How to revise for a physics degree exam

What changes after A-Level

At A-Level the syllabus is a published specification, the past papers run to hundreds of questions, and a mark scheme says which words and steps earn the marks. Revision that works there is revision that learns the patterns: after enough past papers, very few questions are new.

A physics degree changes 3 things at once. The syllabus is the lecture notes, written by whoever taught the module that year. The paper examines the whole module in a sitting, and many papers pair a piece of bookwork, where you derive a standard result from memory, with an unseen problem that uses it somewhere you have not been shown. And the archive is thin: a handful of past papers per module, often without solutions, set by a lecturer who knows you have them.

So the A-Level method does not just work less well; it stops working. Studying worked solutions does make you faster and more accurate on problems built the same way, but in the classic experiment, on algebra, that benefit did not carry to problems built differently.¹ A paper of unseen problems is a paper of problems built differently. The methods below are the ones with evidence behind them, and every one of them feels slower than reading your notes.

Derive it from a blank page, then check

The most useful single change is to stop reading derivations and start producing them. Close the notes, take a blank page, and write the derivation out from the start, every assumption stated, the way the exam will ask for it. Then open the notes and compare line by line.

This is retrieval practice, and it is the best-evidenced technique in the field. In Roediger and Karpicke’s 2006 experiment, students who reread a passage did better on a test 5 minutes later, and worse on one a week later, than students who had tested themselves, and the rereaders were the more confident group throughout.² Pooling 159 comparisons, Rowland found that testing yourself beat rereading by a medium margin, $g = 0.50$, and by more when the practice asked for recall rather than recognition.³ Following a derivation in your notes is recognition: every line looks right while it is in front of you. Writing it from nothing is recall.

Almost nobody revises this way. Asked how they study, 84 per cent of 177 university students listed rereading, and 11 per cent listed testing themselves.⁴ The reason is the feeling. A derivation you have just read feels understood, because each step follows from the one before while the one before is on the page. The blank page takes the steps away, and what is left is an honest measure of what you can do.

The comparison is where the revision happens. Mark the first line you could not produce. It is often not the algebra: it is an assumption, or the choice of where to start, and that line is the one to come back to in a few days.

Use the worked solutions, then take them away

Problem sheets come with solutions sooner or later, and how you use them matters more than whether you do. The best-known study of it was done with university students learning mechanics. The students who learned most from worked physics examples were the ones who talked themselves through why each step worked; the weaker students explained little, overestimated how well they understood, and leaned on copying the example.⁵

That study watched the difference rather than causing it, but prompting students to explain material to themselves has been tested directly, and across 64 studies it helped by about $g = 0.55$.⁶ There is a caveat worth knowing: a 2023 meta-analysis of worked examples in maths found that adding self-explanation prompts to them made them slightly worse.⁷ A reasonable reading of both is to explain the steps you cannot see the reason for, rather than every step.

Then take the support away as you improve. Learners who moved from a complete worked solution to the same problem with the last step blanked, then the last 2, then the bare problem, solved more new problems afterwards than learners who alternated examples and problems, in the same time.⁸ That rests on a single experiment with 78 university students working on probability, but it fits a wider finding: help that clearly serves a beginner loses its value as knowledge grows, and past a point makes performance worse than no help at all.⁹

In practice: on the first pass through a sheet, cover the solution, attempt the problem, and when you are stuck uncover a single line and ask why that line comes next. On the second pass, a week or more later, do the same problems with the solutions shut away.

Mix the problem types

A question in a degree exam does not say which lecture it came from. A block on a slope can be a forces problem, an energy problem or a momentum problem, and choosing the method is often the hardest part. Problem sheets hide that choice, because every question on the week 4 sheet uses week 4.

Practising different kinds of problem mixed together is called interleaving. In a trial across 54 seventh-grade maths classes, students who practised the same problems mixed up scored 61 per cent on an unannounced test about a month later, against 38 per cent for students who practised them grouped by topic.¹⁰ The size depends on the material. Across 238 effect sizes, interleaving helped most when the kinds of problem looked alike from outside, so that telling them apart is the hard part, and maths came out at less than half the size of that trial.¹¹ Physics problems that look alike and need different methods are exactly that case, though none of these studies was run on a physics degree.

So once every topic has been practised on its own, stop working sheet by sheet. Take questions from across the module, shuffle them, and for each one write down the method before you start. Picking the wrong method and finding out why teaches you more than a right answer found by knowing which sheet it came from.

Spread it across the term

The same hours are worth more on different days, and the best gap depends on how far away the exam is. In the study that measured it directly, more than 1,350 people learned facts, reviewed them after a gap of up to 3.5 months, and were tested up to a year later. For a test 1 week away the best gap was about a day; for a test 10 weeks away it was about 3 weeks.¹² A module examined at the end of term is close to the 10-week case, so coming back to each week’s lectures 2 or 3 weeks later is far nearer the measured best than coming back to all of them in the last fortnight. There is more on the gap in the post on spacing.

Do not retire a topic because it feels finished. Learners allowed to set aside the flashcards they judged they had learned recalled less later than learners who kept the whole set in rotation.¹³ A derivation you could do in week 3 is one to try again from a blank page in week 9.

Of 10 study techniques reviewed for how well they generalise, only 2 got the highest rating: testing yourself and spreading study out.¹⁴ Interleaving and self-explanation were rated moderate, which is the honest place to leave them here too.

One derivation, revised this way

Here is the method on a derivation from first-year relativity: why a moving clock runs slow. Try it on a blank page before reading on. The setup is a light clock, 2 mirrors a height $h$ apart with a pulse of light bouncing between them, and a tick for every round trip.

1. In the clock’s own frame the light goes straight up and straight back, a distance $2h$, so a tick takes the proper time, the tick as measured by someone riding with the clock:
   $$\Delta t_p = \frac{2h}{c}$$
2. In a frame where the clock moves sideways at speed $v$, a tick takes $\Delta t$, and during it the clock moves $v\,\Delta t$. Each leg of the light’s path is the slanted side of a right-angled triangle with height $h$ and base $v\,\Delta t/2$.
3. The light still travels at $c$ in this frame. That is the second postulate, and it is where the physics enters. So each slanted leg has length $c\,\Delta t/2$.
4. Pythagoras on that triangle:
   $$\left(\frac{c\,\Delta t}{2}\right)^2 = h^2 + \left(\frac{v\,\Delta t}{2}\right)^2$$
5. Multiply through by 4 and collect the $\Delta t$ terms:
   $$\Delta t^2\,(c^2 - v^2) = 4h^2$$
6. Take the square root, divide top and bottom by $c$, and use step 1 to replace $2h/c$:
   $$\Delta t = \frac{\Delta t_p}{\sqrt{1 - v^2/c^2}} = \gamma\,\Delta t_p$$

Now mark it. Steps 4 to 6 are algebra. The lines that carry the physics are steps 2 and 3, and they are the ones to check. Step 3 is the second postulate: without it the light would simply travel faster along the longer path, and nothing would change. Step 2 carries a quieter assumption that is easy to use without noticing: it takes the same $h$ in both frames. That is only allowed because lengths at right angles to the motion are the same for every observer, and without it the triangle has no known side.

That is the 3-layer shape every Kerfox solution takes: the answer, $\Delta t = \gamma\,\Delta t_p$; the working, the 6 numbered steps; and why the method works, which here is a postulate, a right-angled triangle and an assumption about lengths holding it up. Being able to produce all 3 from a blank page is what revising a derivation means.

The plan, week by week

• Find out what the exam gives you. If there is a formula sheet, what it prints you can look up; what it leaves off is what the paper expects you to produce.
• Every week of term: a derivation from that week’s lectures, from a blank page, a few days after the lecture.
• Every problem sheet, twice: once with the solution covered and uncovered a line at a time, then again a week or more later with it shut away.
• From halfway through term: mixed sets drawn from everything so far, naming the method before solving.
• The last month: past papers under time. Every question that lost marks goes back into the mixed pile, not onto a list of things to reread.
• Nothing is ever finished. A topic that feels done is one to put back on the calendar, not to cross off.

None of this has been tested on Kerfox, or on any app. The studies above were run with other people’s students on other material, and several were not about physics at all. What Kerfox does is take the calendar off your hands: its physics questions open every solution in the 3 layers above, and its memory engine brings each topic back when it is due rather than when it feels due.