Exam Anxiety and Mock Scores: I Scored 96, Then 40

Before you conclude that anxiety is costing you marks, check the arithmetic. On 29 July 2026 I scored 96 in a CAT mock, my best to that point. On 30 July I scored 40. Nothing about my knowledge changed in 48 hours. The largest study I could open on anxiety and mathematics performance - Devine and colleagues, 433 British secondary school children - found maths anxiety explained 11.88% of the variance in girls' performance and 1.5% in boys'. Real, and much smaller than the story you tell yourself after a bad paper.

This post is about what else is in the gap between your mock scores and your real ones, because in my own log the honest answer turned out to be two statistical outliers and one bad decision, not a collapse of nerve.

The 48 hours, section by section

Here are those two papers against my own average across all 20 mocks I logged between 10 May and 8 August 2026.

Put next to the averages, the drama mostly evaporates.

The 96 was not a great all-round paper. VARC was 21 against an average of 20.7 - completely ordinary. DILR was decent. The entire difference was a QUANT score of 50, against a 20-mock average of 27.4 and a range everywhere else of 13 to 36. That was a single-section outlier and nothing more.

The 40 was the mirror image. QUANT came in at 29, which is above my average. VARC at 9 was poor but within my range - I had scored 3 and 5 on other papers. The whole collapse was a DILR score of 2, against an average of 18.1 and a next-lowest of 8.

So two consecutive papers, two outliers in opposite directions, one section each. My note on the second paper, written that same week, reads: "Could have definitely gotten 30 marks." The sets were doable. I picked wrong, committed early, and had no clock left when I noticed. That is a decision failure, and it has a name that is not anxiety: it is triage.

What the anxiety research actually establishes

The most directly relevant open-access study I could find is Devine, Fawcett, Szucs and Dowker (2012) in Behavioral and Brain Functions. It matters because it did something most anxiety studies do not: it measured maths anxiety and general test anxiety separately, so it could ask what each one explains on its own. 433 British secondary school children, school years 7, 8 and 10.

Three things worth sitting with.

The effect is real and it is modest. A correlation of -0.35 is a genuine relationship. It is also about 12% of the variance, which means roughly 88% of the difference between one performance and another is something else. For boys in this study the model explained 1.5% and neither anxiety measure was a significant predictor at all.

Maths anxiety and test anxiety are not the same thing. Once maths anxiety was controlled for, test anxiety no longer correlated with performance in either group. If you feel fine about a history exam and sick about quant, that is a specific thing, and the generic exam-nerves advice is aimed at a different problem.

There was no gender difference in performance, despite girls reporting maths anxiety 0.33 standard deviations higher and test anxiety 0.55 standard deviations higher than boys. Anxiety was elevated and scores were not lower.

The mechanism, and the finding that undoes generic advice

The plausible mechanism is not mysterious. Pekrun, Goetz, Titz and Perry's 2002 review in Educational Psychologist summarises it as test anxiety reducing the working-memory resources available, "leading to an impairment of performance at complex or difficult tasks that draw on these resources". A DILR set with four constraints and a table you are holding in your head is exactly such a task.

But the same paper contains a result that should stop anybody from giving you confident advice about your anxiety, including me. They ran a diary study: 72 students, daily measurements over six weeks before and during their final university exams, then calculated the correlations within each individual over time rather than across the sample.

"It turned out that daily anxiety related differently to variables of learning in different students. For example, whereas anxiety correlated negatively with motivation to learn for some students, it correlated positively for others... Some students may even profit motivationally from their anxieties, whereas others are handicapped."

The group-level correlation hides individuals whose relationship runs the other way. Which means the honest answer to "is anxiety hurting my score?" is that nobody can tell you without your own data, because in the one study I found that looked, the sign of the relationship differed between people.

The same authors add the caveat that decides how you should read any of this about yourself:

"Emotions, learning, and achievement are linked by reciprocal causation, with students' emotions influencing achievement, but feedback of achievement in turn affecting their emotions."

You cannot tell from inside a bad fortnight whether the anxiety produced the scores or the scores produced the anxiety. Both are happening.

One more thing the anxiety literature suggests about how you prepare

Pekrun and colleagues report that in their data, anxiety, anger and shame sometimes correlated positively with the use of rehearsal strategies - rote, repetitive study - while relationships between negative emotions and flexible strategies like elaboration, organisation and critical thinking were negative and weaker.

If that pattern holds for CAT aspirants, and I want to be clear that nobody has tested whether it does, the practical risk of anxiety is not only what it does to you in the hall. It is that it pushes you toward the kind of preparation that feels safe - re-reading formula sheets, redoing topics you already know - and away from the uncomfortable, selection-heavy practice CAT actually rewards. That is a slower and more expensive failure than a bad exam morning.

What to do instead of diagnosing yourself

• Never read one mock as a signal. My 96 and my 40 are two days apart and neither is my level. A five-mock rolling average is the smallest unit worth reacting to.
• Check whether the drop is one section or all three. A single-section collapse next to normal scores elsewhere is almost never a general state problem. Mine was DILR, and DILR alone.
• Write down what the paper was worth to you. After every section, record what you think you could have scored with better selection. The gap between that number and your actual score is triage loss, and it is measurable in a way that "I panicked" is not. Mine was routinely 15-20 marks in DILR.
• Compare your section scores to your own distribution, not to your best day. Half of what feels like a crash is regression from an outlier. Half of what feels like a breakthrough is the same thing running the other way.
• If the pattern is genuinely consistent, treat it as real. Anxiety with a stable, repeated effect across many papers is a different thing from one bad Tuesday, and it is worth taking to somebody qualified rather than to a blog post.

My honest conclusion after three months of logging: what I would have called an anxiety problem was a selection problem with an anxiety-shaped feeling attached. That is not true for everybody. It was true for me, and the only reason I know is that the section scores sat next to each other in a table.

Where this evidence does not reach

• Devine et al. studied 11 to 15 year olds doing mental arithmetic tests. Not graduates sitting a three-hour paper with sectional locks, negative marking and an admissions decision attached. The construct transfers; the coefficients do not.
• Pekrun et al. studied German and Canadian university students. Their diary study had 72 participants. It is the best individual-level evidence I could open and it is small.
• Almost all of this is correlational. You cannot randomly assign people to be anxious, so the direction of causation is genuinely unresolved, and the authors themselves say so.
• I have found no study of test anxiety in Indian competitive entrance exams that I could read in full. That is a real hole. CAT, JEE and NEET have a stakes profile and a cohort size with no equivalent in this literature, and if somebody has measured it properly, I have not found them.
• My own data is one candidate and 20 papers. The outlier analysis above is arithmetic on a small sample, not a controlled test of anything. It is a way to look at your own log, not a finding about yours.
• None of this is clinical advice. There is a difference between the ordinary dread of a hard paper and an anxiety disorder, this post is about the first, and nothing here should be used to talk yourself out of getting help for the second.

Sources

• Devine, A., Fawcett, K., Szucs, D., & Dowker, A. (2012). Gender differences in mathematics anxiety and the relation to mathematics performance while controlling for test anxiety. Behavioral and Brain Functions, 8:33. Open access. All correlations above are from its Results section.
• Pekrun, R., Goetz, T., Titz, W., & Perry, R. P. (2002). Academic emotions in students' self-regulated learning and achievement. Educational Psychologist, 37(2), 91-106. Read in full via the University of Konstanz repository.
• Mock log, 20 attempts, 10 May to 8 August 2026 - first-party, published in full in the 20-mock post. Section averages above are computed across all 20.

The triage-gap number described here is the single most useful thing I recorded all quarter, and it is not a score. Karma Yogi keeps per-section scores and per-mock analysis notes together so the two can be read against each other, which is what turned "I panicked" into "I picked the wrong set at minute four".