Problem-Based Learning

Abstract

Problem-based learning is a form of learning in which students acquire knowledge by working through open-ended problems in small groups before formal instruction, devised for medical education at McMaster University in the late 1960s. Its cognitive rationale is that confronting a problem activates prior knowledge, provokes elaboration, and encodes new material in the context where it will later be used. The method is vigorously contested: critics argue from cognitive load theory that novices need explicit guidance rather than discovery, while defenders reply that problem-based learning is heavily scaffolded, not minimally guided. Meta-analyses find modest advantages for skills and the application of knowledge and little or no advantage for retaining basic facts. This article surveys the tutorial process, the cognitive basis, the guidance debate, and the evidence, with three interactive demonstrations.

Keywords: problem-based learning, self-directed learning, cognitive load, scaffolding, medical education

Problem-based learning reverses the sequence most teaching takes for granted. Instead of presenting a body of knowledge and then setting problems that apply it, the method opens with a problem the student cannot yet solve and treats the resulting gaps as the curriculum (Schmidt, 1983). Students meet in small groups, work out what they need to learn in order to make sense of the case, study those questions on their own, and return to apply what they found. The tutor does not lecture but keeps the group's reasoning moving. The design is a deliberate wager that knowledge sought in order to resolve a concrete problem is understood more deeply, and retrieved more readily in similar situations, than knowledge received and stored for a later exam (Norman & Schmidt, 1992).

What Problem-Based Learning Is

Problem-based learning is an instructional method in which learning is organized around the investigation of problems rather than the coverage of a syllabus. A problem, typically an ill-structured case with no single obvious answer, is presented at the outset, and students work in small collaborative groups to define what the problem requires, to identify what they do not yet understand, and to direct their own study toward those gaps (Barrows, 1986). The defining feature is order: the problem precedes instruction and motivates it, so that the discipline's content is encountered as the answer to a question the student has already felt the force of.

The method was developed for the medical curriculum at McMaster University in the late 1960s, where the founders judged that students who had memorized basic science often could not deploy it at the bedside (Schmidt, 1983). It has since been adopted across medicine, nursing, dentistry, engineering, law, and secondary education, in forms that vary widely in how much structure the problem carries and how much direction the tutor supplies. Barrows captured that variation in a taxonomy that ranges from lecture-based cases at one extreme to fully student-directed, closed-loop problem work at the other, noting that the label “problem-based learning” covers a family of designs rather than a single procedure (Barrows, 1986). MeSH classifies the descriptor as a type of learning and defines it as the instructional use of cases to teach problem-solving and critical thinking.

The Tutorial Process

The engine of problem-based learning is the small-group tutorial, and its cycle is remarkably consistent across programs. Presented with a case, the group first clarifies unfamiliar terms and defines the problem; members then brainstorm using what they already know, which surfaces and activates prior knowledge; the discussion exposes gaps, which the group formulates as explicit learning issues; each student studies those issues independently; and the group reconvenes to pool and apply what was found, testing the new knowledge against the problem (Schmidt, 1983). This progression from problem to self-study to synthesis is what distinguishes the method from a seminar or a case discussion that merely applies material already taught.

Two features of the tutorial carry most of its intended effect. The first is that the group must generate its own learning issues; because the questions arise from the students' own sense of what they cannot yet explain, the subsequent study is self-directed rather than assigned (Loyens, Magda, & Rikers, 2008). The second is the tutor, who functions as a facilitator of reasoning rather than a source of answers, prompting the group to justify claims and to notice contradictions. Later process research argues that the method works less through the mystique of discovery than through a chain of ordinary cognitive events the tutorial reliably sets in motion: activation of prior knowledge, elaboration during discussion, and the sustained situational interest a good problem creates (Schmidt, Rotgans, & Yew, 2011).

The Cognitive Rationale

The psychological case for problem-based learning rests on well-established principles of memory and comprehension. New information is understood and retained to the degree that it can be related to what a learner already knows, and the act of struggling with a problem forces that prior knowledge into play before the new material arrives, giving it something to attach to (Norman & Schmidt, 1992). Discussion in the group then drives elaboration, the generation of explanations and connections that multiplies the retrieval routes to a piece of knowledge. And because the material is learned in the context of a clinical or practical problem, the encoding is bound to cues resembling those of the situations in which it will later be needed, which is meant to ease transfer from classroom to practice.

These mechanisms also reframe what the learner is doing as self-regulated. In deciding what the problem demands, allocating study time, and monitoring whether the group's account holds together, students exercise the planning and self-monitoring of metacognition, and problem-based curricula are explicitly designed to cultivate the self-directed learning that professionals must sustain after formal training ends (Loyens, Magda, & Rikers, 2008). The claim is not that any of these processes is unique to problem-based learning, but that its tutorial structure elicits them more reliably than a lecture does.

The Guidance Debate

Problem-based learning sits at the center of one of education's sharpest disputes: how much guidance instruction should provide. Drawing on cognitive load theory, critics argue that human working memory is severely limited, that novices lack the schemas needed to cope with a complex problem, and that asking them to learn by solving one floods working memory with unproductive search, so that fully guided instruction, worked examples and direct explanation, produces more learning than problem-based, discovery, or inquiry methods (Kirschner, Sweller, & Clark, 2006). On this view the guidance a learner needs is greatest exactly when expertise is least, and the benefit of worked examples over unguided problem solving shrinks and can even reverse as expertise grows, the expertise-reversal effect.

Defenders of problem-based learning reply that the critique attacks a method no serious program uses. Problem-based learning, they argue, is not minimally guided discovery: it embeds extensive scaffolding, the case is chosen to bound the search, the tutor structures the reasoning, and the process itself constrains what students attend to (Hmelo-Silver, Duncan, & Chinn, 2007). Framed this way the disagreement is less about whether guidance helps, which both sides grant, than about whether the guidance in a well-run problem-based tutorial is sufficient and of the right kind. The exchange has been productive: it has pushed the field to specify the scaffolding a problem requires and to stop treating “problem-based” and “unguided” as synonyms (Hmelo-Silver, 2004).

Evidence on Effectiveness

The accumulated evidence tells a consistent, nuanced story rather than a verdict for one side. A meta-analysis of the method's effects found that problem-based learning yields a robust advantage for the application of knowledge and for skills, while its effect on the acquisition of knowledge itself is near zero or slightly negative, so that the answer to “does it work?” depends entirely on which outcome is measured (Dochy, Segers, Van den Bossche, & Gijbels, 2003). Because effects of this kind are reported as standardized differences, comparing programs requires reading an effect size as a difference between distributions, not as a raw score.

Later syntheses confirm and qualify the pattern. An overview of the method's process and impact concluded that problem-based learning is at least as effective as conventional instruction for knowledge and more effective for the retention and application of that knowledge over time (Yew & Goh, 2016). A scoping review of undergraduate medical education similarly found broadly favorable but heterogeneous results, with study quality and the exact form of the intervention varying enough to caution against a single summary number (Trullàs, Blay, Sarri, & Pujol, 2022). A recent meta-analysis in language teaching reported a moderate positive effect on achievement, evidence that the method's benefits are not confined to the health professions where it began (Orhan, 2024).

Worked Example

Suppose a study compares a problem-based course with a conventional one on a test of problem-solving skill and reports a standardized effect size of d = 0.46 in favor of the problem-based group, a value in the range meta-analyses report for skills. What does that number mean for an individual student? The effect size is the difference between the two group means expressed in standard-deviation units, so the average problem-based student sits 0.46 standard deviations above the average conventional student. Reading that position off the normal distribution, Cohen's U3 = Φ(0.46) = 0.677: the average problem-based student scores higher than about 68 percent of conventional students, up from the 50 percent that a zero effect would give.

Now take the other half of the pattern. If the same method shows d = −0.22 on a test of factual recall, the average problem-based student sits just below the conventional average, at Φ(−0.22) = 0.413, scoring higher than only about 41 percent of the comparison group. The two numbers together are the empirical signature of the method: a meaningful push on the application of knowledge, 68 versus 50, bought at a small cost in bare recall, 41 versus 50. The arithmetic shows why a single “does it work?” is the wrong question; the honest answer names the outcome and reads the effect size as a shift between distributions rather than a pass-or-fail (Dochy et al., 2003).

Current Directions

Recent work has moved from asking whether problem-based learning works to specifying the conditions under which it does. Scoping and systematic reviews emphasize the heterogeneity of what travels under the name, and press for reporting that distinguishes the degree of scaffolding, the training of tutors, and the outcomes measured, so that effect sizes can be compared meaningfully rather than pooled across incommensurable designs (Trullàs et al., 2022). Process-oriented accounts continue to unpack the tutorial into its component cognitive events, on the argument that knowing which events carry the effect is what will let the method be engineered rather than merely adopted (Schmidt et al., 2011).

Two frontiers are especially active. One is the extension of problem-based learning beyond the health professions, where meta-analytic evidence now reaches into fields such as language education and reports positive achievement effects, testing how far the method generalizes (Orhan, 2024). The other is the adaptation of the small-group tutorial to online and blended settings, which raises unresolved questions about how the activation, elaboration, and social interaction that the method depends on can be sustained when the group is not in a room together. The open problem common to both is measurement: matching the outcome assessed to the applied, transfer-oriented gains the method is designed to produce, rather than to the factual recall on which it is weakest.

Key Researchers

Howard S. Barrows (1928-2011). McMaster University and Southern Illinois University; the neurologist and medical educator who devised problem-based learning in the late 1960s and gave the method its first formal taxonomy, distinguishing degrees of problem structure and student direction. Wikipedia

Henk G. Schmidt. Erasmus University Rotterdam; he built the cognitive-psychological account of problem-based learning, framing the tutorial as the activation of prior knowledge and elaboration, and later specified the process model of what drives learning in the method. ORCID

Cindy E. Hmelo-Silver. Indiana University; she synthesized what and how students learn in problem-based learning and led the empirical defense of scaffolded problem-based instruction against the minimal-guidance critique. ORCID

John Sweller. UNSW Sydney; he originated cognitive load theory, the basis of the influential argument that problem-based and other minimally guided methods overload the working memory of novices. ORCID

Paul A. Kirschner. Open University of the Netherlands; he co-authored the 2006 critique arguing that minimal guidance during instruction does not work, sharpening the field's debate over how much structure problem-based learning requires. ORCID

Discussion

Problem-based learning endures because it embodies a genuine trade-off rather than a settled result. By putting the problem first, it reliably elicits the cognitive events, activation of prior knowledge, elaboration, and self-directed study, that deepen understanding and support transfer, and the evidence rewards it with advantages in the application of knowledge and in skills (Dochy et al., 2003; Hmelo-Silver, 2004). The same inversion that produces those gains is what the cognitive-load critique targets: a problem given to a novice can squander the very working-memory capacity that learning requires, and the method's weak showing on factual recall is consistent with that concern (Kirschner, Sweller, & Clark, 2006).

The most useful reading of the debate is that both sides are right about different things. Guidance matters most when expertise is least, so a problem-based tutorial that is genuinely scaffolded, rather than a discovery exercise wearing the name, can secure the method's benefits without paying the full cost the critics fear (Hmelo-Silver, Duncan, & Chinn, 2007). What remains is an engineering problem more than a philosophical one: to specify, for a given set of learners and goals, how much structure the problem should carry, how the tutor should intervene, and which outcomes to measure. Seen that way, problem-based learning is best understood not as an alternative to guided instruction but as one carefully guided way of arranging learning around the problems a discipline exists to solve (Schmidt, Rotgans, & Yew, 2011).

Glossary

Cognitive load.

The demand a task places on limited working memory; the basis of the argument that novices need guidance rather than open problems.

Constructivism.

The view that learners actively build knowledge from experience, the educational philosophy often invoked to justify problem-based learning.

Effect size.

A standardized measure of the magnitude of a difference between groups, such as Cohen's d, used to compare instructional methods across studies.

Elaboration.

The generation of explanations and connections around new material, which multiplies its retrieval routes; a process the tutorial discussion is meant to drive.

Expertise-reversal effect.

The finding that guidance which helps novices can hinder experts, so the optimal amount of instructional support falls as expertise rises.

Facilitator.

The tutor in a problem-based group, who guides reasoning and prompts justification rather than delivering content.

Ill-structured problem.

A problem with no single defined path or answer, whose ambiguity is what forces students to define learning issues for themselves.

Learning issues.

The explicit questions a group formulates about what it does not yet understand, which become the agenda for self-directed study.

Metacognition.

Awareness and control of one's own thinking; the planning and self-monitoring that self-directed study in problem-based learning demands.

Prior knowledge activation.

The bringing to mind of what a learner already knows, triggered by the initial problem so that new material has something to attach to.

Problem-based learning.

An instructional method in which an open-ended problem is presented before instruction and drives what students study.

Scaffolding.

Temporary support, from the tutor, the case, or the process, that lets learners manage a task they could not yet handle alone.

Self-directed learning.

Learning in which the student defines goals, chooses resources, and judges progress; a central aim of problem-based curricula.

Self-regulated learning.

The cyclical planning, monitoring, and adjustment of one's own study; the broader capacity self-directed learning draws on.

Tutorial process.

The recurring small-group cycle of defining the problem, identifying learning issues, studying, and synthesizing that constitutes problem-based learning in practice.

Worked example.

A fully solved problem shown to a learner; the guided alternative that cognitive load theorists find more effective than unguided problem solving for novices.

Frequently Asked Questions

What is problem-based learning?
Problem-based learning is an instructional method in which students learn by working through an open-ended problem in small groups before receiving formal instruction, so that the need to solve the problem drives what they study (Schmidt, 1983).

Where did problem-based learning come from?
It was developed for the medical curriculum at McMaster University in the late 1960s, on the judgment that students who had memorized basic science often could not apply it in practice (Schmidt, 1983).

How does a problem-based tutorial work?
A group clarifies and defines the problem, brainstorms from prior knowledge, formulates explicit learning issues, studies those issues independently, and reconvenes to pool and apply what was found against the problem (Barrows, 1986).

Why is problem-based learning thought to aid understanding?
Because confronting a problem activates prior knowledge, prompts elaboration during discussion, and encodes material in the context where it will be used, all of which support comprehension and transfer (Norman & Schmidt, 1992).

What is the main criticism of problem-based learning?
Cognitive load theorists argue that novices lack the schemas to cope with a full problem, so learning by solving one overloads working memory; they hold that guided instruction and worked examples teach more effectively (Kirschner, Sweller, & Clark, 2006).

How do defenders answer that criticism?
They argue that problem-based learning is heavily scaffolded rather than minimally guided, with the case, the process, and the tutor structuring the search, so the critique targets a discovery method the field does not actually use (Hmelo-Silver, Duncan, & Chinn, 2007).

Does problem-based learning improve achievement?
Meta-analyses find a positive effect on skills and the application of knowledge and a near-zero or slightly negative effect on the acquisition of facts, so the answer depends on the outcome measured (Dochy, Segers, Van den Bossche, & Gijbels, 2003).

Is problem-based learning only for medicine?
No. Although it began in medical education, it has spread across the health professions and into fields such as engineering and language teaching, where meta-analytic evidence also reports positive achievement effects (Orhan, 2024).

References