The Pedagogy of Getting It Wrong First
Origin and core claim
Manu Kapur developed Productive Failure through a series of classroom-based design experiments starting in the mid-2000s, mainly in Singapore secondary math and science classrooms, later extended at ETH Zurich (Kapur, 2008; Kapur & Bielaczyc, 2012). The core claim is counterintuitive. Students who fail to solve a novel problem on their own, before being taught the correct method, often understand that method better than students who were taught it first and then practiced. This runs against the instinct that failure wastes time and that struggle without success is inefficient.
The two-phase structure
In the generation and exploration phase, students receive a problem that targets a concept they have not yet been taught. A classic classroom example asks students to invent a way to decide which of two football players is more consistent, given only their goal totals across a season, before they have ever learned about standard deviation or variance (Kapur, 2014). Another common task asks seventh-graders to invent methods for comparing average speeds on different routes before formal instruction on rate. The problem must sit in a narrow difficulty band: hard enough that no student can solve it correctly with existing knowledge, yet rich enough that partial, intuitive, or analogous approaches remain possible. Students draw on whatever they already have, informal reasoning, related procedures, visual intuitions, prior formulas used incorrectly, and generate multiple candidate solutions. Almost all of them are flawed.
In the consolidation phase the teacher does not simply reveal the correct answer. They take the actual student-generated solutions, place them alongside the canonical method, and work through why each attempt captured something real but missed a critical feature. With the consistency problem, a teacher might show how one group’s range calculation ignored the number of games, how another group’s average absolute deviation treated every point equally, and how both ideas point toward the need for a measure that accounts for both spread and frequency. This comparative step is the load-bearing part of the design. Without it, the approach collapses into unguided discovery learning, which the evidence does not support (Kapur, 2016).
Application to an English Home Language classroom
The same structure works with language and literature. Before teaching formal literary analysis, give students a short story or poem and ask them to invent as many ways as possible to explain why the text feels unsettling, powerful, or incomplete. Students generate incomplete or flawed readings: some focus only on plot, some on isolated word choice, some on personal reaction. In consolidation the teacher places those attempts beside the canonical tools of theme, narrative voice, irony, and structure, showing what each student attempt noticed and what it missed.
Before teaching persuasive essay structure, present a complex local issue and ask students to invent the strongest possible argument sequence from scratch. Most will produce lists of points, emotional appeals, or incomplete counter-arguments. The consolidation phase then compares those attempts with the required moves of claim, evidence, concession, and rebuttal. The struggle surfaces the gaps that formal instruction later fills.
The same pattern can be used with vocabulary or grammar. Ask students to convey a precise idea without using a target structure or word set. Their workarounds become the material for later comparison with the conventional form.
Why it works
Kapur and collaborators identify four main mechanisms (Kapur & Bielaczyc, 2012; Loibl, Roll & Rummel, 2017).
Attempting the problem forces students to bring forward everything even loosely relevant. That knowledge becomes active and comparable rather than remaining inert.
Because students try multiple approaches and see them fail in different ways, they notice why certain features of a problem matter. Those features are rarely visible when a method is simply demonstrated.
The variety of failed solutions gives the later correct explanation something to attach to. New information lands in a differentiated structure rather than a blank one.
Get Corbin Vermeulen’s stories in your inbox
Join Medium for free to get updates from this writer.
Struggling makes knowledge gaps visible in a way that prepares students to absorb the fix when it is offered. This is related to what Bjork calls a desirable difficulty.
A striking correlational finding underpins the claim that structured failure is not wasted effort. The number and variety of solutions a student generates during the failure phase predict post-test conceptual understanding, even though none of those solutions were correct (Kapur, 2014). The struggle is valuable partly because of the range of wrong answers it produces.
How PF compares to alternatives
Direct instruction typically wins on immediate procedural fluency and near-transfer items. Students taught the method directly are faster and more accurate right after instruction. Productive Failure tends to catch up and then surpass direct instruction on conceptual understanding and far-transfer items, often showing its advantage on delayed rather than immediate post-tests.
The critical contrast is with pure discovery or unguided problem-based learning. Kapur’s experimental and meta-analytic work is clear that unguided exploration without a strong consolidation phase does not reliably produce these gains and can leave misconceptions uncorrected. The design works only because failure is followed by expert-led, comparative correction.
Design constraints that matter
Not every hard problem given early will work. The problem must admit multiple solution paths and representations. A single-path problem produces uniform failure with little diagnostic value for the consolidation phase. It must also remain within reach of students’ prior knowledge in an intuitive or informal sense, even if the formal tool is missing. Otherwise the failure is only frustration. Time-on-task and problem design matter more than group size; the approach has been shown to work in both individual and small-group settings. The consolidation phase requires real instructional skill. A teacher or system must accurately read the range of student solutions and use them diagnostically, which is harder than delivering a scripted lecture.
Broader context
Productive Failure belongs to a family of related ideas. Schwartz and Martin’s “inventing to prepare for future learning” showed earlier that invention tasks, even unsuccessful ones, improve subsequent learning from instruction (Schwartz & Martin, 2004). Bjork’s desirable-difficulties literature argues that conditions that make learning feel harder in the moment (spacing, interleaving, generation) often produce more durable and transferable knowledge than conditions that feel easier. Productive Failure is a classroom-scale application of the same principle. The discomfort is not a side effect to minimize. It is the mechanism, provided good instruction follows rather than leaving the struggle to resolve itself.
Corbin
Sources:
Kapur, M. (2008). Productive failure. Cognition and Instruction, 26(3), 379–424.
Kapur, M. (2014). Productive failure in learning math. Cognitive Science, 38(5), 1008–1022.
Kapur, M. (2016). Examining productive failure, productive success, unproductive failure, and unproductive success in learning. Educational Psychologist, 51(2), 289–299.
Kapur, M., & Bielaczyc, K. (2012). Designing for productive failure. Journal of the Learning Sciences, 21(1), 45–83.
Loibl, K., Roll, I., & Rummel, N. (2017). Towards a theory of when and how problem solving followed by instruction supports learning. Educational Psychology Review, 29(4), 693–715.
Schwartz, D. L., & Martin, T. (2004). Inventing to prepare for future learning: The hidden efficiency of encouraging original student production in statistics instruction. Cognition and Instruction, 22(2), 129–184.
