What Is Scaffolding in Education? A Guide for Course Designers

Most explanations of scaffolding in education are written for classroom teachers. They describe a teacher modelling a task, then working through it with the class, then stepping back. That picture is right as far as it goes, but it leans on something an online course doesn’t have: a teacher watching each learner and deciding, moment by moment, how much help they need.

A self-paced course has to make those decisions in advance. This entry explains what scaffolding means and where the term came from (it wasn’t Vygotsky), what the research and its meta-analyses found, and how to design support into a course so that it responds to learners and then gets out of their way.

What the research says

Where the term came from

Scaffolding is usually traced to Lev Vygotsky and his zone of proximal development: the idea that a learner can do more with help than alone, and that this is where teaching does its work. His ideas were the inspiration, but not the source of the word. In a history of the metaphor, C. Addison Stone writes of Vygotsky and his colleague Luria that “although the writings of these men were undoubtedly a major impetus for the metaphor, in fact neither theorist used the term” (Stone, 1998).

The first extended use was in a 1976 paper by David Wood, Jerome Bruner and Gail Ross, who studied how an adult tutor helped young children build a wooden puzzle tower. As Stone quotes them, they described a process “that enables a child or novice to solve a problem, carry out a task or achieve a goal” that would otherwise be “beyond his unassisted efforts”, with the tutor “‘controlling’ those elements of the task that are initially beyond the learner’s capacity” (Stone, 1998).

Wood and colleagues listed six things the tutor did. Van de Pol, Volman and Beishuizen’s review names them as recruitment, reduction of degrees of freedom, direction maintenance, marking critical features, frustration control and demonstration (van de Pol et al., 2010). In course-design terms: getting the learner interested in the task, simplifying it to a manageable size, keeping them heading towards the goal, pointing out what matters, keeping frustration in check, and showing how it’s done. Those six still make a useful checklist for designing support.

The three things that make support scaffolding

The word has spread so widely that it now gets used for almost any help. Van de Pol and colleagues note, quoting Puntambekar and Hübscher, that “the scaffolding construct is increasingly being used synonymously with support” (van de Pol et al., 2010). Their review of a decade of research defines scaffolding as “temporary support provided for the completion of a task that learners otherwise might not be able to complete”, and finds three characteristics common to the research:

• Contingency. The support is adapted to the learner, based on what they can currently do.
• Fading. “A teacher is fading when the level and/or the amount of support is decreased over time.”
• Transfer of responsibility. The learner takes increasing control of the task and their own learning.

That definition rules out a lot of what gets called scaffolding. A glossary that every learner reads, forever, is useful support, but it isn’t contingent and it never fades. A template that learners use for every assignment to the end of the course isn’t scaffolding either: nothing has been handed over. Some popular lists also include “adjusting to learning styles”, which isn’t scaffolding and isn’t supported by the evidence (see our entry on learning styles).

What the meta-analyses found

For such a familiar idea, the early evidence was thin. Van de Pol and colleagues found only eight studies measuring effectiveness among the 66 articles they reviewed, concluding that the results “suggest that scaffolding is effective. However, more research is needed” (van de Pol et al., 2010).

Since then, the strongest evidence has come from computer-based scaffolding, the kind built into software rather than given by a teacher, which is also the kind a self-paced course can use. Belland, Walker, Kim and Lefler pooled 144 experimental studies (333 outcomes) of computer-based scaffolding in STEM subjects, from primary school to adult education, where learners worked through problem-based curricula (Belland et al., 2017). They found:

• A consistent positive effect. Scaffolding improved cognitive outcomes by an average of g = 0.46 (a moderate effect), and the effect was significantly greater than zero across every age group and level of assessment.
• The largest effect among adult learners, and when learning was measured at the level of principles (understanding why), rather than concepts or application.
• No difference by whether the support faded. The effect didn’t vary with whether the scaffolding changed over time, or with how that change was decided. Nor did it vary with whether the support was generic or written for the specific task.

A companion analysis by the same team, a network meta-analysis of 56 studies, found the effect strong across most problem-centred teaching models and education levels, with the largest averages among college and graduate learners and the smallest in secondary school (Belland, Walker & Kim, 2017).

For online courses specifically, Doo, Bonk and Heo analysed 64 effect sizes from 18 studies of online higher education published between 2010 and 2019. Scaffolding had a large average effect on learning outcomes, g = 0.87, and the effect was largest on meta-cognitive outcomes (learners’ awareness and control of their own learning) (Doo et al., 2020). They grouped scaffolds into four types, taken from Hannafin and colleagues: conceptual (pointing to the key ideas), meta-cognitive (helping learners monitor and reflect), procedural (helping them use the course’s tools and resources) and strategic (offering other ways to tackle the task).

In online university courses, meta-cognitive scaffolding had the largest effect, g = 1.10

Average effect on learning outcomes by type of scaffolding, with 95% confidence intervals, from a meta-analysis of 64 effect sizes (18 studies, 2010–2019). Conceptual and strategic rest on five effect sizes each. · effect size (Hedges' g)

Meta-cognitive scaffolding was both the most studied and among the most effective. Procedural help, the kind most courses already have in their orientation module, had the smallest effect. Computers were the most common source of scaffolding (44 of the 64 effect sizes), with an average effect close to that of instructors.

How strong is the evidence?

We rate it moderate. The direction is consistent: two large meta-analyses and one focused on online higher education all find positive effects, and the largest effects among adult and university learners, which is the audience most online courses serve. Three limits hold the rating back.

First, “scaffolding” covers very different things in these studies, from prompts in a simulation to peer feedback, so a pooled effect says that support of this broad kind tends to help, not that any particular scaffold will. Second, the feature that defines scaffolding, fading, is the part with the least support: Belland’s meta-analysis found no extra effect when support changed over time. That doesn’t mean fading is pointless. It means the evidence for it comes more from the logic of the definition and related work on the expertise reversal effect than from scaffolding studies themselves. Third, the online meta-analysis is small: 18 studies, coded by a single researcher, English-language only, and the authors acknowledge possible publication bias (Doo et al., 2020).

How to apply it in a course

The design job is to decide what support learners get, how it responds to them, how it goes away, and how you’ll know they can do the task alone. Here’s how we’d do that in a self-paced course, using one invented example: a short course teaching new team leaders to write a one-page project brief.

1. Start from the unsupported task

Scaffolding is defined by where it ends, so start there. Write the final task first: “Write a one-page brief for a real project in your team, with a goal, scope, risks and a timeline, that your manager could approve.” That is the performance learners should manage without help, and it becomes the course’s final assessment. Our guide to backward design covers this step in full.

2. Choose the scaffolds, using Wood’s six functions

Go through the six functions and decide what will do each job in your course:

• Recruitment: open with a short scenario showing a project going wrong because nobody wrote down its scope.
• Reduction of degrees of freedom: give a template with the four headings and a word limit for each, so learners decide content, not structure.
• Marking critical features: an annotated example brief, with notes beside the lines that make the scope specific and the risks real.
• Demonstration: a short screen recording of someone drafting a brief, thinking aloud.
• Direction maintenance: a progress list showing which sections are done.
• Frustration control: hints on practice tasks, and a note that first drafts are expected to be rough.

Keep each scaffold small. Cognitive load theory is a reminder that support is itself something to process, and too much of it at once defeats the purpose.

3. Build in contingency

This is the hard part online. A recorded demonstration says the same thing to everyone. To make support respond to the learner, you have to design the branches:

• Hints on demand, or after an error. A practice question that offers a hint only after a wrong answer gives help to the learners who need it and leaves the rest alone.
• Feedback that depends on the answer. Explanations written for each wrong option, not a single “Incorrect”, tell learners what to do next. Our entry on effective feedback covers what makes feedback useful.
• A short check that routes learners. Three quick questions at the start of a module can send experienced learners past the worked example and straight to practice.
• Human feedback where it counts. An instructor or peer review of the first draft is the most contingent support a course can offer. Use it on the task that matters most, not everywhere.

4. Plan the fade

Write down how each scaffold goes away. In the project-brief course: the first practice task uses the full template plus the annotated example; the second uses the template with headings only; the third uses a blank page and a four-line reminder; the final assessment uses nothing. The same pattern works for worked examples, which can lose one solved step at a time. The expertise reversal effect explains why this matters: support that helps a novice can slow down a learner who no longer needs it.

Keep the fade simple. Belland’s finding that fading didn’t add to the effect in computer-based studies is a good reason not to build an elaborate adaptive system before you know the basic version works.

5. Hand over responsibility with reflection prompts

The last scaffold to go should be the one that turns into the learner’s own habit. Replace the template with a self-check list (“Could someone outside the team say what’s in and out of scope?”) and ask learners to rate their draft against it before submitting. Prompts like these are meta-cognitive scaffolding, the type with the strongest showing in Doo and colleagues’ online meta-analysis.

6. Check the hand-over worked

Compare how learners do on scaffolded practice with how they do on the unsupported final task. A big drop means the fade was too fast or the scaffolds were doing the thinking for them. Near-identical scores with complaints about pace suggest the support could go sooner. Regular formative assessment gives you the data for both.

We often build scaffolding in early modules that help learners build a lexicon of knowledge that is needed for later lessons, and then slowly reduce those signposts later. But we have found that it’s good to leave breadcrumbs back to the scaffolding (ie. a glossary of new terms) as some students may need to refresh their memory in later modules, and others may just like reviewing the content. Within our LMS, we also use mastery paths to reduce scaffolding as a course progresses for advanced learners, but keep the scaffolding there for those who need more repetition and support throughout a unit.

How the platforms support it

Scaffolding in a self-paced course depends on a handful of platform features. We haven’t yet published our reviews of Thinkific, LearnWorlds, Teachable and Kajabi, and features change often, so check these in your platform’s current documentation, on the plan you’re paying for:

• Hints and answer-specific feedback. Can quiz questions show a hint, and a different explanation for each wrong answer?
• Branching or conditional content. Can a quiz score send learners to different lessons, or unlock a skip route past the worked examples?
• Downloadable and optional resources. Can you attach templates and examples to a lesson, and mark a lesson as optional without breaking progress or certificates?
• Assignments with instructor or peer feedback. Can learners upload a draft and get a response from a person before the final task?
• Lesson and question analytics. Can you compare scores on scaffolded practice with the final assessment, and see where learners drop off?

If your platform can’t branch, the simple version still works: put the support in clearly labelled optional lessons, and tell learners in plain words when they can skip ahead.

Further reading

• Stone (1998), “The metaphor of scaffolding”. The clearest history of the term and a sharp critique of how loosely it’s used.
• van de Pol, Volman & Beishuizen (2010), “Scaffolding in teacher–student interaction”. The review that set out contingency, fading and transfer of responsibility. Open access.
• Belland, Walker, Kim & Lefler (2017), “Synthesizing results from empirical research on computer-based scaffolding in STEM education”. The largest meta-analysis, and the best source on what built-in support can do.
• Doo, Bonk & Heo (2020), “A meta-analysis of scaffolding effects in online learning in higher education”. The one closest to online course design, with results by scaffold type and source. Open access.

For the rest of the principles behind good course design, browse our research library.