What counts as one?
Person, firm, node, city?
Choosing a unit determines what variation is visible.
Side 78 · Special
Scale
A study of what changes when the unit, level, size, time horizon or spatial extent changes. Scale asks whether a relationship survives aggregation, whether constraints change with size, and whether conclusions at one level travel to another.
unit→level→aggregation→scaling relation→consequence
06scale lenses
05scaling problems
05cross-level traps
78Side
The same system can be described at several levels.
Cells, organisms, organizations, markets and ecosystems each expose different regularities.
Where is the phenomenon defined?
Micro, meso, macro?
Some properties belong to individuals; others only make sense for groups or systems.
What is included?
System extent.
Boundary choice changes inputs, outputs and apparent efficiency.
Does the metric mean the same thing at each level?
Comparable or redefined?
Averages, rates and totals behave differently under aggregation.
Which level actually generates the effect?
Avoid level confusion.
Macro patterns may arise from micro interaction but require macro description.
Aggregation can create, erase or reverse patterns.
Combining units is not a neutral mathematical step.
Compress variation.
Means hide distribution, tails and subgroup structure.
Normalize by exposure.
Different denominators can reverse apparent comparisons.
Group mix matters.
Aggregate change can reflect changing composition rather than within-group change.
Trend reverses after grouping.
Aggregation across a lurking variable can produce opposite conclusions.
Group relation ≠ individual relation.
Area-level associations cannot automatically be assigned to individuals.
Individual relation ≠ group relation.
Micro-level mechanisms do not guarantee the same macro-level pattern.
Many systems change predictably with size.
Scaling relations reveal whether doubling size doubles, less-than-doubles or more-than-doubles some outcome.
Linear
Output grows proportionally.
Doubling scale roughly doubles the measured quantity.
Sublinear
Output grows more slowly than size.
Shared infrastructure can create economies of scale.
Superlinear
Output grows faster than size.
Interaction-rich systems can generate accelerating outputs and costs.
Allometry
Biological traits scale nonlinearly with body size.
Geometry and transport constraints create systematic size relationships.
Network effect
Possible interactions grow with participants.
Value or complexity can grow faster than node count under some conditions.
Spatial and temporal scale change what looks causal.
Processes that appear stable over one window can look volatile or reversed over another.
| Scale | Question | Typical consequence |
|---|---|---|
| Seconds | What is the immediate response? | Transient dynamics dominate |
| Years | What persists or adapts? | Learning, depreciation, selection appear |
| Local | What happens at one site? | Context specificity is visible |
| Regional | How do flows connect places? | Networks and spillovers matter |
| Global | What aggregates across systems? | Large-scale feedback and distribution emerge |
Scaling can be discontinuous.
Systems can work smoothly until size crosses a threshold and a new constraint becomes dominant.
More participants increase interaction burden.
Informal communication can fail beyond team-size thresholds.
Shared capacity saturates.
Delay can rise sharply near practical limits.
Interfaces multiply.
Adding components can increase relationships faster than component count.
Control systems change with organizational size.
Direct supervision often gives way to standards, metrics and delegation.
Fixed systems become worthwhile.
Large scale can justify capital-intensive shared infrastructure.
Rare events become routine somewhere in a large system.
Scale changes expected incident frequency even when individual probabilities stay constant.
Scale errors are reasoning errors.
Before transferring a conclusion, ask whether the unit, mechanism and boundary stayed the same.
Check unit
Did the analysis switch from individuals to groups or systems?
Check denominator
Did the underlying exposure or population change?
Check composition
Did subgroup mix create the aggregate movement?
Check mechanism
Does the causal process still operate at the new level?
Check threshold
Did scaling activate a new bottleneck, network effect or governance need?
ScaleGeoffrey West · biological and urban scaling
Hierarchylevels and organization
Ecological Inferencecross-level reasoning
Complexitysize, interaction and emergent behavior