Partitioning the metabolic cost of human running: a task-by-task approach
Key Findings
Review of the biomechanical basis for the metabolic cost of running. Running is classically modelled as a spring-mass system; the "cost of generating force" hypothesis (Taylor et al., 1980) and Kram & Taylor (1990) — metabolic rate proportional to body weight and inversely proportional to foot-ground contact time — motivate partitioning the cost into discrete biomechanical tasks.
Quantitative Data — Distribution of Net Metabolic Cost
| Biomechanical task | Individual partitioning (early studies) | Synergistic approach (later studies) |
|---|---|---|
| Body weight support | 74% | ~80% together with forward propulsion (interactive, not separable) |
| Forward propulsion | 37% | (see above) |
| Leg swing | 20% | at most ~7%, independent of support and propulsion |
| Lateral balance | — | 2% |
| Arm swing | — | reduces cost by ~3% (net benefit) |
| Sum | 131% (paradoxical overestimate) | 89% of net cost accounted for |
Note: the individual estimates sum to 131% because the tasks are not independent; the synergistic estimates are the ones to quote.
Key Mechanistic Findings
Weight support and propulsion dominate: together ~80% of the net metabolic cost of running
Leg swing is cheap: ~7% at most, and independent of the other two tasks
Arm swing pays for itself: active arm motion reduces net cost by ~3%
Lateral balance is minor: 2%
Practical Application
~80% of the net metabolic cost of running is supporting body weight and propelling forward (Kram group's synergistic partitioning), leg swing ≤7%, lateral balance 2%, and arm swing saves ~3%. For a cyclist adding running: the cost is dominated by carrying and pushing body mass, which is why running is expensive at any pace and why body mass matters more to running than to flat cycling.
Inference, not tested: because weight support dominates, body-mass changes and vertical oscillation should affect running economy more than technique details of the swing phase — this review gives no per-kilogram or per-technique percentages.
Let the arms swing naturally; restricting them costs energy.
Limitations
From the design as the abstract reports it: a review of the Kram group's own task-by-task experiments, not a systematic review; the percentages come from experiments that altered one task or combination at a time (body weight support, aiding horizontal forces, leg-swing assist).
The early individual partitioning sums to 131%, showing the tasks are not independent; the synergistic estimates account for 89% of net cost, so part of the cost remains unassigned.
The abstract gives no participant characteristics or running speeds, so how the shares change with speed or runner level is not addressed.
How this entry was checked and changed
Content correction 2026-09-27
"Losing 2 kg yields ~8% running economy improvement but <3% cycling improvement" and "technical gains yield >2% economy improvement" -> not in the abstract; removed.
Table mixed the early individual partitioning (74/37/20%) with the later synergistic totals -> both sets now shown as the abstract presents them (synergistic: support+propulsion ~80%, leg swing ≤7%, lateral balance 2%, arm swing −3%, total 89%).
"Cadence interaction" bullet -> not in the abstract; removed. Cross-reference to REF-003 as a "weight-loss strategy" -> REF-003 is a sleep-low TT performance study; removed.
Sport line "cycling, running, sprint" -> "running".
2026-09-28: added Limitations from the abstract (Europe PMC PMID 24838747).