Starting from the base or from the task? What the learning pyramid doesn't settle

Bottom-up or top-down approach: two coaches facing the same failed movement

The bottom-up approach (process-oriented) works on the functions assumed to sit beneath the movement while waiting for the movement to follow. The top-down approach (task-oriented) trains the target activity directly. In children with coordination difficulties, the task clearly wins, and the 2019 international recommendations advise training any targeted function inside the task as well.

Hello to you, the movement professional,

Change of direction, plant on the left leg: at the moment the athlete looks for the incoming ball, his knee drifts inward. On the right, nothing. Two coaches watch the video in slow motion.

The first shakes his head: "He doesn't have the basics." He sees a plant that gives way and wants to send the athlete back to a few weeks of foundational work, on one foot, eyes closed. The other shrugs: a change of direction is learned by doing it, and he puts the athlete straight back into the movement, ball and defender included.

The one who talks about basics reasons like the learning pyramid of Williams and Shellenberger: what cracks at the top may rest on a tier below. The second has on his side most of the intervention trials run since 1995 in children with coordination difficulties, where people improve mostly on what they practice. Each therefore holds a piece of the truth. Occupational therapists have put these two reflexes to the test in children, and their results don't let either coach win alone.

Starting from the base: the bottom-up approach

Starting from the base is betting that a failed movement betrays a fragile function underneath, and that by strengthening this function the movement will follow without being trained. We speak of a bottom-up approach, or a process-oriented approach. In 2001, Angela Mandich, Helene Polatajko and their colleagues tied it to neuromaturational and hierarchical theories of development: the underlying functions were brought back up to level in the hope of then seeing motor skills progress.

This is where Ayres sensory integration belongs, which offers the child, through play, organized vestibular, tactile and proprioceptive experiences. The family also includes kinesthetic sensitivity training, and any isolated work on a bodily function, balance for example.

Its logic is that of the pyramid: when the sensory foundation of the pyramid holds poorly, everything built above it costs more. It fits the picture of a hierarchy of motor control where the top works better when the bottom has done its job. What remains to be seen is whether the balance gained with eyes closed will carry over by itself into the change of direction.

Starting from the task: the top-down approach, from CO-OP to Neuromotor Task Training

The top-down approach takes the problem from the other end: you train the activity that poses difficulty, in its real conditions, and the functions organize themselves around it. Mandich and her colleagues tied it to human movement science, for which skill emerges from the interaction between the child, the task and the environment. You recognize the organism, task and environment constraints that you already manipulate when you shrink a playing area.

CO-OP (Cognitive Orientation to daily Occupational Performance) was developed in Canada by Helene Polatajko's team during the 1990s. The child chooses goals, riding a bike or writing more legibly, then learns a four-step strategy: goal, plan, action, check (Goal-Plan-Do-Check). The professional guides through questions, and the method's fidelity checklist sums it up in a formula worth posting in every locker room: "Ask don't tell". One of its criteria even requires that no exercise to bring underlying functions back up to level be observed during a session.

Neuromotor Task Training (NTT) comes from the Netherlands. In 2007, Niemeijer, Smits-Engelsman and Schoemaker evaluated it in children aged 7 on average: twenty-six followed nine weekly sessions of thirty minutes. Thirteen others, recruited separately and without randomization, received nothing. Only the trained group improved on the motor tests, and above all on tasks close to those practiced. People improve on what they practice, hardly beyond: the strength of the task approach already contains its limit.

Task-oriented or process-oriented approach: what the meta-analyses say

In 2013, Bouwien Smits-Engelsman and her colleagues gathered twenty-six intervention studies published between 1995 and 2011 in children with coordination difficulties. The task-oriented approach obtained a strong effect (weighted d of 0.89), standard physiotherapy and occupational therapy a close effect (0.83), the process-oriented approach a weak effect (0.12). For reference, a d of 0.8 is commonly read as strong and 0.2 as weak. The team advised against process-oriented approaches for improving these children's motor skills.

The 2018 update, on thirty studies published from 2012 to February 2017, nuances the picture. The overall effect is large (d = 1.06) and very scattered. Benefits appear for activity-oriented approaches, but also for approaches centered on a bodily function when they are combined with activities. Working on a bodily function can therefore pay off, provided it is paired with activities.

Iona Novak and Ingrid Honan widened the focus in 2019 to pediatric occupational therapy: 129 articles, 135 indications ranked by traffic light. Among motor interventions, the seven indications with negative, weak or strong evidence all came from bottom-up approaches, and the eight rated green all from top-down approaches. CO-OP was among the green lights.

A 2026 meta-analysis tempers the enthusiasm. Zengin Yazıcı and her colleagues, in Istanbul, gather thirteen randomized trials conducted in children with disabilities. Against a waiting list, CO-OP seems to improve the goals chosen by the child, with low certainty; against other active approaches, it shows no clear superiority, with very low certainty. Starting from the task helps, and the label of the method perhaps weighs less than we think.

Sensory integration or task-oriented approach: a debate that depends on the population

On the base side, the most discussed file remains Ayres sensory integration. In 2019, Sarah Schoen and her colleagues screened the studies from 2006 to 2017 in autistic children. Of nineteen studies, three were ultimately evaluated, including two randomized controlled trials that met 100% and 85% of the criteria of the Council for Exceptional Children. Their conclusion: from 4 to 12 years, this approach can be considered an evidence-based practice.

Bradley S. Stevenson challenged this reading the same year. According to the response from Schoen's team, he questioned the methodological soundness of one of the two trials, yet this framework requires two solid randomized trials and at least sixty participants in total. The team maintains its conclusion.

In 2012, the American Academy of Pediatrics judged research on sensory integration, taken in the broad sense, "limited and inconclusive". It accepted the approach as a possible component of a wider plan, after a trial period. For the motor skills of children with coordination difficulties, the 2019 international recommendations, coordinated by Rainer Blank, still hold the evidence to be inconclusive.

The two files therefore concern different children and different goals. To improve a specific motor skill, the task clearly wins; for other goals, the debate remains open among serious researchers. Before choosing your side, ask yourself what for, and in whom.

Comparison table of the bottom-up and top-down approaches

LabO RNP synthesis based on the cited sources; the last two rows reflect our field reading.

What this changes for the learning pyramid

Read as a staircase, the pyramid would require validating each tier before climbing, and no study supports this forced march. It keeps all its value for reading: when the top stagnates, it helps you formulate the hypothesis of a fragile tier.

For acting, the text by Blank and her colleagues gives the most useful formula for the field. Since we still don't know how transfer operates, the authors recommend activity-oriented approaches that stay close to the activities to be improved. When a bodily function is targeted, they advise training it inside those activities as well. The task stays at the center, and the base is worked in a targeted way before being handed straight back to the movement.

A still fragile sensorimotor development, or hand-eye coordination that drops out at the perceptual-motor tier, can justify targeted work. In our reading, this work returns into the task within the same session, and it is by the movement that its value is judged.

Combining the bottom-up and top-down approaches in the field

Take the athlete from the beginning again. What follows describes a LabO RNP reading logic, with no value as a validated protocol; the content of the session remains your business, to be built for that athlete.

Everything starts from the real movement. You observe the change of direction in its game conditions, ball included, and you film. Note what happens when the gaze goes looking for the ball: the knee's trajectory, the plant time, the stability of the head. This observation serves as your test.

It gives rise to a hypothesis, only one at a time. If the knee drifts only when the head turns, the lead of gaze and postural adjustment with a moving head comes before that of strength. The vestibulo-ocular reflex that keeps the gaze stable comes into play as soon as the head turns, and the article on dynamic balance explains why standing still on one foot prepares poorly for the field. The targeted work focuses on this one function, designed from the start to return to the movement.

The base then returns to the movement: same change of direction, same conditions, same camera. That is your retest, in the logic of the test, stimulation, retest that we teach in Neuro-Postural Reprogramming. When you vary the task afterward (where the ball comes from, who closes the angle), you take up the spirit of the constraints-led approach. You can also ask a question rather than give an instruction, as CO-OP prescribes: "What changed when the ball arrived later?"

It is the retest that decides. If the movement improves, the hypothesis holds and the targeted work keeps its place, always handed back to the task; if it doesn't budge, the fragile tier lies elsewhere, or the task alone is enough. A gain observed the same day remains an immediate performance, distinct from learning. Only a retest after some delay, then on unfamiliar ground, tells you about the transfer of a skill.

In a child, if a base seems to give way to the point of weighing on school or daily life, refer to the physician who follows the child, an occupational therapist or a psychomotor therapist.

The blind spots of the debate between the base and the task

The figures in this article almost all come from children, with coordination difficulties or autism, and their transfer to the athlete remains a transposition that nobody has tested as is. The categories also overlap. In 2006, a pilot study on NTT associated the best progress with the physiotherapists who gave cues, asked questions and explained the why of the movement. On nineteen children, it is a cue and nothing more, but it suggests that how you teach perhaps weighs as much as the label of the approach.

An average like 0.12 aggregates different studies and methods, and it can hide a method that is useful for a specific profile. Our logic puts itself to the test: in an athlete whose movement degrades mostly with a moving head, combining targeted work and task should improve the delayed retest more than the task alone. If the two are equal, the targeted work was superfluous for him.

To place this debate within the whole of the five tiers, go back to the learning pyramid and its limits.

Work the base as much as needed, then hand it straight back to the task: it is in the movement that it will be judged.

👉 I want to understand why, not just what.

Frequently asked questions about the bottom-up and top-down approaches

What is the difference between a bottom-up and a top-down approach?

The bottom-up approach, called ascending or process-oriented, strengthens the functions assumed to sit beneath a movement, such as balance, while waiting for the movement to improve. The top-down approach, called descending or task-oriented, trains the target activity directly. In children with coordination difficulties, the meta-analyses give a clear advantage to the latter for motor skills.

What is the CO-OP approach?

CO-OP (Cognitive Orientation to daily Occupational Performance) is a task-oriented, cognitive, child-centered approach developed in Canada by Helene Polatajko's team. The child chooses goals, learns the strategy goal, plan, action, check, and finds solutions thanks to the professional's questions. Novak and Honan (2019) rate it green.

What is Neuromotor Task Training?

A Dutch task-oriented approach. In the non-randomized controlled trial by Niemeijer, Smits-Engelsman and Schoemaker (2007), nine weekly sessions of thirty minutes improved the motor skills of children with coordination difficulties, especially on tasks close to those trained. The 2019 international recommendations cite it among the recommended approaches.

Is Ayres sensory integration validated?

It all depends on the population. For autistic children aged 4 to 12, the review by Schoen et al. (2019) considers it an evidence-based practice, a conclusion challenged by Stevenson the same year. For improving the motor skills of children with coordination difficulties, the 2019 international recommendations judge the evidence inconclusive.

Does the learning pyramid require starting with the base?

No study shows that each tier must be consolidated before climbing. The pyramid of Taylor and Trott, taken up by Williams and Shellenberger, helps spot a fragile tier. The evidence invites keeping the task at the center and training the base in a targeted way, then straight away in the task.

How do you combine the base and the task in a session?

Keep the real movement as your test, draw from your observation a single hypothesis about the function that seems to give way, work it in a targeted way, then hand it back to the movement under the same conditions to retest it. Only a delayed retest tells you whether this targeted work deserves its place. This is a LabO RNP logic, without formal validation, and each session is built for the athlete.

And if the difficulty seems to come from a lower tier in a child?

The movement professional observes and works on movement; the assessment belongs to the physician who follows the child, the occupational therapist or the psychomotor therapist, to whom he refers. LabO RNP is neither medical nor paramedical, and comes as a complement to a care pathway, never in its place.

Sources

Mandich A.D., Polatajko H.J., Macnab J.J., Miller L.T. (2001). Treatment of children with developmental coordination disorder: what is the evidence? Physical and Occupational Therapy in Pediatrics, 20(2-3), 51-68.

Polatajko H.J., Mandich A.D., Miller L.T., Macnab J.J. (2001). Cognitive orientation to daily occupational performance (CO-OP): part II, the evidence. Physical and Occupational Therapy in Pediatrics, 20(2-3), 83-106.

McEwen S., Polatajko H., Wolf T., Baum C. (2012, revised by the CO-OP Academy in 2015 and 2018). CO-OP Fidelity Checklist (CFC), version "As Approved March 19 2018". CO-OP Academy.

Niemeijer A.S., Schoemaker M.M., Smits-Engelsman B.C.M. (2006). Are teaching principles associated with improved motor performance in children with developmental coordination disorder? A pilot study. Physical Therapy, 86(9), 1221-1230.

Niemeijer A.S., Smits-Engelsman B.C.M., Schoemaker M.M. (2007). Neuromotor task training for children with developmental coordination disorder: a controlled trial. Developmental Medicine and Child Neurology, 49(6), 406-411.

Smits-Engelsman B.C.M. et al. (2013). Efficacy of interventions to improve motor performance in children with developmental coordination disorder: a combined systematic review and meta-analysis. Developmental Medicine and Child Neurology, 55(3), 229-237.

Smits-Engelsman B. et al. (2018). Evaluating the evidence for motor-based interventions in developmental coordination disorder: a systematic review and meta-analysis. Research in Developmental Disabilities, 74, 72-102.

Blank R. et al. (2019). International clinical practice recommendations on the definition, diagnosis, assessment, intervention, and psychosocial aspects of developmental coordination disorder. Developmental Medicine and Child Neurology, 61(3), 242-285.

Novak I., Honan I. (2019). Effectiveness of paediatric occupational therapy for children with disabilities: a systematic review. Australian Occupational Therapy Journal, 66(3), 258-273.

Zengin Yazıcı G., Yıldırım K., Tanrıverdi M. (2026). Effects of the Cognitive Orientation to Daily Occupational Performance approach on occupational performance and related outcomes in children with disabilities: a systematic review and meta-analysis. Australian Occupational Therapy Journal, 73(5), e70129.

Schoen S.A. et al. (2019). A systematic review of Ayres Sensory Integration intervention for children with autism. Autism Research, 12(1), 6-19.

Stevenson B.S. (2019). Response to "Application of the Council for Exceptional Children's Standards for Evidence-Based Practices in Special Education" by Schoen et al. [2019]. Autism Research, 12(8), 1152-1153.

Schoen S.A. et al. (2019). Ayres Sensory Integration meets criteria for an evidence-based practice: a response to Stevenson [2019]. Autism Research, 12(8), 1154-1155.

Zimmer M., Desch L., Section on Complementary and Integrative Medicine, Council on Children with Disabilities, American Academy of Pediatrics (2012). Sensory integration therapies for children with developmental and behavioral disorders. Pediatrics, 129(6), 1186-1189.

By the LabO RNP team