What an assessment is trying to measure
Every assessment is a sample. A three-hour exam cannot cover a whole course, so it samples some topics and some kinds of thinking and hopes the result reflects the rest. Two questions decide whether the sample is useful. Is it valid, meaning does it measure what you actually care about? And is it reliable, meaning would a similar learner get a similar result on a different day or from a different marker?
Comparing methods is easier once you ask these questions of each one. No format is best for everything, and the evidence on which methods give the most accurate picture of learning is mixed, so what follows is about trade-offs rather than winners.
Timed exams
The classic invigilated exam asks learners to answer under controlled conditions, usually on paper and without notes.
Strengths. Everyone faces the same conditions, and it is reasonably clear who produced the answers. Exams can test recall and quick reasoning, and they suit subjects where fluent skill matters, such as arithmetic or recognizing vocabulary.
Limits. Time pressure and anxiety can depress performance without reflecting knowledge. Exams also tend to favor what is easy to write quickly. Marking long answers by hand is slow, and markers can differ in how they judge the same script.
Essays and extended writing
Essays ask learners to build an argument, use evidence and explain their reasoning.
Strengths. They show how a person organizes ideas, which a short-answer test often cannot. They can be set as take-home work, which reduces time pressure and allows drafting.
Limits. Marking is subjective unless there is a clear rubric, and even then two markers may disagree. Essays take a long time to mark, so feedback often arrives late. For take-home essays, authorship is harder to confirm, which links to the wider topic of avoiding plagiarism and to the integrity questions covered under online assessment.
Online quizzes
Online quizzes usually mean multiple choice, true or false, matching, numeric entry or short text, marked automatically.
Strengths. They give instant feedback, scale to thousands of learners, and can be retaken. Low-stakes quizzes pair well with retrieval practice, because the act of answering helps memory in its own right. They also produce data about which questions many learners miss.
Limits. Well-written questions take skill. Multiple choice can reward guessing: on a quiz of 20 questions with four options each, a learner choosing at random would be expected to get 5 right (20 multiplied by one quarter), which is why question quality and sufficient length matter. Automatic marking also struggles with nuance, so quizzes mostly test recognition and recall rather than extended reasoning, although well-built items can reach further than people expect.
Projects, portfolios and practical tasks
A fourth group sits between the other categories. Learners produce something: a report, a design, a recording, a small program, a collection of work over time. This is often called authentic assessment because it resembles real tasks. Projects can reveal skills that exams miss, but they are time-consuming to mark and need clear criteria to be fair. Group projects add the difficulty of judging each member’s contribution.
A side-by-side summary
| Method | Good for | Weak at | Typical marking effort |
|---|---|---|---|
| Timed exam | Recall, speed, controlled conditions | Depth, reducing stress effects | Moderate to high |
| Essay | Argument, use of evidence | Consistency between markers | High |
| Online quiz | Frequent checks, instant feedback | Extended reasoning, guessing | Low once built |
| Project or portfolio | Applied skills, real tasks | Comparability, authorship, time | High |
The marking effort column describes a typical pattern, and it varies with class size and tool support.
Does moving online change what is measured?
Taking an exam on a screen does not by itself change the skill being tested, but several things shift.
- Conditions. At home, rooms, devices and connections differ, so fairness takes more planning. Not every learner has a quiet space or a stable connection, a point tied to the digital divide.
- Integrity. Remote invigilation tools exist, but they raise privacy and fairness concerns of their own. Many educators respond by redesigning tasks rather than watching harder, for example by using open-book questions that ask for explanation rather than lookup.
- Feedback speed. Online systems can respond in seconds, which helps learning if the feedback is specific and learners act on it.
- Accessibility. Digital formats can allow screen readers, adjustable text and extra time more easily, but only if the tool is built with accessibility in mind.
Combining methods
Because each method has blind spots, many courses use a mix. A common pattern is frequent low-stakes quizzes for practice, one or two larger pieces of writing or a project for depth, and a smaller number of controlled tasks where identity and conditions matter. The right balance depends on the subject, the age of the learners and what the qualification is for.
A few questions help when designing a mix.
- What should a learner be able to do at the end, and is that written down as clear learning objectives?
- Does each task actually require that skill, or only something nearby, like fast typing?
- Can learners get feedback early enough to use it?
- Is the workload for markers realistic?
- Would a learner with a disability or a poor connection be unfairly disadvantaged?
Reading claims about assessment carefully
Articles and vendors sometimes claim that one format produces better learning outcomes than another. Look for what was actually measured, how large and varied the group was, and whether the comparison was fair. Results often depend heavily on the subject and on how well the assessment was designed, so a general ranking is rarely trustworthy. A modest, well-built assessment used thoughtfully usually beats an elaborate one used carelessly.