Interruption and delay are the main triggers of activation loss

What it is

A loss-of-activation error has no single identifiable trigger, but it does have identifiable accelerating conditions — interruption and delay. An interruption forces an unrelated cognitive activity into the middle of an otherwise continuous execution sequence, and that inserted activity competes for the very same cognitive resources needed to maintain the goal. Delay simply stretches out the time window between when a goal is set and when it gets cashed out, giving decay more time to happen. Neither is a trigger in the data-driven or associative sense of "something firing" — both are cases where the cash-out that would otherwise have happened naturally gets pushed back, giving decay room to work.

Why it happens

Maintaining a not-yet-cashed-out goal requires continuously spending limited cognitive resources to fight off natural decay; a new task inserted by an interruption competes for that same resource pool — the more resource the new task demands and the longer it lasts, the less is left to maintain the original goal, and decay speeds up accordingly. Delay is a separate, independent pathway: even with no new task inserted, the sheer passage of time lowers activation strength on its own — the longer the wait, the higher the odds the goal drops away. These two pathways can stack: a long interruption that also inserts a high-load task is more likely to cause goal loss than either a short delay alone or a low-load interruption alone.

Studying it

A common way to study this relationship is to have participants, partway through a multi-step task, get asked to complete a different task before returning to the original one, manipulating the cognitive load and the duration of the inserted task, and measuring whether participants can recall and execute the original goal at the correct moment. A recurring finding in this line of work is that higher load in the inserted task raises the rate at which the original goal is forgotten; simply extending the inserted task's duration without raising its load also raises the forgetting rate, but usually by less than raising the load does — suggesting load is a more direct driver than duration, and the two shouldn't be read as interchangeable.

Where it stops holding

This relationship holds on the premise that the interruption or delay occurs before the goal has been cashed out — once a goal has already been executed, whatever interruption happens afterward can no longer cause that goal itself to be lost. Also, the load and delay manipulated in a lab are usually controlled and sharply bounded, while real-world interruptions tend to have fuzzy boundaries and hard-to-quantify load; the specific forgetting rates measured in the lab shouldn't be applied directly to any particular product scenario — only the relative relationship (load matters more than duration) can be borrowed to decide which variable to address first.

Applying it

To estimate whether a flow carries a meaningful risk of goal loss, first check whether it requires the user to cross a potentially interruptible time window between setting an intention and cashing it out; if so, prioritize shortening that window rather than counting on the user's memory. If the window can't be shortened, reduce the cognitive load an interruption is likely to impose — for example, surface all the information needed to return to the original task at once, so the user isn't left reassembling scattered context on resumption. To verify: compare goal-completion rates on the same multi-step flow under three conditions — no interruption, a low-load short interruption, and a high-load long interruption; if the high-load long-interruption condition shows a markedly lower completion rate than the other two, the flow carries a meaningful goal-loss risk and needs a targeted safeguard.

Related

• Same group: A10.12.1 Mid-sequence, the goal representation driving an action gradually decays until it is lost · A10.12.2 After a goal is lost, the action may stop or get replaced by a more salient one · A10.12.4 Keeping a goal visible or providing progress cues maintains its representation
• Nearby: A10.03 Omission vs. commission errors · A9.03 Task-switching cost
• Search terms: interruption · activation decay · working memory load