Unchanged primary-task performance does not mean load hasn't increased — resources may simply have slack left

What it is

When primary task performance itself is used as a cognitive load indicator, a common misjudgment is to equate "performance hasn't gotten worse" with "load hasn't increased." In fact, a performance measure only starts to decline once the resources it consumes approach or exceed the total available — before that threshold is reached, load can keep rising while performance stays flat.

Why it happens

People typically have more processing resources available than strictly needed for a task; this margin is called reserve capacity. As long as task demand stays within that reserve, added load gets absorbed by the slack and the performance curve stays nearly flat over that range. Only once demand exceeds the total capacity, reserve included, does performance visibly decline. Performance and load are therefore not linearly related — the relationship looks more like a curve that's flat at first and drops sharply later — and looking at performance alone during the flat stretch gives no way to tell whether load has just started rising or is already close to the critical point.

Studying it

Finding this inflection point usually requires systematically manipulating task difficulty across a graded design and recording the full trajectory of the performance measure as difficulty changes, rather than comparing just two or three conditions. A common methodological trap is setting up only a "low load" and a "high load" condition, finding no performance difference between them, and concluding that "the load difference has no effect on operation" — if both conditions happen to fall within the range the reserve capacity can absorb, that conclusion doesn't hold. The difficulty gradient needs to be extended until performance is actually observed to start declining before it can be confirmed that the resource boundary has been crossed.

Where it stops holding

This problem is especially pronounced for skilled users or simple tasks with large reserve capacity, where the range in which performance is insensitive to load is wider. In scenarios where reserve capacity is already small — novice users, or a situation with other tasks already running in parallel — this insensitive safety margin shrinks, and performance changes reflect rising load earlier.

Applying it

• Don't conclude that a newly added feature or piece of information hasn't increased load just because "speed and accuracy haven't changed" — this lag between load and performance is especially pronounced in high-frequency operations designed for skilled users.
• Supplement with other evidence (subjective reports or secondary-task performance) for cross-validation rather than relying on the performance measure alone.
• Verification: where feasible, push the task difficulty up one more notch and test again to see whether performance starts to show a decline inflection, to gauge how much margin remains between the current design and the actual reserve-capacity boundary.

Related

• Same group: A9.08.2 The secondary-task method infers the primary task's remaining resources from the performance drop on an added task · A9.08.3 The secondary task's own presence changes how the primary task is performed — the measurement interferes with what it measures · A9.08.4 The secondary-task method is sensitive when the two tasks compete for the same resource, and underestimates load when they don't
• Adjacent: A9.02 Measuring Load
• Search terms: primary task performance · reserve capacity · workload-performance curve