HSW-0245
A student writes a target at the top of a study plan: 85% on the next test.
The number is specific. It is measurable. It may even be motivating. But none of those properties tell us whether 85% is a good forecast, a useful stretch, an impossible wish, or a target so easy that it changes nothing.
That distinction matters because goal setting is often taught as though precision were enough. “Make the goal specific” is good advice only after a second question has been answered: specific relative to what evidence?
In a 2026 panel study of 426 learners in a six-month online literature-exam preparation course, Maxim Boitcov, Kseniia Adamovich and Aleksandra Getman examined not only whether students set score goals, but how accurately those goals matched subsequent test performance. Learners set targets repeatedly across five waves, allowing the researchers to examine within-person changes as well as differences between learners.
The study found that setting specific goals was associated with higher test scores, and that goal accuracy was also associated with performance. Students whose targets substantially exceeded their eventual scores tended to perform lower than accurately calibrated students, while students whose targets understated their eventual scores tended to perform higher. Prior knowledge and greater course engagement were associated with a lower likelihood of overestimating the next result.
Those findings are useful, but they do not prove that choosing a realistic number causes better marks. The design was observational panel research, not a randomised intervention assigning learners to accurate and inaccurate goals. A learner who is better prepared may both predict more accurately and score better. Calibration can therefore be a diagnostic signal even when the causal pathway remains unsettled.
HSW-0245 calls the study problem goal calibration: aligning the target for a future performance with the evidence currently available about knowledge, task difficulty, time, reliability and uncertainty—then updating the target when new evidence arrives.
The direct answer
A study target is useful when it does two jobs at once. It should create direction for action, and it should remain anchored to evidence about the learner’s current state. Specificity without calibration can produce false precision.
The practical rule is not “lower your ambitions.” It is: separate aspiration from forecast. You may aspire to an A, admission to a selective course, or mastery of a difficult subject. Your forecast for the next test should still be updated from recent independent evidence. The gap between aspiration and forecast then becomes a planning problem rather than a confidence contest.
Three numbers that should not be collapsed into one
Students often use one number to carry three different meanings:
- Aspiration: the result you ultimately want.
- Forecast: the result current evidence suggests you are likely to produce under the next set of conditions.
- Operating target: the performance level that should guide the next block of work.
They can be identical, but they need not be. A learner may aspire to 90, forecast 68 on a full unseen paper today, and set an operating target of 75 for the next fortnight. That is not pessimism. It is a control system.
When all three are collapsed into “I want 90,” a missed target tells us little. Was the goal unrealistic? Was the study plan weak? Was the test unusually difficult? Did execution fail despite good knowledge? Without separating the jobs, the number cannot diagnose the failure.
What the 2026 panel study actually measured
The learners were aged 16–18 and enrolled in a paid online preparatory course for a literature examination. Before each upcoming test, they could enter a numerical target from 0 to 100. Goal setting was voluntary, and instructors did not coach students toward particular values. The goal remained visible in the learning system until the next test.
Researchers classified a goal as accurate when it fell within ten points of the eventual test score. A goal more than ten points above performance counted as overestimation; a goal more than ten points below counted as underestimation. They also considered prior test performance, pretest knowledge and the proportion of webinars viewed.
Because the same learners contributed repeated observations, the panel design could track how goals and outcomes changed across time. That is stronger for studying dynamics than a single one-off survey, but it still does not randomly manipulate calibration. The results therefore support relationships, not a simple causal statement that “accurate goals raise scores.”
The mathematical trap inside “overestimation”
There is an important interpretive trap. If overestimation is defined partly by the eventual score, students who score unexpectedly low will mechanically be more likely to land in the “overestimated” category. That does not make the observed relationships meaningless, but it is one reason not to interpret the large score differences reported between calibration groups as the causal effect of choosing the wrong goal.
A better educational use is diagnostic: repeated overestimation tells us that the learner’s internal forecast is not tracking performance well enough. We can then ask why. Perhaps practice is easier than the test. Perhaps marking is too generous. Perhaps only favourite topics are being sampled. Perhaps recent success is being extrapolated too far. Perhaps the learner is forecasting peak performance rather than typical performance.
Illustrative case: the student whose best paper becomes the forecast
This is an illustrative teaching case, not data from the 2026 study.
Maya completes four timed mathematics papers: 63, 66, 81 and 68. She predicts 82 for the next paper because “I already got 81 once.”
The 81 is real evidence. The mistake is treating a peak as though it represented the distribution. A more defensible forecast asks what usually happens, under what topics, with what error types, and whether the strong paper contained conditions likely to recur.
Maya may keep 82 as an aspiration. Her operating forecast might be closer to the high 60s until she can reproduce stronger performance across multiple unseen papers. The study job then becomes precise: identify what made the 81 possible and make those conditions portable rather than emotionally promoting the 81 into a guarantee.
Calibration needs independent evidence
A target calibrated only against familiar practice can still be wrong. If every revision set repeats the same examples, the learner may become fluent with the study environment while remaining weak on unfamiliar tasks.
Useful calibration evidence therefore comes from conditions that resemble the future performance enough to be diagnostic: closed-book retrieval when the examination is closed book, unseen questions when transfer matters, timed work when speed is part of the requirement, and external marking when self-marking has become generous.
This does not mean every study session should imitate an exam. Learning and measuring are different jobs. But the forecast should periodically be anchored by a measurement that is not secretly carrying the answer.
Why prior success can distort the next target
The 2026 study found that a higher previous test score was associated with greater likelihood of later overestimation, while stronger prior knowledge and greater webinar engagement were associated with less overestimation. The authors discuss the possibility that situational success can inflate confidence while deeper engagement provides more information for self-assessment.
This should not be turned into a universal psychological law. The associations came from one course, a predominantly female sample, one subject and one goal format. Still, it suggests a useful check after a very good result: Was this evidence of a new stable level, or one successful sample?
A calibration protocol for real studying
1. State the aspiration separately
Write the long-run result you want. Do not dilute it merely because present evidence is weaker.
2. Build a current evidence window
Use several recent independent samples rather than the best or worst attempt. Record score, task type, assistance, timing, marking standard and dominant error families.
3. Forecast a range before a point
“Probably 66–72” is often more honest than “70 exactly.” The range makes uncertainty visible. A point target may still be useful for action, but it should not pretend the future is known more precisely than the evidence permits.
4. Identify what must change for the higher target to become plausible
Turn the aspiration–forecast gap into mechanisms: fewer algebra sign errors, stronger source-evidence selection, faster planning, broader vocabulary retrieval, better checking, or higher accuracy on a named topic.
5. Update after new evidence
A goal should not become sacred merely because it was written down. If three independent checks show the forecast has moved, update it. Calibration is a repeated inference, not a promise to your past self.
When an ambitious target is still useful
A target can be deliberately above the current forecast if its job is motivational or developmental. Coaches, teachers and learners often use stretch goals to create effort. The error is not ambition; it is confusing an intentionally difficult target with a prediction of what will happen.
If the target is 80 while current independent performance is 65, say so. Then the plan must specify which bottlenecks are expected to close the fifteen-point gap. If no mechanism can be named, the number is not yet a plan.
When underestimation is not automatically healthy
In the panel study, underestimating students tended to score higher than the accurately calibrated group. That does not mean students should deliberately predict low scores. Underestimation can come from caution, anxiety, uncertainty, unfamiliarity with the standard or a desire to protect against disappointment.
A learner who repeatedly predicts 60 and scores 85 is also poorly calibrated. They may underspend effort on opportunities, avoid suitable challenges or carry unnecessary anxiety. Calibration aims at useful correspondence, not optimism or pessimism.
Delayed and independent performance check
To test whether calibration is improving, make the forecast before seeing the next task’s answers or marking. Record both a point estimate and a plausible range. Complete the assessment independently. Then compare forecast with outcome and, more importantly, compare the predicted error pattern with the actual one.
Repeat across several assessments. One accurate prediction can be luck. Better calibration should appear as smaller and less systematically biased errors across changing topics and conditions.
Then run a transfer check: forecast performance on a task format you have practised less. If accuracy collapses, your calibration may be local to a familiar environment rather than a portable model of your capability.
For parents: ask for the evidence beneath the number
When a child says, “I think I’ll get 80,” the most useful response is not immediate praise or correction. Ask: “What evidence makes 80 plausible?”
A good answer may include recent unseen papers, teacher feedback, topic coverage and recurring errors. A weak answer may rely on hope, one easy worksheet or the last unusually strong result. The conversation shifts from confidence to evidence without punishing ambition.
For tutors and teachers: goals should generate decisions
A useful goal changes what happens next. If two students both write “75” but one is currently at 72 and the other at 45, they do not need the same intervention. The number becomes educational only when linked to a diagnosis, a time horizon and evidence that can update the forecast.
The existing eduKateSG page How Self-Evaluation Works owns the broader job of judging one’s own work against criteria. HSW-0245 has a narrower job: calibrating a future performance target against repeated evidence, while keeping aspiration, forecast and operating target distinct.
Misconceptions
“Specific goals always improve performance.” Goal-setting research is more heterogeneous than that. A 2025 systematic review of 60 higher-education studies found wide variation in how goal-setting activities were implemented and relatively few experimental tests of their effects.
“A high target proves high motivation.” A number can reflect motivation, poor calibration, social pressure, optimism or strategic stretching. The target alone does not reveal the mechanism.
“An accurate prediction causes the good score.” Not established by the 2026 panel study. Accurate forecasts and strong performance can share causes such as better knowledge, engagement or monitoring.
“Underpredicting is safer, so it is better.” Persistent underestimation is also miscalibration and can distort decisions.
“One mock exam is enough to calibrate.” One observation cannot tell you whether performance is stable. Use multiple samples and note the conditions.
Evidence and limits
The primary recent source is Boitcov, Adamovich and Getman, “Accuracy of learning goals and academic achievement: a panel study in an online course”, published 30 April 2026 in Psychological Science and Education. The study used five waves of panel data from 426 learners aged 16–18 in one online literature-exam preparation course.
Goal setting and goal accuracy were associated with academic performance, and prior knowledge and engagement were associated with lower odds of overestimation. The sample was highly imbalanced by gender, came from one course and did not include several psychological variables such as motivation and self-efficacy. Most importantly, the design supports association, not a clean causal effect of calibration.
For the wider goal-setting landscape, Martins van Jaarsveld and colleagues’ systematic review of goal setting in higher education, published in January 2025, included 60 studies and reported substantial heterogeneity in goal-setting implementations, with relatively few studies experimentally testing the effect of goal-setting activities. That review is a reason to resist turning one goal format into a universal recipe.
The return
A target is not valuable because it is written in bold.
Its value comes from what it coordinates: attention, effort, strategy and the next decision. For that coordination to work, the learner needs two forms of honesty at the same time—the courage to want more and the discipline to forecast from evidence.
Keep the aspiration. Measure the present. Forecast with uncertainty. Name the gap. Change the work. Then let the next independent result update the model.
That is what calibrated studying looks like: not smaller ambition, but better navigation.
Continue through the How Studying Works Numbered Series Reading Index, or return to the How X Works Hub.