Essays · Action and Control

Prospective Memory: Remembering to Act

An intention can be formed, valued and carefully planned without returning when action becomes possible. Prospective memory connects what we intend now with the time, event or context that makes it relevant later.

By Yona Ole Lobulu ·

Essay11 min readD8.5

Topic
Action and Control
Read first
3 pieces should be read before this one
Reading time
About 11 minutes of reading
Difficulty
Late reading: this page sits at the far end of the Library, after many other pieces

The question

How does an intention formed in one moment return to influence behaviour in another, and why is remembering to act only one of the conditions required for the intended action to occur?

Definition

Prospective memory is the capacity to remember that an intended action must be performed when an appropriate future time, event or context occurs.

Many intentions cannot be acted on when they are formed.

You decide to send a document after a missing figure arrives. You intend to make a call when an office opens. You plan to take something with you when you leave the house. The decision is made now, but the conditions required for action belong to a later moment.

Attention cannot remain fixed on every pending intention. Other activities have to continue. The intention must remain retrievable across a period in which it is not the main object of thought—and return when a later situation makes it relevant.

This capacity is called prospective memory: remembering that an intended action must be performed when an appropriate future condition occurs.

Its defining challenge is not simply retaining information. It is allowing an intention formed in one moment to influence behaviour in another.

Remembering information is not the same as remembering to act

Prospective memory is sometimes described as "memory for the future," but people do not remember events that have not yet happened. What they retain is an intention created in the past and meant to guide future action.

This involves two distinguishable components.

The retrospective component concerns what was intended and the information needed to perform it.

The prospective component concerns its future relevance: is this the moment when I am supposed to do it?

A person might remember perfectly well that a document needs a missing figure. That does not mean the plan to update and send it will return when the colleague's message arrives. Conversely, the person might sense that something needs to be done without recovering the complete intended action.

The components normally work together, but they can fail differently. Remembering what an intention contains and remembering that it is now due are related rather than identical achievements.

Deferred intentions can leave awareness

After an intention has been formed, attention usually returns to other demands.

The person waiting for the figure may begin another document, enter a meeting or respond to unrelated messages. The original intention is no longer in focal awareness. This does not necessarily mean that it has been lost.

A deferred intention can remain retrievable without being continuously rehearsed. The person may occasionally think about it, deliberately check for the relevant opportunity or encounter a cue that brings it back with little deliberate checking.

Different tasks and circumstances can involve different mixtures of monitoring, intermittent reactivation and cue-triggered retrieval. The intention does not have to occupy one fixed state throughout the delay.

Allowing an intention to recede is not merely a weakness of memory. It makes ongoing behaviour possible. If every unfinished intention had to remain continuously conscious, attention would be overwhelmed by actions that could not yet be performed.

The challenge is therefore not to keep every intention permanently active. It is to preserve a route through which the right intention can return under the right conditions.

Future conditions create opportunities for retrieval

An intended action usually becomes relevant under some future condition.

That condition may be an event:

When the colleague sends the figure, I will update the document.

It may be a time:

At three o'clock, I will send the document.

Or it may be a context:

When I return to the office, I will leave the document on the manager's desk.

These conditions can organise later retrieval by connecting an intention to something that may occur while attention is elsewhere.

But a future condition is only a potential retrieval opportunity. It does not mechanically return an intention to awareness.

The message can arrive without the document plan coming to mind. Three o'clock can pass unnoticed. The person can return to the office without encountering anything that restores the intended action.

For the opportunity to affect behaviour, the relevant condition must be detected or processed in a way that can reactivate the intention. This does not require deliberate conscious analysis, but it does require some effective connection between what is happening now and what was intended earlier.

A cue can be present without being detected

Several transitions are compressed when we say that something "reminded" a person.

The relevant condition must occur. The person must encounter or have access to it. It must be detected, and its relationship to the earlier intention must become available.

The colleague's message may arrive while the person is concentrating on another task. The notification may not be noticed. The message may be seen but left unopened. It may even be read without retrieving the plan to update the document.

The cue is present in every case, but it does not exert the same effect.

A cue is described as focal when processing it as part of the ongoing activity overlaps with features relevant to the intended action. Reading a message specifically about the missing figure may therefore support retrieval of the document plan. A less directly related signal may require more deliberate checking.

Focality is not simply visibility or salience. A bright notification can attract attention without restoring the relevant intention. Focality describes a relationship among the cue, the intention and the way the cue is processed.

A cue processed in a way that overlaps with the encoded intention can sometimes bring that intention to mind with little deliberate monitoring. It can still be missed. Conversely, a less focal cue can succeed when the person is actively looking for it.

Events and time organise retrieval differently

Event-based and time-based intentions often create different retrieval demands.

An event can provide an observable change in the environment: a message arrives, a person enters the room or an object becomes visible. Yet events vary in distinctiveness, predictability and relevance. Some are readily detected; others blend into ongoing activity.

Time does not always announce itself through a distinctive change. Remembering to act at three o'clock may require checking the time, maintaining readiness around the relevant period or arranging an external alert.

Time-based intentions may therefore depend more heavily on self-initiated checking when no external signal is available. This does not make them inherently harder.

A distinctive alarm can make a time-based task straightforward. An ambiguous event can make an event-based task difficult. Performance depends on the cue, the demands of the ongoing activity, the required timing and the support provided by the environment.

The two forms do not create a universal hierarchy of difficulty. They organise retrieval opportunities differently.

Ongoing activity is part of the problem

Prospective memory operates inside other behaviour.

During the delay, people read, work, travel, talk, solve problems and respond to new demands. Ongoing activity is not background noise added to an otherwise pure memory task. It is part of what prospective remembering has to overcome.

A person can deliberately monitor for a future condition, but monitoring uses attention. Repeatedly checking the time or watching for a message can slow or interrupt the activity already in progress.

An intention may also return with relatively little deliberate checking. Monitoring and cue-triggered retrieval are therefore complementary possibilities rather than exclusive explanations. Their contribution can vary with cue properties, intention encoding, ongoing-task demands and the opportunity for strategic checking.

A demanding activity can make the relevant condition less likely to be detected. It can also make the intended response harder to prioritise after the intention has returned.

Retrieval must reorganise behaviour

Remembering an intention does not perform it.

When the figure arrives, the original plan may return immediately. Yet the person is already engaged in something else. The remembered intention must now influence the organisation of behaviour.

This can involve recognising that the action is due, interrupting or reprioritising the current activity, selecting the intended response, initiating it and carrying it through. These are functional distinctions rather than a rigid sequence: several may overlap, and simple actions may compress them into a single transition.

Failure can nevertheless occur between them.

The person may remember the document but decide that the current task cannot be interrupted. They may open the document but become diverted before inserting the figure. They may update it but fail to send it.

An intention can therefore return to awareness without guiding action to completion. Prospective memory makes the intended action available when it becomes relevant; it does not settle questions of opportunity, priority, capacity or choice.

Planning can build a route back to action

Planning can improve the structure of a future intention.

"Finish the document later" leaves the retrieval condition unclear. "When the colleague sends the figure, insert it and send the document" links the action to a more specific future event.

An implementation intention makes this relationship explicit through an if–then structure:

If the figure arrives, then I will insert it and send the completed document.

Such planning can strengthen the relation between a future condition and a response. It can also reduce uncertainty about what should happen when that condition occurs.

But planning and prospective memory are not identical. A plan can specify the correct action without ensuring that it returns. The cue must still occur and be processed, and the person must still be able to initiate the response.

Under some experimental conditions, a previously specified cue–action link can also continue to influence behaviour after the intention is no longer relevant. Greater specificity is therefore a form of support, not a guarantee of appropriately timed action.

Planning can create a clearer route from opportunity to action. Prospective memory is what allows that route to become relevant later.

Environments can support remembering

Prospective remembering does not have to depend entirely on unaided internal memory.

People place objects near doors, write lists, flag messages, schedule calendar events and set alarms. In this sense, the environment can carry part of the demand.

This is a form of cognitive offloading. The person externalises part of the retrieval burden instead of relying only on internal monitoring.

A reminder must still be created and configured. It must remain available, and its signal must be noticed and interpreted. The associated intention must become accessible, and the person must still initiate the action.

External support can therefore increase reliability while introducing different dependencies. A reminder can be scheduled incorrectly. A device can be unavailable. A notification can be obscured by others or dismissed without action. Expecting an external reminder can also reduce the effort devoted to internal remembering, creating vulnerability if the support later fails.

Using reminders is not evidence that memory is weak. It is one way future-directed behaviour can be coordinated across a person and their environment.

Remembering is only one condition of acting

When an intended action does not occur, its absence does not reveal one definite cause.

The intention may have been weakly encoded. The future condition may never have occurred. The cue may have been missed or detected without retrieving the intention. The action may have become impossible, unnecessary or less important. The person may have remembered it but been unable—or unwilling—to interrupt another activity.

Importance can influence the support given to an intention. A valued action may receive clearer planning, more monitoring or stronger reminders. But importance is not itself a retrieval mechanism. People can forget actions they deeply care about.

Nor does failure to act prove deficient motivation. Remembering, wanting, choosing and being able to act are connected but distinguishable conditions.

The same boundary separates prospective-memory failure from procrastination. In the distinction relevant here, procrastination at minimum involves delaying an action that remains available as something to be done. If the intention has not returned at all, the person is not deliberately postponing it at that moment.

The processes can still interact. A remembered action can be deferred, creating a new demand to remember it later. But procrastination cannot be explained by prospective memory alone.

Completion must also be recognised

The process does not always end when the intended action is performed.

Successful coordination also requires a distinction between intentions that remain pending and those already completed. Otherwise, an old cue may continue to retrieve an action that is no longer required.

A person may send the document but later remain uncertain about whether it was sent. The unresolved status can produce unnecessary checking or, under some conditions, repetition.

These aftereffects are not inevitable. Completed intentions do not always remain unusually accessible, and unfinished intentions do not always intrude on thought. The relevant point is narrower: reliable prospective remembering includes recognising when an intention no longer needs to guide behaviour.

Coordinating discontinuous time

Prospective memory connects an intention to a later opportunity without requiring it to remain continuously conscious.

A relevant time, event or context can provide a route back, but not a guarantee. The condition must be processed, the intention retrieved and ongoing behaviour reorganised. Planning and external reminders can support these transitions, while successful completion requires the action to stop remaining pending.

The central achievement is not permanent awareness. It is the return of an intention when the world makes it relevant to action.

Prospective memory allows an intention to leave awareness without losing its route back into action.

Behind this page

The claims this essay makes, the evidence behind them, and the limits it accepts.

Evidence status

High confidence

Strongly supported, though resting on synthesis or principle rather than a single decisive body of evidence.

Claims

  1. Many intentions cannot be acted on at the moment they are formed

    Canonical inference

    What this does not assert: The problem prospective memory addresses.

  2. Remembering what an intention contains and remembering that it is now due are related rather than identical achievements

    Canonical inference

    What this does not assert: Component boundary.

  3. After formation, an intention usually leaves focal awareness as attention returns to other demands

    Established

    What this does not assert: Normal course, not failure.

  4. Leaving focal awareness does not mean an intention has been lost

    High confidence

    What this does not assert: Availability ≠ activation.

  5. A deferred intention can remain retrievable without continuous rehearsal

    Established

    What this does not assert: Central to the delay problem.

  6. Intentions may be sustained through occasional thought, deliberate checking or cue-triggered return

    High confidence

    What this does not assert: Multiple routes.

  7. Tasks and circumstances involve different mixtures of monitoring, intermittent reactivation and cue-triggered retrieval

    High confidence

    What this does not assert: Dynamic multiprocess account.

  8. An intention need not occupy one fixed state throughout the delay

    High confidence

    What this does not assert: Against static accounts.

  9. Allowing intentions to recede makes ongoing behaviour possible

    Canonical inference

    What this does not assert: Functional rather than defective.

  10. If every unfinished intention remained conscious, attention would be overwhelmed by actions that cannot yet be performed

    Canonical inference

    What this does not assert: Why receding is adaptive.

  11. The requirement is to preserve a route by which the right intention returns under the right conditions

    Canonical inference

    What this does not assert: Reframes the task.

  12. Attention cannot remain fixed on every pending intention while other activity continues

    Established

    What this does not assert: Capacity constraint; see D5.10.

  13. An intended action becomes relevant under a future condition that may be an event, a time or a context

    Established

    What this does not assert: Three condition types.

  14. Event-based intentions link an action to an occurrence in the environment

    Established

    What this does not assert: Standard task class.

  15. Time-based intentions link an action to a moment or interval

    Established

    What this does not assert: Standard task class.

  16. Context-based intentions link an action to a situation or place

    High confidence

    What this does not assert: Often treated as a variant of event-based.

  17. Future conditions organise retrieval by connecting an intention to something that may occur while attention is elsewhere

    High confidence

    What this does not assert: Mechanistic role.

  18. A future condition is only a potential retrieval opportunity and does not mechanically return an intention to awareness

    Established

    What this does not assert: Key boundary.

  19. A message can arrive, a time can pass or a place can be entered without the intention coming to mind

    Established

    What this does not assert: Everyday failure cases.

  20. The relevant condition must be detected or processed in a way capable of reactivating the intention

    Established

    What this does not assert: Detection requirement.

  21. Reactivation does not require deliberate conscious analysis of the cue

    High confidence

    What this does not assert: Spontaneous retrieval is possible.

  22. Saying that something 'reminded' a person compresses several distinct transitions

    Canonical inference

    What this does not assert: Analytic decomposition.

  23. Prospective memory is remembering that an intended action must be performed when an appropriate future condition occurs

    Established

    What this does not assert: Working definition.

  24. The condition must occur, be encountered, be detected, and its relation to the intention become available

    Canonical inference

    What this does not assert: Four transitions.

  25. A cue can be present without being noticed, noticed without being processed, or processed without retrieving the intention

    Established

    What this does not assert: Independent failure points.

  26. A cue is focal when its processing within the ongoing activity overlaps with features relevant to the intended action

    Established

    What this does not assert: Definition of focality.

  27. Focal cues generally support retrieval with less deliberate monitoring than non-focal cues

    Established

    What this does not assert: Robust but not absolute.

  28. Focality is not the same as visibility or salience

    High confidence

    What this does not assert: Common conflation.

  29. A salient notification can attract attention without restoring the relevant intention

    High confidence

    What this does not assert: Salience is insufficient.

  30. Focality describes a relationship among the cue, the intention and the way the cue is processed

    High confidence

    What this does not assert: Relational, not a cue property alone.

  31. A focal cue can still be missed, and a non-focal cue can succeed when the person is actively looking for it

    High confidence

    What this does not assert: No deterministic mapping.

  32. Events vary in distinctiveness, predictability and relevance

    Established

    What this does not assert: Not all events are equal cues.

  33. Some events are readily detected while others blend into ongoing activity

    Established

    What this does not assert: Detection varies.

  34. The defining challenge is allowing an intention formed in one moment to influence behaviour in another, not merely retaining information

    Canonical inference

    What this does not assert: Canonical thesis.

  35. Time does not always announce itself through a distinctive environmental change

    Established

    What this does not assert: Why time-based tasks differ.

  36. Time-based intentions may depend more on self-initiated checking when no external signal is available

    High confidence

    What this does not assert: Conditional, not universal.

  37. Time-based tasks are not inherently harder than event-based tasks

    High confidence

    What this does not assert: No universal difficulty ranking.

  38. A distinctive alarm can make a time-based task straightforward and an ambiguous event can make an event-based task difficult

    High confidence

    What this does not assert: Support matters more than class.

  39. Performance depends on the cue, the ongoing activity, the required timing and environmental support

    Established

    What this does not assert: Multi-factor determination.

  40. Event-based and time-based forms organise retrieval opportunities differently rather than forming a hierarchy

    Canonical inference

    What this does not assert: Canonical framing.

  41. Prospective memory operates inside other ongoing behaviour

    Established

    What this does not assert: Defining feature of the paradigm.

  42. Ongoing activity is part of what prospective remembering must overcome, not neutral background

    Canonical inference

    What this does not assert: Reframing.

  43. Deliberate monitoring for a future condition consumes attention

    Established

    What this does not assert: Monitoring has a cost.

  44. Monitoring can slow or interrupt the activity already in progress

    Established

    What this does not assert: Measured as ongoing-task cost.

  45. People do not remember events that have not yet happened; what is retained is an intention created in the past

    Canonical inference

    What this does not assert: Corrects the phrase 'memory for the future'.

  46. Ongoing-task costs are heterogeneous across tasks, cues and conditions

    Established

    What this does not assert: Systematic-review finding.

  47. Monitoring and cue-triggered retrieval are complementary possibilities rather than exclusive explanations

    High confidence

    What this does not assert: Dynamic multiprocess position.

  48. Their relative contribution varies with cue properties, encoding, ongoing-task demands and opportunity for checking

    High confidence

    What this does not assert: Conditions of variation.

  49. A demanding ongoing activity can make the relevant condition less likely to be detected

    Established

    What this does not assert: Attentional competition.

  50. A demanding activity can also make the intended response harder to prioritise after retrieval

    High confidence

    What this does not assert: Post-retrieval interference.

  51. Remembering an intention does not perform it

    Canonical inference

    What this does not assert: Retrieval ≠ action.

  52. A retrieved intention must reorganise behaviour already in progress

    High confidence

    What this does not assert: Coordination demand; see D8.2.

  53. Reorganisation can involve recognising the action is due, interrupting current activity, selecting the response, initiating it and completing it

    Canonical inference

    What this does not assert: Functional distinctions, not a rigid sequence.

  54. Several of these transitions may overlap, and simple actions may compress them into one

    Canonical inference

    What this does not assert: Against rigid staging.

  55. Failure can occur at any transition between retrieval and completion

    High confidence

    What this does not assert: Independent failure points.

  56. Prospective memory has a retrospective component concerning what was intended and how to perform it

    Established

    What this does not assert: Standard componential distinction.

  57. A person may remember an intention and decide the current task cannot be interrupted

    Canonical inference

    What this does not assert: Priority, not memory, failure.

  58. A person may begin an intended action and be diverted before finishing it

    Established

    What this does not assert: Interruption during execution.

  59. An intention can return to awareness without guiding action to completion

    Canonical inference

    What this does not assert: Central boundary.

  60. Prospective memory does not settle questions of opportunity, priority, capacity or choice

    Canonical inference

    What this does not assert: Scope limit.

  61. Vague intentions leave the retrieval condition unclear

    High confidence

    What this does not assert: Encoding quality matters.

  62. Specifying a future condition and a response links an intention to a more identifiable retrieval opportunity

    Established

    What this does not assert: Encoding specificity.

  63. Implementation intentions make the cue–action relationship explicit through an if–then structure

    Established

    What this does not assert: Definition.

  64. Implementation intentions can strengthen the relation between a future condition and a response

    Established

    What this does not assert: Effect sizes vary by domain.

  65. Planning and prospective memory are not identical

    Canonical inference

    What this does not assert: See D8.4.

  66. A plan can specify the correct action without ensuring that it returns when relevant

    Canonical inference

    What this does not assert: Specification ≠ retrieval.

  67. Prospective memory has a prospective component concerning whether this is the moment the action is due

    Established

    What this does not assert: Standard componential distinction.

  68. A specified cue–action link can persist and influence behaviour after the intention is no longer relevant

    High confidence

    What this does not assert: Commission errors under experimental conditions.

  69. Greater specificity is a form of support rather than a guarantee of appropriately timed action

    High confidence

    What this does not assert: Qualified benefit.

  70. Prospective remembering need not depend entirely on unaided internal memory

    Established

    What this does not assert: Distributed across person and environment.

  71. People use lists, placed objects, flagged messages, calendars and alarms to carry part of the retrieval demand

    Established

    What this does not assert: Everyday offloading.

  72. Cognitive offloading externalises part of the retrieval burden instead of relying on internal monitoring

    Established

    What this does not assert: Definition.

  73. Offloading intentions can improve prospective-memory performance

    Established

    What this does not assert: Demonstrated experimentally and in daily life.

  74. A reminder must be created, configured, remain available, be noticed and be interpreted

    Canonical inference

    What this does not assert: Chain of new dependencies.

  75. External support increases reliability while introducing different failure modes

    High confidence

    What this does not assert: Trade-off, not elimination of risk.

  76. Expecting an external reminder can reduce effort devoted to internal remembering

    High confidence

    What this does not assert: Creates vulnerability if support fails.

  77. Using reminders is not evidence of weak memory

    Canonical inference

    What this does not assert: Explicitly rejected framing.

  78. A person can retain the content of an intention without it returning at the relevant moment

    Established

    What this does not assert: Dissociable failure mode.

  79. A missed intended action does not identify one definite cause

    Canonical inference

    What this does not assert: Underdetermination.

  80. Failure may arise from weak encoding, an absent condition, a missed cue, changed relevance, or inability to interrupt

    Canonical inference

    What this does not assert: Alternative explanations.

  81. Importance can influence the planning, monitoring and reminders given to an intention

    High confidence

    What this does not assert: Indirect route.

  82. Importance is not itself a retrieval mechanism, and people forget actions they care about

    High confidence

    What this does not assert: Blocks motivational inference.

  83. Failure to act does not prove deficient motivation

    Canonical inference

    What this does not assert: Remembering, wanting, choosing and being able are distinguishable.

  84. Procrastination at minimum involves delaying an action that remains available as something to be done

    High confidence

    What this does not assert: Boundary definition; see D8.16.

  85. If an intention has not returned, the person is not deliberately postponing it at that moment

    Canonical inference

    What this does not assert: Prospective-memory failure ≠ procrastination.

  86. A remembered action can be deferred, creating a new demand to remember it later

    Canonical inference

    What this does not assert: The processes interact.

  87. Procrastination cannot be explained by prospective memory alone

    Canonical inference

    What this does not assert: Scope limit.

  88. Coordination requires distinguishing intentions still pending from those already completed

    Established

    What this does not assert: Completion monitoring.

  89. A person can sense that something must be done without recovering the complete intended action

    High confidence

    What this does not assert: Converse failure mode.

  90. Without completion updating, an old cue may continue to retrieve an action no longer required

    High confidence

    What this does not assert: Source of commission errors.

  91. Uncertainty about whether an action was performed can produce unnecessary checking or repetition

    High confidence

    What this does not assert: Observed under some conditions.

  92. Aftereffects of completed intentions are not inevitable and vary across conditions

    High confidence

    What this does not assert: Systematic-review qualification.

  93. Unfinished intentions do not always intrude on thought

    Contested

    What this does not assert: Qualifies classic accounts.

  94. Reliable prospective remembering includes recognising when an intention no longer needs to guide behaviour

    Canonical inference

    What this does not assert: Narrow, defensible claim.

  95. Prospective memory connects an intention to a later opportunity without requiring continuous consciousness

    Canonical inference

    What this does not assert: Summary statement.

  96. A relevant time, event or context provides a route back, not a guarantee

    Canonical inference

    What this does not assert: Summary boundary.

  97. Prospective memory allows an intention to leave awareness without losing its route back into action

    Canonical inference

    What this does not assert: Locked final line.

Where to go from here

Next published piece

Metacognition

Metacognition allows people to evaluate and regulate aspects of their cognition. It can reveal uncertainty, misunderstanding and possible error, but it does not provide complete self-knowledge or guaranteed control.

Continue through the Library →See where this sits in the graph →

Back to the Library →