is that the body does not learn like a machine converging on a fixed solution. It learns like a complex system navigating a landscape, and the way to help it navigate is to make that landscape richer, stranger, and less predictable. Variability is not the noise you train around. It is the mechanism. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.