Creating the Conditions for Learning: Key Principles

Avraham Kluger and Angelo DeNisi published a meta-analysis in Psychological Bulletin in 1996 that should have changed how schools and workplaces operate. It has not.

They gathered 131 usable papers, extracting 607 effect sizes from 12,652 participants across 23,663 observations. On average, feedback interventions improved performance at a moderate size, d = .41. The number that matters sits underneath that average: more than 38 percent of the effects were negative. Feedback made performance worse in over a third of cases.

Their explanation is specific rather than vague. Feedback moves the learner’s attention, and where it lands determines the result. Attention on the task produces improvement. Attention drawn upward toward the self, meaning the learner’s ability, standing, or worth, reduces effectiveness. Same intervention, opposite outcomes, depending on what the recipient starts thinking about.

Feedback is the single most recommended condition for learning, in classrooms and in workplace training alike. It carries a one-in-three chance of backfiring, and almost nobody teaching or managing has been told this.

That sets the tone for everything below. The conditions that reliably produce learning are not the ones that feel best, and several of the practices most widely promoted have weak or absent evidence. This guide covers what the phrase means, why comfort misleads, five conditions supported by evidence, what to stop doing, three worked examples, two templates, and a term-length plan.

Every study named here comes with its authors and journal. Where the evidence is thin, I say so.

What the Phrase Means

Conditions for learning get discussed at three separate levels, and mixing them produces the vague advice this topic is known for.

Most writing on this subject stays at the environmental level, because it is visible and easy to change. Repainting a classroom is simpler than restructuring when practice happens.

Working definition: creating the conditions for learning means arranging the cognitive and social circumstances so that effort produces durable knowledge rather than temporary familiarity.

The word doing the work there is durable. Anyone can produce a room full of people who understand something at 3 pm. The question is whether they still have it in March.

Also Read | How to Build Systems That Achieve Your Goals in 90 Days

The Comfort Trap

There is a distinction underneath everything in this article: performance during learning is a poor guide to learning itself.

Watch someone reread a chapter. Comprehension feels smooth, the material feels known, and confidence rises. Test them a fortnight later and much of it has gone. The fluency was real; it was fluency with the page in front of them rather than with the material.

Robert Bjork’s term for the alternative is desirable difficulties: conditions that slow acquisition, feel harder, produce more errors during practice, and result in better long-term retention. Spacing, self-testing, and varied practice all fit that description. They feel worse and work better.

That produces an awkward situation for anyone designing learning. The methods learners rate most highly are frequently the ones that serve them least, and satisfaction scores at the end of a training session measure comfort rather than retention. A trainer optimising for feedback forms will drift toward the least effective methods available, and be rewarded for it.

My position, stated plainly: if your learners never feel slightly frustrated, the conditions are probably wrong.

Condition One: Spacing

Nicholas Cepeda, Harold Pashler, Edward Vul, John Wixted and Doug Rohrer published a meta-analysis in Psychological Bulletin in 2006 covering 839 assessments of distributed practice, drawn from 317 experiments across 184 articles. Spreading study across separated sessions beat massing the same total time into one block.

The finding that rarely gets reported is the more useful one. The interval between study sessions interacts with how long you need to remember the material. The gap producing maximum retention grew as the retention interval grew. Short gaps suit material needed next week. Longer gaps suit material needed next year.

That converts a slogan into a decision.

These are working approximations drawn from the shape of the finding rather than fixed prescriptions, and the research is clearer about the pattern than about exact numbers for any given subject.

The practical version for a teacher: stop teaching a topic once and moving on permanently. Return to it briefly three times across the term. For a self-directed learner: four thirty-minute sessions across two weeks beat one two-hour session, using identical total time.

Condition Two: Retrieval Instead of Review

Rereading and highlighting are the two most common study methods and among the least productive. Both leave the material in front of the learner, so the work of recall never happens.

Retrieval means closing the book and producing the answer. Blank page, from memory, then check. The effort of pulling something out of memory strengthens it in a way that putting it in again does not.

Practical forms:

• Close the material and write everything you remember, then compare
• Answer questions before rereading rather than after
• Explain the concept aloud to someone with no notes in front of you
• Start each session by recalling last session’s content before adding new material

That final one costs about five minutes and combines this condition with spacing.

Learners resist this, and the resistance is informative. Retrieval reveals what you cannot do, which is uncomfortable and precisely the point. Rereading conceals it.

Also Read | 8 Money Mindset Shifts That Actually Change Behaviour

Condition Three: Feedback Aimed at the Task

Return to Kluger and DeNisi. Their Feedback Intervention Theory holds that effectiveness falls as attention moves away from the task and toward the self.

That gives a usable test for any piece of feedback: after hearing it, is the person thinking about the work or about themselves?

Praise is not exempt. “You are so clever” points at the self as firmly as criticism does, which is why it can lower subsequent performance on harder tasks. “That approach worked because you checked the units first” points at the task.

Three rules follow. Describe what the work does rather than what the person is. Give one thing to change rather than five. Deliver it close enough to the work that the learner can still act on it.

Condition Four: Managed Cognitive Load

Working memory holds very little at once. Any material that overruns it produces the appearance of teaching without the substance of learning, since nothing survives to be stored.

Three common overloads, with fixes.

• Split attention. A diagram on one page and its explanation on another forces the learner to hold one while finding the other. Put labels on the diagram.
• Redundancy. Slides read aloud verbatim while the audience reads them creates two competing streams of the same content. Speak or display, not both.
• Missing foundations. New material resting on shaky prerequisites consumes all available capacity on the basics. Check the foundation before adding the floor.

Worked examples deserve particular mention. For genuine beginners, studying a fully worked solution generally produces more learning than attempting the problem unaided, because the unaided attempt consumes capacity on searching rather than on understanding the method. That advantage shrinks and eventually reverses as expertise grows, so the useful rule is to move learners from worked examples toward independent problems as competence develops.

Also Read | The Practical Life Skills Nobody Teaches You in School

Condition Five: Safety to Be Wrong

Every condition above requires learners to expose errors. Retrieval surfaces gaps. Desirable difficulties generate mistakes by design. Feedback requires someone to hear that something did not work.

A learner who believes errors will be treated as evidence about their ability will avoid all of it. They will reread instead of self-test, stay quiet instead of asking, and choose easier tasks that protect their standing. Every one of those choices is rational given their assessment of the risk, and every one damages their learning.

This connects back to the feedback research directly. Comments that point at the self teach learners that errors are identity claims, which then makes them avoid the conditions that would help them most. The social layer and the cognitive layer are the same problem viewed from two angles.

The practical signal is simple. Count how often learners volunteer that they do not understand something. In an environment where that number is near zero, the material is either trivially easy or nobody feels safe saying it.

What Does Not Work

Some widely promoted conditions do not survive examination, and removing them frees time for the ones that do.

• Matching instruction to learning styles. Harold Pashler, Mark McDaniel, Doug Rohrer and Robert Bjork reviewed this for Psychological Science in the Public Interest in 2008. Validating the idea requires a specific experimental result: learners classified with one style should do best with matching instruction, while learners with another style should do best with different instruction. They found that this core evidence was essentially missing, and that studies using adequate designs produced negative results.
• Two clarifications, since this finding gets overstated in both directions. People do have preferences, and those preferences are real. Some material genuinely suits one presentation over another, for everyone. What lacks support is the specific claim that tailoring the format to an individual’s diagnosed style improves their learning.
• Comfort as a proxy. Pleasant surroundings are worth having. They are not conditions for learning in any measurable sense once basic adequacy is met, and effort spent there is often effort not spent on sequencing and practice.
• Satisfaction scores as evidence. They measure how a session felt. Given the comfort trap, they can move in the opposite direction to retention.

Three Worked Examples

Constructed to show the reasoning, following the research above rather than any individual’s practice.

Example 1: The Teacher Who Stopped Reteaching

A secondary teacher noticed her class performed well on end-of-topic tests and poorly on the end-of-year paper. Her response had been to reteach weak topics in June, which produced the same pattern the following year.

She changed the timetable instead. Each topic now returns for ten minutes, three times, spread across the following two terms, always beginning with recall before any re-explanation. Coverage of new content slowed by roughly a week across the year.

Lesson: the problem was retention rather than initial teaching, and reteaching in June treats the symptom.

Example 2: The Training Day Nobody Remembered

A company ran a full-day compliance session with high satisfaction scores and negligible behaviour change. Managers concluded the content needed to be more engaging.

The redesign kept the same material and changed its distribution: a ninety-minute opening session, then four fifteen-minute sessions across six weeks, each beginning with three recall questions about the previous one. Satisfaction scores fell slightly. Error rates in the audited process fell substantially.

Lesson: the scores measured the day. The audit measured the learning. They disagreed, and only one of them mattered.

Example 3: The Student Who Studied Harder Than Anyone

A student read every chapter three times, highlighted extensively, and produced colour-coded notes. His confidence going into examinations was consistently higher than his results.

The change was mechanical. Same hours, redistributed: read once, then close the book and write what he could recall, then check and mark the gaps, then return to those gaps two days later. His confidence dropped, and his marks rose.

Lesson: the earlier method produced fluency with his own notes. The new one produced the ability to retrieve without them, which is what an examination measures.

Also Read | How to Build Leverage When You’re Starting From Zero

Mistakes That Cost the Most

Two of these do most of the damage. Feedback aimed at the person carries a documented risk of making performance worse, and it is delivered constantly with good intentions. Judging by satisfaction scores steers every other decision in the wrong direction, because it rewards the comfortable methods over the effective ones.

Two Templates

Template A: The Session Structure

Works for a lesson, a training block, or a solo study hour.

First 5 minutes:   Recall from the previous session. No notes. Check after.
Next 15 minutes:   New material, one idea at a time, worked example first.
Next 20 minutes:   Practice, unaided, errors expected and visible.
Last 5 minutes:    Learners write what they now know, from memory.
Afterwards:        Schedule the return date for this content. Actually put it in the calendar.

The final line is the one that gets dropped, and it is the one that turns a good session into retained knowledge.

Template B: The Feedback Sentence

Observation:  "The conclusion appears before the supporting evidence."
Effect:       "A reader reaches the claim without a reason to accept it."
Change:       "Move the evidence above the claim in the next draft."
Check:        "What made you order it that way?"

Nothing in that structure refers to the learner. The final question exists because roughly a third of the time the answer reveals a reason worth knowing, and asking costs nothing when it does not.

A Term-Length Plan

Weeks 1 to 2: Measure what is actually retained

• Test something taught six weeks ago, unannounced, low stakes
• Compare the result against what the original assessment suggested
• Ask learners which study methods they use

Weeks 3 to 6: Change the distribution

• Add a five-minute recall opening to every session
• Schedule three return visits for each new topic before teaching it
• Replace one rereading activity with a retrieval activity

Weeks 7 to 10: Change the feedback

• Audit ten recent comments. Do they point at the work or the person?
• Give one change per piece of work rather than several
• Stop praising ability. Praise the method that worked.

Weeks 11 to 12: Check

• Repeat the delayed test on newer material
• Compare against the week 1 baseline
• Keep what moved the delayed results, drop what only moved the mood

The baseline in week one is what makes this more than an opinion at the end.

Also Read | Factors That Affect Talent Development and How to Reach Your Full Potential

Honest Limits

Effect sizes are averages. A meta-analysis reporting d = .41 describes a distribution, and Kluger and DeNisi’s central point was that the distribution contained a great deal of harm hidden inside a positive mean. Your context might sit in either tail.

Much of the cognitive research uses simple materials. Word lists and short passages under laboratory conditions are not a term of secondary chemistry. The spacing and retrieval findings replicate widely in classrooms, which is reassuring, though the precise intervals do not transfer with any precision.

Conditions are not the whole picture. A hungry child, an unstable home, an under-resourced school and a workplace that gives nobody time to practise are not problems of sequencing. Advice about spacing intervals aimed at those situations is misdirected effort, and pretending otherwise moves a structural failure onto individual teachers and learners.

Motivation still matters. None of these conditions operates on someone who has no reason to engage. The evidence base here is about making effort productive rather than about generating it.

Summary

Conditions for learning operate at environmental, cognitive, and social levels, and the strongest evidence sits in the last two. Kluger and DeNisi’s meta-analysis of 607 effect sizes found feedback improved performance on average yet made it worse in more than 38 percent of cases, with the difference turning on whether attention lands on the task or the self. Cepeda and colleagues, synthesising 839 assessments across 317 experiments, established that distributed practice beats massed practice and that the useful gap grows with how long the material must last. Retrieval beats review. Cognitive load must be managed rather than filled. All of it depends on learners feeling safe enough to expose what they do not know, and none of it feels as good as the methods it replaces.

Frequently Asked Questions

1. What are the conditions for learning?

They divide into environmental, cognitive, and social. The evidence is strongest for the cognitive conditions, meaning how material is spaced, practised and loaded, and for the social condition of whether errors are safe to show. Physical surroundings matter mainly at the extremes.

2. Does feedback always help learning?

No. Kluger and DeNisi’s meta-analysis of 607 effect sizes found more than 38 percent of feedback interventions reduced performance. Feedback directing attention to the task tends to help. Feedback directing attention to the person, including praise about ability, tends to hurt.

3. How long should I wait between study sessions?

The useful gap grows with how long you need to retain the material. Days for something needed next month, weeks for something needed next year. Cepeda and colleagues found that the interval producing maximum retention increased alongside the retention interval.

4. Why does rereading feel effective when it is not?

Rereading produces fluency with the page rather than with the material. That fluency feels like knowledge and disappears once the page does. Retrieval feels harder because it exposes gaps, which is what makes it useful.

5. Are learning styles real?

Preferences are real. The claim lacking support is that matching teaching format to a diagnosed style improves outcomes. Pashler, McDaniel, Rohrer, and Bjork found the core evidence for that interaction missing, with adequately designed studies producing negative results.

6. What is cognitive load and why does it matter?

Working memory processes a small amount at once. Material that exceeds it cannot be stored, so teaching happens without learning. Split attention, redundant presentation, and missing prerequisites are the common causes.

7. Does a comfortable environment improve learning?

Beyond basic adequacy, the effect is small compared with sequencing, practice, and feedback. Comfort during learning can even mislead, since the methods that feel smoothest often retain the least.

8. What is the difference between performance and learning?

Performance is what someone can do now, with the material available. Learning is what remains weeks later without it. The two frequently move in opposite directions, which is why methods that feel effective often are not.

9. How do I make workplace training stick?

Distribute it rather than delivering it in one day, open each part with recall of the previous one, and build in practice on real tasks. Judge it with a delayed check on behaviour rather than with satisfaction forms collected at the end.

Key Takeaways

• Feedback improved performance on average across 607 effect sizes yet worsened it in more than 38 percent of cases, depending on whether attention landed on the task or the self.
• Distributed practice beats cramming, and the gap between sessions should grow with how long the material must last.
• Retrieval strengthens memory in a way that rereading does not, and it feels worse for exactly that reason.
• Performance during a session is a poor predictor of retention weeks later.
• Satisfaction scores measure comfort, and comfort can run opposite to learning.
• Matching instruction to diagnosed learning styles lacks supporting evidence, though preferences themselves are real.
• Every effective condition requires learners to expose errors, which makes safety a precondition rather than a nicety.
• Structural problems such as hunger, instability or no time to practise are not solved by better sequencing.

Conclusion

The uncomfortable pattern running through this evidence is that the conditions producing durable learning are the ones people rate lowest while experiencing them. Spacing feels disjointed. Retrieval feels like failing. Task-focused feedback feels colder than praise. Meanwhile, the methods everyone enjoys, rereading, smooth delivery, encouraging comments about ability, perform badly on delayed measures.

Anyone designing learning therefore faces a choice between being liked in the room and being useful in six months. Those are different jobs, and the feedback you receive will mostly reward the first.

Pick one thing to change and measure it properly. Test something you taught six weeks ago, unannounced, before you change anything. The gap between what that test shows and what you expected is the real starting point, and it is usually wider than anyone wants it to be.

Also Read | How to Overcome Procrastination and Get More Done Every Day