Evidence-Based L&D Series: Article 3 of 8

Picture Maria, three weeks into her new role as a financial advisor.

Her manager has just assigned her first real client—a 52-year-old teacher who needs retirement planning. Maria opens the spreadsheet. Stares at it. Closes it. Opens the training manual. Scans for the formula. Tries a calculation. Gets $847,000. Recalculates. Gets $1,200,000. Both numbers can’t be right.

Her hands are shaking. The client meeting is in two hours.

This isn’t learning. This is drowning.

But wait—didn’t we just establish that difficulty is essential for learning in the last article?

In the previous piece in this series, we looked at desirable difficulties: how struggle, errors, and effortful retrieval strengthen memory. The research is clear: when learning feels too easy, it often doesn’t stick.

So here’s the apparent paradox: I’m about to argue that worked examples—which feel easier to learners than problem-solving—actually produce better results for novices. In many studies, novices who study worked examples outperform those who only solve problems, often showing 20–40% better performance on delayed tests.

So how can both be true?

The answer lies in a critical distinction that even experienced learning pros often miss:

Not all difficulty is created equal.

There’s difficulty that builds expertise. And there’s difficulty that just overwhelms. There’s productive struggle and destructive struggle.

I this post, we’ll see how to tell the difference—and why worked examples eliminate the bad kind of difficulty while amplifying the good kind.

Because right now, scenes like Maria’s play out in training programs everywhere. We hand novices complex problems and expect them to “figure it out” because that’s how we think learning works.

Practice makes perfect, right?

Not for novices.

Research shows that jumping straight to problem-solving can actually impair learning by overwhelming working memory and blocking schema construction (Sweller & Cooper, 1985). Novices don’t need more practice problems. They need better examples.

In this article, we’ll unpack the Worked Example Effect: a robust finding from Cognitive Load Theory showing that novices learn complex skills faster and retain them longer by studying expert solutions than by solving problems independently.

Even better, you’ll see how to use progressive fading—a structured way to move people from full worked examples to independent problem-solving as their expertise grows.

---

The Problem With “Just Practice”

Here’s the standard approach to skills training:

1. Explain the principles
2. Show maybe one example
3. Hand learners a stack of problems
4. Repeat until time runs out

A cybersecurity module explains threat detection, shows one phishing email, then immediately asks learners to analyze 15 suspicious emails on their own.

On paper, this looks “active” and “hands-on.” In practice, it fails novices because they face heavy intrinsic cognitive load—the inherent difficulty of the material itself.

Without existing schemas to guide them, novices have to:

• Decode what the problem is even asking
• Recall the relevant principles
• Decide which approach to use
• Execute each step correctly
• Monitor their progress
• Check whether the solution makes sense

All at the same time.

That’s a lot.

Working Memory Has Real Limits

Working memory is the brain’s “mental workspace.” It’s where conscious thinking happens. But it has severe limits.

Under simple, familiar conditions, working memory can hold about four separate elements at once (Cowan, 2001). With complex, unfamiliar material—like Maria’s retirement calculation—that effective capacity can drop to one or two elements unless the learner already has helpful schemas.

So when we hand novices a complex problem and tell them to “figure it out,” we’re often pushing their working memory past its limits. Once that happens, learning stalls. Information doesn’t get organized and doesn’t transfer into long-term memory in a useful form.

The Result?

• Random trial-and-error instead of systematic thinking
• Focus on surface features, not deep principles
• Weak schemas that don’t transfer to new situations
• Slower time to competence
• Higher frustration and lower confidence

A recent meta-analysis of worked-example studies found that students who learned only through problem-solving scored about 30% lower on delayed retention tests than those who studied worked examples (Barbieri, Clerjuste, & Chawla, 2023).

That’s not a rounding error.

Why such a big difference? To answer that, we need a closer look at how the brain handles complex information—and why “just practice” often violates the way our cognitive system actually works.

---

The Cognitive Architecture Behind Worked Examples

Cognitive Load Theory gives us a useful lens for this. Three key ideas explain why worked examples work so well for novices.

1. Working Memory Constraints

Working memory is:

• Limited in capacity
• Short in duration without rehearsal
• Easily overloaded when tasks are complex

Again, “about four elements” is a best-case estimate for simple and familiar material. For complex tasks, those “elements” quickly multiply.

When a task demands more processing than working memory can handle, the system overloads. Learners may still do things, but they’re not encoding organized, stable knowledge.

2. Schema Construction

The long-term goal of learning isn’t “remember this slide.” It’s build schemas.

Schemas are organized knowledge structures in long-term memory that:

• Chunk information (many elements → one meaningful unit)
• Guide problem-solving (“when X, use Y”)
• Support transfer across situations

For example:

• A novice looking at a phishing email sees five separate items:
  • Sender address
  • Urgent language
  • Suspicious link
  • Spelling errors
  • Credential request
• That’s five separate things to juggle.
• An expert sees one thing: “Phishing attempt”

That single schema automatically activates the right response. Same input. Completely different cognitive load.

Experts don’t have a bigger working memory. They have better-organized knowledge that compresses complexity.

3. The Worked Example Effect

Here’s the core finding:

For novices, studying complete expert solutions leads to faster learning and better retention than solving equivalent problems from scratch.

Why? It’s about where mental effort goes.

When novices solve a complex problem on their own, their effort is split:

• Search – “What should I do next?”
  • High cognitive load
  • Low learning value by itself
• Execution – “Now I plug in this number, click this menu, send this email.”
  • Medium load
  • Medium learning value
• Understanding – “Why this step? When does this approach apply?”
  • High learning value
  • Often skipped because there’s no capacity left

Worked examples change the game. They remove the need for search and lower execution effort. That frees up mental resources for understanding:

• Why each step works
• When this approach is appropriate
• What to look for next time

In other words, worked examples let novices spend their scarce mental energy on building schemas, not just guessing what to do next.

The evidence here is strong: a meta-analysis across 55 studies and 181 effect sizes found a medium to large effect size (g = 0.48) for worked examples in math learning (Barbieri et al., 2023).

---

Research Spotlight: The Classic Evidence

In a classic geometry study, Sweller & Cooper (1985) compared two groups:

• Worked Example Group:
  Studied complete, step-by-step solutions with explanations
• Problem-Solving Group:
  Solved standard practice problems on their own

Results:

• Immediate test: No big difference
• One-week delayed test: Worked example group scored about 30% higher
• Time required: Worked example group completed training in about half the time

The key twist:

This advantage is not permanent. It’s primarily a novice effect.

As learners gain experience, the same worked examples that once helped can start to get in the way. That’s the expertise reversal effect (Kalyuga et al., 2003).

Novices need guidance. Experts need room to think.

---

The Five-Stage Fading Framework

Effective worked-example instruction isn’t just “show some examples.” It’s a process of gradually fading support as learners build expertise.

Here’s a five-stage framework, inspired by Renkl, Atkinson, and colleagues (Atkinson, Renkl, & Merrill, 2003; Renkl, 2014).

---

Stage 1: Full Worked Examples (Novice)

Purpose: Build initial schemas and prevent overload.

You present complete solutions with every step visible and explained.

Workplace Example: Excel Data Analysis

Problem: Calculate quarterly sales growth.

Region A Q1 Sales: $125,000
Region A Q2 Sales: $142,000

Step 1: Calculate the change

• Formula: Q2 Sales – Q1 Sales
• Calculation: $142,000 – $125,000 = $17,000
• Why: We need the absolute difference in dollars.

Step 2: Divide by the original amount

• Formula: Change ÷ Q1 Sales
• Calculation: $17,000 ÷ $125,000 = 0.136
• Why: This gives growth as a proportion of the starting value.

Step 3: Convert to a percentage

• Formula: Proportion × 100
• Calculation: 0.136 × 100 = 13.6%
• Why: Percentages are easier to interpret and compare.

Answer: Region A grew 13.6% from Q1 to Q2.

Excel formula:
=(B2-A2)/A2*100
Where A2 = Q1 sales, B2 = Q2 sales.

Critical design elements:

• Label each step clearly
• Explain why, not just what
• Expose expert decision points
• Use dual coding (text + formulas or visuals)
• Add self-explanation prompts like: “Why divide by Q1 instead of Q2?”

When to use: First 2–4 examples when introducing a new procedure.

---

Stage 2: Completion Problems (Advanced Beginner)

Purpose: Start activating retrieval while maintaining support.

You provide most of the solution and let learners complete specific steps.

Workplace Example: Customer Service De-escalation

Scenario: An angry customer calls about a delayed shipment.

Partially worked example:

Step 1: Lower your voice and slow your speech

• Script: ✅ “I can hear you’re frustrated about this delay. Let me help you resolve this.”
• Why: Matching their volume escalates. Lowering yours slows things down.

Step 2: Acknowledge their emotion specifically

• Your turn: Write what you would say to acknowledge this customer’s frustration.
• Why: Generic “I understand” lines don’t land. Naming the emotion does.

Step 3: Take ownership

• Script: ✅ “I’m going to personally make sure we get this fixed.”
• Why: Moves from defensive to collaborative.

Step 4: Ask clarifying questions

• Your turn: Write two questions you’d ask.
  
• Why: Questions show you’re listening and gather details.
  

Step 5: Propose specific next steps

• Your turn: Based on the scenario, what will you propose?
• Why: Vague promises don’t rebuild trust. Concrete steps do.

Critical design elements:

• Remove steps learners can now handle
• Give immediate feedback
• Fade easier steps first, keep harder ones as examples
• Keep the full structure visible

When to use: After 2–4 full examples, use 3–5 completion problems.

---

Stage 3: Faded Examples (Competent)

Purpose: Increase cognitive engagement while still giving a safety net.

You show the structure and key decision points, but learners generate most of the content.

Think:

• Project parameters: given
• Framework: visible but mostly empty
• One expert example: provided as a model
• Hints: available but optional

Critical design elements:

• Provide frameworks without filling them in
• Include one complete expert example
• Offer “show hint” or “see how an expert did this” options
• Gradually move from structured to open-ended tasks

When to use: After roughly 5–8 total examples (full + completion).

---

Stage 4: Problem-Solving with Examples Available (Proficient)

Purpose: Build independence while keeping expert models handy.

Now the problem comes first. Worked examples sit in the background as optional support.

Workplace Example: Sales Objection Handling

Problem:

Customer: “Your competitor offers the same service for 30% less. Why should I pay more?”

Your response:

• Learner writes their own reply first.

Then they see a button:

Need help? See expert example.

Clicking it reveals a full expert response with reasoning.

Critical design elements:

• Always let learners try first
• Make expert models available, not mandatory
• Encourage comparison:
  
  “What’s similar in your response? What’s different?”

When to use: After about 10–12 structured examples.

---

Stage 5: Independent Problem-Solving (Expert)

Purpose: Automation and flexible application.

At this point, you drop the scaffolding. Learners tackle authentic, messy problems on their own.

Workplace Example: Leadership Coaching

Scenario:

One of your direct reports has missed deadlines for three weeks. Other team members are blocked as a result. You have a 1:1 today.

Your task:

• What will you say?
• What questions will you ask?
• What outcomes are you aiming for?
• How will you follow up?

No template. No hints. Just the learner and the problem.

Critical design elements:

• High variability across scenarios
• Interacting factors and trade-offs
• No rigid formats or scripts
• Realistic workplace context

When to use: After learners show reliable performance on scaffolded tasks.

---

Critical Implementation Principles

Principle 1: Fade Based on Performance, Not Time

Wrong:

Week 1 = full examples. Week 2 = completion. Week 3 = independent.

Right:

Move forward when learners demonstrate mastery at the current level.

Mastery indicators:

• They can explain why each step matters
• They complete steps accurately and quickly
• They can transfer the approach to slightly different problems
• They can spot when a familiar approach does not apply

A simple approach:

• After each stage, give 1–2 “check problems” at that same level.
• If accuracy is ≥80%, move on.
• If it’s <80%, stay at that stage and add more examples.

---

Principle 2: Include Self-Explanation Prompts

Learners learn much more from worked examples when you nudge them to explain the reasoning in their own words (Atkinson, Renkl, & Merrill, 2003; Renkl, 2014).

Some levels of prompts:

• Surface-level (less effective)
  
  “What is the next step?”
• Principle-based (better):
  
  “Why is this step necessary? What would happen if we skipped it?”
• Transfer-focused (best):
  
  “In what other situations would you use this same approach?”

Implementation tips:

• Add 2–3 prompts per worked example
• Aim at decision points and key principles
• Require a short written response, not just a checkbox
• Show an expert explanation afterwards so learners can compare
  

---

Principle 3: Beware the Expertise Reversal Effect

The same worked examples that supercharge novice learning can hurt experts (Kalyuga et al., 2003).

In one line:

For experts, worked examples often become redundant and annoying.

They already have strong schemas, so walking through every step:

• Wastes cognitive effort
• Splits attention between “what I already know” and “what this example shows”
• Reduces time for more challenging, meaningful practice

Think of it this way:

• For novices, examples save mental effort.
• For experts, examples cost mental effort.

Practical solution: route learners based on prior knowledge.

Quick pre-assessment (2–3 minutes):

• Score 0–40% → Novice path (Stages 1–5, heavy scaffolding)
• Score 41–70% → Intermediate path (start around Stage 2–3)
• Score 71–100% → Expert path (jump straight to Stage 5 problems)

You can also offer learner choice, but give clear guidance:

“If you’ve done X many times before, skip ahead to practice. If this is new or fuzzy, start with the examples.”

---

Principle 4: Use the Completion Strategy for Complex Skills

For especially complex skills, van Merriënboer’s 4C/ID model is your friend (van Merriënboer & Kirschner, 2018).

At its core, 4C/ID is about:

• Whole tasks that resemble real work
• Broken down into manageable parts
• Supported by just-in-time information and systematic fading

The completion strategy fits perfectly here:

1. Full worked examples for whole tasks
2. Completion assignments where learners finish partially-completed tasks
3. Conventional problems where they solve whole tasks independently

A useful twist:

• Early on, give the strategic thinking and ask learners to complete execution.
• Later, give the execution steps and ask learners for the strategy.

Example: Troubleshooting Network Issues

Early completion (strategy given, execution required):

Problem: Users can’t access SharePoint.
Diagnosis: DNS resolution issue, not general connectivity.
Evidence: Can ping IPs but not domain names.

Your task (execution):

1. What command checks current DNS settings?
2. What should the DNS server address be set to?
3. How will you verify the fix worked?

Later completion (execution given, strategy required):

Given steps:

1. Run ipconfig /all
2. Check DNS server addresses
3. If incorrect, run a command to set the right DNS

Your task (strategy):

• Why is DNS the likely culprit for this symptom?
• What other issues could look similar?
• When would this fix not be appropriate?
  

---

Principle 5: Vary Examples at Each Stage

If every example looks almost identical, learners memorize procedures, not principles.

Poor variety:

• Calculate percentage growth for:
  • Region A Q1→Q2
  • Region B Q1→Q2
  • Region C Q1→Q2

New numbers, same structure. The brain doesn’t have to generalize.

Better variety:

• Region A quarterly sales growth
• Department B annual turnover rate
• Product C year-over-year market share change

Same underlying concept (percentage change), different contexts.

Best variety:

• Calculate quarterly sales growth
• Determine if growth is statistically meaningful
• Compare growth rates across unequal time periods

Same core concept, but with increasing complexity and different demands.

Varied examples push learners to extract the deep structure, not just copy the surface pattern.

---

Common Mistakes (And How to Avoid Them)

Mistake 1: Too Few Examples Before Fading

The issue: Moving to faded examples after just one full worked example.

One example isn’t enough to build a robust schema, especially in complex domains.

Research guidance: Aim for 2–4 full worked examples before introducing completion problems (Atkinson et al., 2003).

How to decide:

• If learners can explain why each step works → 2 may be enough.
• If explanations only describe what happens → go to 3–4.
• If they can’t transfer to slightly different problems → add more varied examples.

A simple sequence:

• Examples 1–3: Full worked examples (varied)
• Examples 4–6: Completion problems
• Examples 7–9: Faded examples
• Examples 10+: Independent problems

---

Mistake 2: Fading Too Quickly

Warning signs:

• Accuracy drops below ~70% when you remove support
• Learners keep asking for more examples
• Time to complete tasks jumps up
• Errors reveal real misunderstandings, not just typos

Fix:
If performance tanks when you fade, go back one stage. Add more examples at that level before advancing.

---

Mistake 3: Examples Without Explanations

Just showing the final answer isn’t enough.

Weak version:

Problem: Calculate a 15% tip on $47.50
Solution: $7.13

Learner takeaway: “Type this into a calculator somehow.”

Stronger version:

Problem: Calculate a 15% tip on $47.50

Step 1: Convert the percentage to a decimal

• 15% = 15 ÷ 100 = 0.15
• Why: We multiply using decimals, not percentage symbols.

Step 2: Multiply the bill by the decimal

• $47.50 × 0.15 = $7.13
• Why: “15% of $47.50” literally means 0.15 × 47.50.

Tip: $7.13

Explanations direct attention to principles, not just procedures.

---

Mistake 4: Ignoring the Expertise Reversal Effect

Forcing experienced people through novice-style worked examples is a great way to:

• Waste their time
• Frustrate them
• Lower performance
• Damage the credibility of your training

Always include a quick prior-knowledge check or a clear “choose your own path” option.

---

Putting It All Together: A Complete Example

Topic: Writing professional emails to executives
Audience: New hires and junior employees (novices)
Goal: Write concise, action-oriented emails that get responses

Stage 1: Full Worked Example

Scenario: Your executive wants a status update on the Q3 campaign.

Novice email (what they often write):

Subject: Update

Hi Sarah,

I wanted to reach out and touch base with you about the Q3 marketing campaign that we’ve been working on. So far things are going pretty well overall. We’ve been making good progress on a number of fronts. The team has been working really hard and putting in a lot of effort. There have been a few small challenges here and there but nothing we can’t handle. I think we’re generally on track to hit most of our goals, though there might be some slight adjustments needed as we move forward. Let me know if you want to discuss any of this in more detail.

Thanks!
Mike

Expert email (model):

Subject: Q3 Campaign Status – On Track, One Risk

Sarah,

Q3 campaign status:
✓ Launched on time (Sept 1)
✓ 47,000 leads generated (goal: 40,000)
✗ Conversion rate at 2.1% (goal: 3.5%)

Key issue:
Low conversion suggests messaging misalignment. We’re testing revised landing page copy.

Action needed from you:
Approve $15K additional budget for landing page optimization (approval attached).

Timeline: Results from the revised copy expected Oct 15.

Mike

Why this works:

• Subject line as preview.
  “Q3 Campaign Status – On Track, One Risk” tells her what it’s about, that it’s mostly good news, and that there’s one thing to pay attention to—before she even opens it.
• Bottom line up front.
  Status bullets are visible at a glance. No hunting.
• Data over adjectives.
  “47,000 leads (goal: 40,000)” beats “making good progress.”
• Clear action request.
  “Approve $15K additional budget” beats “Let me know if you want to discuss.”
• Brevity.
  Fewer words, more clarity, less cognitive load.

You’d walk learners through each of these moves.

---

Stage 2: Completion Problem

Scenario: Your executive needs to decide whether to attend next week’s product demo.

Template:

Subject: [Your turn: write a specific subject line]

[Executive name],

[Your turn: 2–3 bullet status points with data]

Key issue:
Product demo on Tuesday conflicts with board meeting.

Action needed from you:
[Your turn: what decision do you need?]

Timeline: [Your turn: when do you need the decision?]

[Your name]

Then you show an expert version and a checklist so learners can compare.

---

Measuring Success: How to Know It’s Working

Don’t stop at “they finished the course.” Look for evidence that learning actually happened and sticks.

Useful measures:

1. Near transfer tests
   • New problems that share the same structure but different surface details.
2. Far transfer tests
   • New formats or situations that still rely on the same principles.
3. Delayed tests
   • Administered 1–2 weeks later to check durable learning, not just short-term memory.
4. Real-world performance metrics
   • Executive email response rates
   • Time to competency
   • Error rates
   • Number of help requests

Signs you’re on track:

• ≥80% accuracy on independent problems
• Strong performance on transfer tasks
• Good retention on delayed tests
• Noticeable reduction in time to competence
• Fewer “how do I…?” questions after training

---

Conclusion: The Paradox of Effort

In the last article, we talked about desirable difficulties: spacing, retrieval practice, interleaving—all the things that make learning feel harder but lead to stronger long-term memory.

That research is solid.

So, how can we also say that making early learning feel easier with worked examples is a good idea?

Here’s the key:

• Desirable difficulties help once there’s something to retrieve and apply—when learners already have some schemas in place.
• Worked examples help before that, by preventing overload and building those schemas in the first place.

Or in short:

First build the schema. Then make using it harder.

For novices, worked examples:

• Reduce destructive cognitive load
• Prevent drowning-by-problem-set
• Build initial schemas efficiently

For developing learners and experts, desirable difficulties:

• Make retrieval and application effortful
• Strengthen and refine schemas
• Support flexible transfer

Same learner. Different stage. Different needs.

Your Action Plan

• Assess prior knowledge before assigning content
• Start novices with 2–4 full worked examples that show expert thinking
• Add self-explanation prompts to deepen understanding
• Fade support gradually: full examples → completion problems → faded examples → independent practice
• Vary examples to support transfer, not memorization
• Measure transfer and retention, not just completions

Worked examples aren’t a shortcut or a cheat code. They’re a research-backed way to give novices the right kind of difficulty at the right time—and a faster, more reliable path to real expertise.

---

References

Atkinson, R. K., Renkl, A., & Merrill, M. M. (2003). Transitioning from studying examples to solving problems: Effects of self-explanation prompts and fading worked-out steps. Journal of Educational Psychology, 95(4), 774–783.

Barbieri, C. A., Clerjuste, S. N., & Chawla, K. (2023). A meta-analysis of the worked examples effect on mathematics performance. Educational Psychology Review, 35, Article 45.

Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114.

Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31.

Renkl, A. (2014). Toward an instructionally oriented theory of example-based learning. Cognitive Science, 38(1), 1–37.

Renkl, A. (2017). Instruction based on examples. In R. E. Mayer & P. A. Alexander (Eds.), Handbook of research on learning and instruction (2nd ed., pp. 325–348). Routledge.

Renkl, A., & Atkinson, R. K. (2003). Structuring the transition from example study to problem solving in cognitive skills acquisition: A cognitive load perspective. Educational Psychologist, 38(1), 15–22.

Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59–89.

Sweller, J., van Merriënboer, J. J. G., & Paas, F. (2019). Cognitive architecture and instructional design: 20 years later. Educational Psychology Review, 31(2), 261–292.

van Merriënboer, J. J. G., & Kirschner, P. A. (2018). Ten steps to complex learning: A systematic approach to four-component instructional design (3rd ed.). Routledge.

van Merriënboer, J. J. G., Clark, R. E., & de Croock, M. B. M. (2002). Blueprints for complex learning: The 4C/ID-model. Educational Technology Research and Development, 50(2), 39–64.