Are Final Exams Necessary

The Case Against Traditional Final Exams

I spent seven years teaching introductory computer science at a community college before moving into curriculum design, and the final exam has always been the most contentious part of that job. People argue about it endlessly, and honestly, most of the arguments just circle the same points without actually solving the problem. The core issue with final exams isn't that they're evil. It's that they measure something very specific — the ability to recall and apply information under time pressure in a controlled setting — and then everyone pretends that skill is the same as learning or competence. It isn't. A student who can solve differential equations on a three-hour test under fluorescent lights is not necessarily better at engineering than someone who designs robust systems over weeks of iterative work. The test is a proxy, and it's a blunt one.

Why They Persist Anyway

Final exams survived because they are defensible. They are cheap to administer at scale, hard to cheat on without elaborate proctoring infrastructure, and produce a single data point that administrators can sort into a gradebook. That last part matters more than anything else. When you have ten thousand students and six faculty members, you need a system that reduces complexity, and a two-hour closed-book exam does exactly that. The alternative — continuous assessment through projects, portfolios, and participation — is better at measuring actual understanding. I ran a version of that in a graduate seminar once where the final grade was entirely project-based, and the learning outcomes were clearly superior. It also required approximately four hundred hours of additional grading time that didn't exist in the budget. So the exam stayed.

What Research Actually Says

The research on final exams is maddeningly inconsistent because the studies measure different things. Some show that exams improve long-term retention when they are high-stakes, which makes sense — students study harder when there is something at stake. Other studies show that the anxiety induced by exams actually impairs performance for a significant subset of students, sometimes 20 to 30 percent, and that this subset disproportionately includes first-generation college students and people with documented test anxiety. A 2018 meta-analysis in Review of Educational Research found that cumulative final exams produced modest gains in retention compared to non-cumulative ones, but the effect size was small — around d = 0.25. That means the difference between giving a cumulative final and not giving one is real but not dramatic. Meanwhile, student satisfaction with exams drops sharply regardless of format, which suggests the problem isn't the exam itself but the experience of taking one.

Practical Alternatives That Actually Work

If you are redesigning a course and considering whether Are Final Exams Necessary, here is what I have seen work in practice across multiple institutions. Take-home exams with a narrative component. Instead of abstract problems, give students a realistic scenario they need to work through over 48 hours. I built a final for a data structures course where students had to debug a broken codebase with documented production incidents. It took three times longer to grade but revealed far more about who actually understood the material. The tradeoff is that you need a sufficiently complex scenario, and writing one well takes about forty hours of development time. Replace the final with a capstone project. This works best in upper-division courses where students have already demonstrated baseline competence. The project should have clear rubrics and milestone checkpoints so you are not grading a single artifact at the end of term. One of my colleagues at the curriculum design firm replaced finals in our advanced algorithms course with a semester-long optimization project, and pass rates went up by twelve percentage points while student engagement scores improved noticeably.

Cumulative but open-resource exams. If you keep the exam format but allow any resources — notes, textbook, calculator, even AI — you change what the test measures from memorization to application. I ran this in a certification prep course and found that students who performed poorly on closed-book exams actually performed better on open-resource ones, which suggested the issue was anxiety and recall speed, not knowledge gaps.

Where Exams Fail Completely

Let me be direct about the situations where final exams produce useless or actively harmful data. They fail in courses where the learning objective is collaboration. A computer science final written individually tells you nothing about a student's ability to contribute to a team, review code, or communicate technical decisions. If your course involves group work and you grade with a solo exam, you are grading a different skill than the one you taught. They fail when the stakes are too high relative to the assessment window. A single three-hour exam for a course worth four credits compresses roughly fourteen weeks of learning into a narrow time slice. Any factor — illness, a bad night's sleep, a family emergency, simple bad luck with question ordering — can distort the result significantly. I watched a student in the top quarter of the class miss the exam due to a hospital visit and fail the course despite having an A going in. The university allowed a makeup but required documentation that took two weeks to obtain, by which point the semester was over. That is not an exam problem. That is a policy problem wearing an exam costume.

They fail in early-term assessment for remedial courses. Students in developmental math or English composition are often still building foundational skills. A single high-stakes test at the end punishes the learning curve rather than measuring the destination. Portfolio assessment works better here because it shows growth trajectory, not a snapshot.

What to Do If You Keep the Exam

Sometimes you keep the final exam. Budget constraints, accreditation requirements, or departmental policy force your hand. In that case, make it less harmful. Make it cumulative but weighted at ten to fifteen percent of the grade, not thirty or forty. The weight determines how much anxiety students feel, and fifteen percent is enough to matter without being catastrophic if something goes wrong. Give students a practice exam under real conditions two weeks before the actual one. I started doing this after noticing that students who took the practice exam scored eight points higher on average, which suggests a lot of the variation was test-taking familiarity rather than content knowledge. The practice also reveals which concepts students are actually struggling with before the real stakes hit.

Offer a choice of exam formats if possible. Let students pick between a traditional written exam and a project submission. You will get pushback from administrators who worry about grade inflation or fairness, but the research supports this approach, and students who can demonstrate competency through multiple modalities tend to perform closer to their actual ability level. The conversation about whether Are Final Exams Necessary will continue because the answer depends entirely on what you are trying to measure, what resources you have, and who your students are. There is no universal solution. The best courses use exams as one tool among several, not as the final verdict on what someone learned.