Why Learning to Struggle With Math Is More Valuable Than Getting the Right Answer

There is a particular kind of silence in a tuition centre that most Singapore parents would recognise as success. A child leans over a worksheet. The tutor points. The child writes. The answer is correct. Everyone feels good.

Manu Kapur, a learning scientist who spent years at the National Institute of Education in Singapore before moving to ETH Zurich, spent years studying that silence. What he found was uncomfortable: that moment of smooth, guided correction might be one of the least effective things that can happen to a child who is trying to learn mathematics.

What Productive Failure Actually Is

Kapur's research programme, which began in Singapore secondary schools in the mid-2000s, started with a simple and counterintuitive hypothesis: that struggling with a problem before receiving instruction might produce better learning than receiving instruction first.

The standard model of mathematics education — and the model that almost all tuition follows — is what researchers call "direct instruction followed by practice." You teach the method. You demonstrate it. The student practises it. This feels logical because it is logical. You wouldn't ask someone to fix a car engine before showing them what an engine is.

Kapur designed studies where he reversed this sequence. One group of students received direct instruction first, then practised problems. A second group was given complex problems with no instruction and asked to solve them — knowing they likely couldn't — before instruction was given. He called the second condition "productive failure."

The results, published across multiple studies in journals including Instructional Science and Cognition and Instruction, were consistent: students in the productive failure condition significantly outperformed their peers on tests of conceptual understanding and the ability to transfer knowledge to new problems. Not just marginally — the effects were large enough to be educationally meaningful.

This was not a fringe finding. Subsequent replications and meta-analyses, including work by researchers Tanmay Sinha and others building on Kapur's framework, have extended these findings across subjects, age groups, and countries.

Why Struggle Produces Better Learning: The Mechanism

The finding is only counterintuitive if you think the goal of learning is to accumulate correct answers. If you think the goal is to build a mental structure that can handle novel problems, the mechanism makes complete sense.

When a student attempts a problem they cannot yet solve, several things happen cognitively that do not happen when they are guided to a correct answer:

They activate prior knowledge. The brain searches what it already knows and tries to connect it to the new problem. Even failed attempts create a kind of cognitive scaffolding — a landscape of "things I tried that didn't work" that makes subsequent instruction land differently.

They develop awareness of their own confusion. This is called metacognitive awareness, and it is one of the most robust predictors of long-term academic performance. A student who can accurately identify what they don't understand is in a fundamentally different position from one who has been guided to correct answers without encountering the gap.

They notice the deep structure of the problem. Kapur's research found that students who struggled first were better able to identify the core mathematical features of a problem — the parts that actually matter — rather than pattern-matching on surface features. Surface-feature matching is the learning equivalent of memorising the shape of an answer rather than understanding why it has that shape.

When instruction comes after a genuine struggle, students have a prepared mind. They already know what the instruction is for. The concept lands in context rather than in a vacuum.

What This Looks Like in Singapore's Tuition Landscape

Singapore's mathematics results are genuinely exceptional. The country consistently performs at the top of TIMSS and PISA rankings. This is not an accident — the curriculum is coherent, teachers are well-trained, and the pedagogical approach to mathematics has genuine depth.

But the private tuition industry operates on a different logic. The incentive for a tuition centre is for the student to get better grades, faster. The incentive for parents is to see improvement in the near term, which means fewer red marks and less visible distress. These incentives are understandable. They are also, according to the research, in moderate tension with the conditions that produce durable mathematical competence.

When a tuition teacher consistently intervenes the moment a student shows difficulty, they are short-circuiting the productive failure cycle. The student gets the right answer. The parent is satisfied. But the cognitive work that would have produced a more robust mental model — the searching, the failed attempts, the awareness of confusion — never happens.

This is not a reason to abandon tuition. It is a reason to be very specific about what kind of tuition actually helps.

The Distinction That Parents Rarely Hear

There is a useful distinction between two types of errors in learning:

Errors during practice — mistakes made when a student is applying a method they have already learned — are worth correcting promptly. Allowing wrong practice to embed is genuinely counterproductive. If a student has learned the wrong procedure for solving simultaneous equations, you want to catch that quickly.

Errors during exploration — mistakes made when a student is grappling with an unfamiliar problem type or concept — are different. These errors are generative. They are not a sign that learning has failed; they are the mechanism by which deeper learning happens. Rescuing a student from this kind of error too quickly robs them of the cognitive work the error was producing.

Most tuition treats both categories identically. The pedagogically useful intervention is to distinguish between them and respond differently.

The Long Shadow of Exam Preparation

There is an obvious objection to all of this: Singapore children are being prepared for high-stakes examinations, and those examinations reward correct answers under time pressure, not struggle. If your child is sitting PSLE mathematics next year, is productive failure really the priority?

This objection is worth taking seriously. Short-term exam performance is a real and legitimate goal.

But a few things are worth holding alongside it.

First, the research suggests that productive failure improves performance on transfer problems — novel problems that require applying knowledge in unfamiliar contexts. This is precisely what examination setters try to create with their harder questions. A student who only practises familiar problem types will be well-prepared for the routine questions but vulnerable on the ones that differentiate the scores.

Second, the students who consistently perform at the highest level in mathematics — the ones who go on to do well in O-Level Additional Mathematics, H2 Mathematics, and beyond — are almost always students who have developed genuine conceptual understanding, not just procedural fluency. Procedural fluency can be drilled. Conceptual understanding requires the kind of cognitive engagement that productive struggle produces.

Third, and most importantly: your child will eventually hit a level of mathematics where the answer cannot be retrieved or pattern-matched. Every student reaches this point somewhere — for some it is at PSLE, for others at O-Levels, for others at university. When they reach it, the question is whether they have any experience of sitting with difficulty and working through it. A child who has never been allowed to struggle has no practice in struggle. When the difficulty finally arrives — and it always arrives — they have no resources to draw on.

What Good Mathematical Help Actually Looks Like

This is not an argument against helping children with mathematics. It is an argument for a specific kind of help.

The research literature on effective tutoring and teaching converges on a few principles that are directly relevant here:

Ask questions rather than provide methods. "What have you tried?" "What does this remind you of?" "What would happen if the number were smaller?" These questions keep the student's cognitive engine running. A tutor who says "here's how you do this" the moment difficulty appears is doing the intellectual work that the student needed to do.

Allow a productive amount of unresolved tension. Not indefinitely — genuine confusion that persists too long is demoralising and counterproductive. But there is a meaningful difference between a student who says "I've been stuck for five minutes" and a student who has genuinely wrestled with a problem and is ready to receive an explanation.

Make failure visible and normal. Students who believe that struggle means they are not smart enough avoid difficult work. Students who understand that struggle is how the brain builds new structures approach difficulty differently. This is Carol Dweck's growth mindset research, and while it has been somewhat oversimplified in popular culture, the underlying finding is robust: a student's theory of what struggle means has significant effects on how they respond to it.

Debrief wrong attempts explicitly. When a student tries something that doesn't work, the most valuable pedagogical move is often to examine why it didn't work. This is where the conceptual learning happens. Getting to the right answer and moving on skips the most useful part of the problem.

A Note for Parents Who Feel the Pressure

The tuition arms race in Singapore is not entirely irrational. It is a response to real competitive pressure in a system with real stakes. Opting out entirely carries genuine risk for individual families even if broad participation makes everyone worse off collectively.

But there is a version of supplementary education that is coherent with what the research shows actually builds mathematical ability — one that uses the time to develop genuine understanding, to practise working through difficulty, to build the metacognitive skills that transfer across problems and years. And there is a version that simply drills correct procedures under adult supervision, producing short-term grade improvements while quietly undermining the capacity for independent mathematical thought.

The question worth asking of any tuition arrangement — including ours — is: is my child getting better at mathematics, or are they getting better at doing mathematics when someone is in the room?

Those are not the same thing. And only one of them will serve them in the examination hall, and beyond it, when no one is in the room at all.

The Research, If You Want to Go Deeper

The work referenced in this piece is real and accessible. Manu Kapur's foundational papers — particularly "Productive Failure in Mathematical Problem Solving" (2010, Instructional Science) and subsequent meta-analytic work — are worth reading if you want more than a parent's summary. The broader literature on desirable difficulties in learning, synthesised by Robert Bjork at UCLA, provides the cognitive science context. Kapur's book Productive Failure: Unlocking Deeper Learning (2023) is the most accessible single resource.

This research is not obscure. It is taught in teacher education programmes. The gap is between what the research shows and what the incentive structures of private tuition produce. Closing that gap is, in our view, the actual job.