Generalisation Determines Real-World Change

A behaviour can be available where it was learned and difficult to access where it is needed.

Someone may express a boundary clearly with a supportive friend but struggle to do it with an authority figure. They may deliver a presentation fluently while alone and lose access to the same level of performance when an audience is present.

These differences do not necessarily mean the earlier capability was false. Something can be genuinely learned without yet becoming reliably available across the conditions where it must function.

This is the problem of generalisation.

Acquisition establishes that something was learned. Retention establishes that some of the learning lasted. Generalisation determines whether it reaches the situations in which it must become useful.

Learning Can Become Tied to the Conditions That Support It

Learning occurs under particular conditions.

A behaviour is acquired while certain cues are present, particular demands are being made and specific forms of support are available. The physical setting may matter, but so may the person standing in front of us, the instructions we receive, the pressure we feel or the consequences we expect.

Some of these conditions can become functionally involved in retrieving or performing what was learned.

Experiments on context-dependent memory have shown that recall can improve when learning and retrieval occur under matching conditions. But context effects are selective. Not every detail of the original environment becomes necessary, and changing the setting does not automatically make learning disappear.

What matters is whether a feature of the situation became involved in encoding, retrieving or executing the response.

Someone who learns a task with constant prompts may partly learn to respond to those prompts. Someone who practises only in private may become capable under low social evaluation without yet becoming equally capable under scrutiny.

The learning may be real while its current availability remains conditional.

Retention Is Not Generalisation

Retention asks whether learning remains after time has passed.

Generalisation asks whether a relevant benefit remains when something important about the situation changes.

A person may remember every step of a behaviour and reproduce it days later in the original setting. That is evidence of retention. It does not yet establish that the behaviour will remain available with different cues, different demands or less support.

Motor-learning research separates retention tests from transfer tests for precisely this reason. A skill can persist while transferring unevenly to related tasks or altered performance conditions. Different forms of transfer can even respond differently to the same training.

This protects against two opposite mistakes.

The first is assuming that retained learning must work elsewhere.

The second is assuming that difficulty elsewhere proves nothing was learned.

A capability may be available across a narrower range than its intended purpose requires. The size of the gap depends on what changed between practice and use.

The Distance Between Practice and Use Matters

Some transfer demands remain close to the original learning. Others require the capability to travel much farther.

Near transfer involves a new situation or task that shares important cues, demands or processes with training. Those shared elements can make transfer more plausible, but they do not guarantee it.

Far transfer concerns an outcome more substantially removed from what was practised. Applying a principle in an unfamiliar relationship is more distant than repeating the same response with a different person. Expecting a specialised training exercise to improve a broad life outcome requires a larger transfer.

Cognitive-training research makes the distinction visible. People may improve on exercises they repeatedly perform without showing equivalent improvement on broader, untrained abilities. Some studies report benefits beyond the trained task; others find little or no near or far transfer.

These mixed findings do not show that far transfer is impossible. They show that broad benefit depends on the intervention, the outcome and how transfer is measured. It must be demonstrated directly.

If someone becomes better at a digital attention exercise, the real-world question may be whether they can remain engaged during work, study or conversation. Improvement on the exercise cannot answer that question by itself.

Generalisation must be demonstrated at the distance that matters.

Generalisation Does Not Mean Doing the Same Thing Everywhere

It is tempting to think that generalisation means reproducing one behaviour identically across situations.

Real environments do not always call for identical action.

A boundary might be expressed through direct refusal in one situation, a delayed answer in another and physical distance in a third. The visible behaviours differ, yet each may reflect the capability to recognise and protect a limit.

A writing routine that works as a long morning session at home may need to become a shorter protected interval during travel. A movement learned in rehabilitation may require adjustment when it is used on uneven ground or under fatigue.

If the surface form had to remain unchanged, appropriate adaptation would look like failure.

Functional generalisation asks:

Does the learned capability still produce an effective and appropriate response under the new conditions?

Not every different behaviour qualifies. The adapted response must remain traceably connected to what was learned, preserve the relevant function and fit the demands of the new situation.

Generalisation can involve stable reproduction, appropriate adaptation or both.

Variation Can Help, but It Does Not Guarantee Transfer

One way to challenge narrow learning is to practise across relevant variation.

The principle is straightforward: variation can expose the learner to changes they may later encounter and create opportunities to discover what can change without losing the capability's function.

Research provides conditional support. Motor-learning studies sometimes find that variable practice improves later transfer. Fear-extinction research has also found that learning across multiple contexts can reduce the return of fear in a new setting. Other studies find limited, unstable or task-specific benefits.

Variation creates opportunities. It does not guarantee transfer.

Relevance matters more than variety for its own sake.

Practising a presentation in three empty rooms adds environmental variation while leaving social evaluation untouched. Practising with different audiences, interruptions or question formats may better represent the demand that later performance requires.

There is no universal number of contexts that produces generalisation. The useful question is whether practice varies along the dimensions that real use will change.

Failure to Transfer Reveals the Current Boundary of Learning

When performance deteriorates under changed conditions, the quickest conclusion is often that the change was not real.

That may be premature.

A cue may be missing. The task may now require adaptation. Pressure may interfere with retrieval or execution. Guidance available during practice may no longer be present.

Stress research shows that acute pressure can impair memory retrieval. Physiological-state experiments also suggest that internal conditions can sometimes influence access to learned information. These findings do not prove that every behavioural capability becomes state-dependent. They show why performance demonstrated under calm conditions cannot always be assumed to remain equally accessible under stress.

One difficult performance cannot establish the complete boundary of learning. Even repeated context-linked differences do not prove one mechanism; fatigue, capacity, task demands and measurement conditions may also contribute.

But a recurring pattern of success under one condition and difficulty under another provides evidence worth investigating.

It reveals where the learning currently appears to travel—and where access or adaptation may still be limited.

Real-World Change Must Work Where It Matters

Generalisation does not require a behaviour to work in every imaginable situation.

The required range depends on the behaviour's purpose. A presentation skill need not survive every possible audience. A boundary need not look identical in every relationship. A regulation capability need not take the same form under every degree of activation to remain functionally present.

The relevant questions are:

Where is the capability already available?

What changes when performance becomes difficult?

Does the response need to be reproduced or adapted?

What outcome would demonstrate that the learning transfers far enough?

These questions replace the binary idea that change is either present or absent. They reveal its current range.

Acquisition tells us that learning occurred. Retention tells us that it lasted. Generalisation tells us whether the acquired and retained learning reaches the conditions where it must function.

Generalisation does not replace the other requirements of change. It determines whether they extend into real-world use.

Change becomes real-world change when it can leave the conditions in which it was learned without losing what made it useful.