Switch cost is higher when tasks are similar

What it is

The more two tasks share — the same set of response keys, the same stimulus features, differing only in the judgment criterion applied to them — the higher the switch cost between them tends to be, not lower. Intuition says similar tasks should be easier to pick up from one another, but task-switching research shows the opposite direction: similarity produces crosstalk between task-sets, where the previous task's rules compete for judgments and responses that the new task should own exclusively. This runs counter to the intuition that overlapping content makes recovery easier, and it is a different mechanism from remembering where an interrupted task left off — what's competing here is which rule set governs behavior right now, not information about past progress.

Why it happens

When two tasks use the same stimulus features or the same response keys, each task's rule set produces a candidate "what to respond" answer to the same input at the same time. Even when only one rule set is supposed to run, the unselected one still gets partially activated and competes with the correct one for output. The more the rules overlap, the fiercer this competition, and extra inhibition is needed to suppress the rule set that shouldn't fire — the time spent on that suppression is exactly what shows up as elevated switch cost. Conversely, when two tasks share no overlap in stimulus type or response mode (say, one uses a visual judgment and a keypress, the other uses spoken number naming), each rule set occupies its own input-output channel and barely competes at all, so the pure reconfiguration cost sits closer to its floor.

Studying it

A common manipulation varies how many stimulus or response dimensions two tasks share: at one end, fully non-overlapping task pairs; at the other, pairs sharing stimulus features or response keys (bivalent stimuli, where the same stimulus admits an answer under either rule set). Switch cost is then compared across overlap levels.

Common independent variables: number of overlapping stimulus/response dimensions, the specific type of overlap (shared keys vs. shared stimulus features). Common dependent variables: switch cost (RT/error difference between switch and repeat trials), error type (whether the response given matched what the previous task would have required).

This manipulation is often used to evaluate designs where the same control carries different meanings across modes — for example, one key triggering different actions in two working modes; higher overlap tends to raise the error rate on mode switches.

Methodological note: "similarity" here specifically means overlap at the level of task rules (stimulus-response mappings), not surface topical similarity. Two tasks that feel thematically close but have entirely different stimulus-response structures will not show this effect.

Where it stops holding

• The effect mainly shows up when two tasks are actually interleaved in time; if the two tasks are never switched between in close succession, overlap alone doesn't translate into extra cost.
• Overlap-driven cost can be weakened by physically separating the two rule sets — different keys, different input channels — even while the content theme stays similar.
• This entry concerns competition between two rule sets active at the same moment; it does not cover how a task's activation lingers over time after it ends, which is a different mechanism.

Applying it

• In interfaces requiring users to move back and forth between two functional modes, avoid letting the same control or key set carry different operational meanings across modes; less overlap means fewer mis-hits and less delay.
• When two modes must genuinely share a control (e.g., the same gesture means different things in edit mode versus browse mode), use a clear visual state indicator so users can consciously suppress the previous mode's habitual response.
• Verification: track the rate of "performed the action the previous mode required" errors in the first few actions right after a mode switch; if these errors cluster around high-overlap controls, crosstalk is occurring and the controls or visual states need further separation.

Related

• Same group: A9.03.1 Switching tasks requires reloading rules and goals · A9.03.3 Apparent parallel processing is mostly rapid switching · A9.03.4 Switch cost has an active goal-reset component and a passive carryover-interference component · A9.03.5 Advance warning of a switch only partly reduces the cost, never eliminating it · A9.03.6 Switching from a simple to a complex task and the reverse carry asymmetric costs · A9.03.7 Frequent small switches can accumulate more total cost than a few large ones
• Nearby: A5.08 Interruption cost and task resumption
• Search terms: task-set crosstalk · bivalent stimuli · switch cost · task-set competition