# Why Engineering Students Blank on Problems They've Already Solved

> The formula sheet is handed out at the exam. Knowing which method a problem wants is the part that decays.

Source: https://getmemset.app/blog/memset-for-engineering-students  
Published: 2026-08-07

---

Halfway through a thermodynamics final, a friend of mine sat in front of a problem he could not start. The official formula sheet was on the desk. First law, second law, ideal gas relations, all of it printed and permitted. He read the problem four times and nothing happened.

Afterwards a classmate looked at the same question and said "that's a control volume, one inlet, one outlet, steady state." The whole thing came apart in about ten seconds. He knew every equation involved. He had derived most of them. What he could not do, with a clock running, was read a paragraph about a turbine and see which of the twelve things he knew applied to it.

That gap gets misdiagnosed constantly. Students walk out of exams convinced they need to memorise more. What they actually lost was the mapping between a problem's description and the method that solves it, and no amount of extra memorising rebuilds it.

I relearned this the slow way across several years of engineering coursework, and it is part of why I ended up building [a review planner](https://getmemset.app) that schedules returns to material rather than trying to store the material itself. Engineering knowledge does not really live in your notes. It lives in how fast you recognise what you are looking at.

![Four minimal engineering schematics (a cantilever beam, a circuit loop, a pipe bend, and a spring with a block) drawn on pale graph paper, with faint lines converging from each one toward a single glowing point at the centre](https://getmemset.app/blog/memset-for-engineering-students.webp)

## What experts see that everyone else doesn't

In 1981 three cognitive scientists ran an experiment that explains the thermodynamics story better than anything I have read since. They handed physics problems to PhD students and to undergraduates who had just finished an introductory course, and asked both groups to sort the problems into whatever categories made sense to them.

The undergraduates sorted by appearance. Inclined planes with inclined planes, springs with springs, pulleys with pulleys. The PhD students sorted by the principle that would solve the problem, so a spring question and a block sliding down a ramp landed in the same pile because [both were conservation of energy problems](https://doi.org/10.1207/s15516709cog0502_2). Identical problems, completely different filing system.

The undergraduates were not less intelligent, and they were not missing the equations. They had organised their knowledge around **surface features**, because that is what a semester of chapter-by-chapter problem sets trains you to notice.

The same pattern shows up in the chess research from the seventies. Masters shown a real game position for five seconds could [reconstruct it far more accurately than weaker players](https://doi.org/10.1016/0010-0285%2873%2990004-2), and lost that advantage completely when the pieces were scattered at random. They were not seeing thirty-two pieces. They were seeing a handful of familiar configurations.

Engineering works the same way. Each discipline has maybe forty recurring problem shapes, and a lot of what gets called ability is really the speed of naming the shape in front of you.

## The part that fades is the part nobody writes down

A formula is cheap to lose. Forget the exact form of Darcy-Weisbach and you find it again in eleven seconds. Exams increasingly concede this by handing out reference sheets, and the FE gives you a searchable handbook for the entire subject.

Recognition works differently, because it is not written down anywhere. It exists only as something you can do quickly or cannot do at all. You also cannot audit it by looking at it, the way you can glance at a formula sheet and confirm the equations are still there. You find out during the exam.

This matters more in engineering than in most degrees, because the curriculum is a chain. Statics feeds mechanics of materials. Differential equations feed signals and systems. Thermodynamics feeds heat transfer, which feeds most of mechanical design. Every course assumes the pattern library from the previous one arrived intact, and nobody checks.

You can watch it break in specific ways. Students who did fine in statics turn up to mechanics of materials able to write the equilibrium equations from memory and unable to look at a loaded beam and decide where to cut it. The equations survived. The judgement about when to reach for them did not.

## Grinding more problems does not build it

The standard advice is to do more problems, which is not wrong so much as incomplete.

Working twenty problems from chapter seven in one sitting is **blocked practice**. Before you read the first word of problem one you already know the method, because the chapter told you. Every problem after that confirms it. You are practising execution, the algebra and the bookkeeping and the unit tracking, which genuinely needs practice. You are not practising the choice, because the choice was made for you by the table of contents.

Then the exam pulls from chapters four through eleven in no particular order, and the choice becomes the entire difficulty.

> A problem set that announces its method in the chapter heading is not testing the thing your exam will test.

There is a clean study on this with mathematics students, which sounds distant from a fluids exam and tests exactly the same thing. Students practised several problem types either blocked by type or shuffled together. On a test one week later, [the shuffled group performed vastly better](https://doi.org/10.1007/s11251-007-9015-8), even though the shuffling made their practice sessions slower and more error-prone while they were happening.

That trade is the point of [interleaving](https://getmemset.app/learn/interleaving-vs-blocked-practice), and it is worth knowing in advance that it feels bad by design. Mixed practice produces more mistakes today and better retention in a month. Students read those mistakes as evidence the method is not working, and go back to blocking.

## What a review session actually looks like

The instinct is to reread worked solutions. That feels like reviewing and does close to nothing, because reading a solution you already understand produces recognition without retrieval. You nod along at each step, and nodding along is not the skill being tested.

A version that works takes about seven minutes. Pull up three problems you solved weeks ago. Read only the problem statement and keep the solution covered. For each one, say out loud or write down what kind of problem it is, which principle applies, and what the first two steps would be. Then stop. Do not do the algebra.

Now uncover the solution and compare. The distance between what you produced and what the solution actually did is the whole lesson, and it is usually small and specific: you reached for energy where the problem wanted momentum, you forgot the control volume was moving, you missed that the beam was statically indeterminate.

Three problems, seven minutes, no calculator.

That is a review session aimed at what engineering exams actually test, and [pairing retrieval with spacing](https://getmemset.app/learn/active-recall-and-spaced-repetition) is the method in full. Copying your notes into a neater format is not.

## Four hard courses, one calendar

Spacing has an obvious failure mode, and engineering students hit it early, because the course load is dense and every course generates review debt at the same rate.

By week nine you might be carrying differential equations, thermodynamics, circuits, and mechanics of materials, plus a lab that eats a full afternoon. If every topic in every course gets its own schedule, some Tuesday in November says: review Laplace transforms, review entropy for a control volume, review node voltage analysis, review Mohr's circle, and clear three leftovers from last month. Nobody does that day. They open the list, feel the weight of it, and close it.

What gets blamed afterwards is discipline. What actually happened is that **the schedule stacked badly** and there was no rule for deciding what to drop.

A cap is that rule. Decide in advance how many items from a given course can surface on the same day, and let everything past the cap roll forward. Being three days late costs you very little. Across [254 studies and more than 14,000 participants](https://doi.org/10.1037/0033-2909.132.3.354), spreading a fixed amount of study over separate sessions beat massing it into one, with the advantage growing the longer the material had to be retained. Hitting the interval exactly matters far less than the review happening at all.

In Memset that cap is what a Category does, and it is the single setting most likely to keep the habit alive through a genuinely bad week.

## Where flashcards fit and where they don't

Engineering does have material that suits cards. Unit conversions, material properties, standard constants, sign conventions, definitions you want available instantly during a lab. Card-based tools handle those well and there is no reason to avoid them.

The recognition layer fits cards badly, for a mechanical reason. The front would have to be a full problem statement, usually with a figure, and the back would be a worked solution running half a page. At that point you have not made a card. You have retyped a page out of your own problem set.

This is the distinction that pushed me toward building a planner instead of a card editor, and it is the practical difference [between Memset and Anki](https://getmemset.app/compare/memset-vs-anki): Anki wants the material rebuilt as cards, which is a good trade for vocabulary and a poor one for a fluids problem set. A memset is a reference. "Thermo, Ch. 6, problems 4 to 9" is enough, because the problems are already written and already sitting in your folder.

## The part that carries

The case for any of this is not really the next midterm. You can survive a midterm on caffeine and one desperate night, and most of us do at least once.

The case is that engineering keeps asking for the same material back. The next course assumes it. The [FE exam](https://ncees.org/exams/fe-exam/) is written separately for each discipline and usually sat near the end of the degree. The mechanical version packs 110 questions across mathematics, statics, dynamics, mechanics of materials, fluid mechanics, thermodynamics, and engineering economics into one sitting, most of it from courses finished two or three years earlier. Then interviews ask you to reason about a system out loud with no reference sheet in front of you. Then work hands you a problem with no chapter heading attached to it at all.

At every point in that chain, the formula is the easy part. You will have the reference. The hard part is looking at a situation and knowing, within a few seconds, what kind of problem it is.

That skill gets built by returning to problems you have already solved, spread out over months, in short sessions that are mostly thinking rather than calculating. It also dies quietly the moment the schedule gets heavy, which in engineering is somewhere around week six. Keeping that habit survivable when four courses want the same Tuesday is what [Memset](https://getmemset.app) is built to do.