Should final exams be comprehensive?

Question

I teach in the school of sciences and engineering in the university where I work. I have asked professors several times if the final exams have to be comprehensive, and I usually get mixed answers.

So I want to ask, for a course with two midterms and a final:

1. If there is some criteria or organizations that require the final exams to be comprehensive (for instance, does ABET accreditation or legal statute require it?), or does it depend on the school or the professor?

2. What are the pros and cons of having a comprehensive final?

3. Will your answers to the above questions be different if the course components were 2 best out of 3 midterms plus a final, or if the course was a graduate course instead of undergrad?

Answer 1

All of my final exams are cumulative, for a couple reasons:

1) I've always felt odd about non-cumulative exams. It seems to send a message I'd rather not send: “It's okay to forget about what you learned earlier in the course – you won't need that any more.”

2) Often, what gets taught later in the course builds on earlier concepts. If A lays the foundation for B, and B lays the foundation for C, it can be rather challenging to test on C without testing on A and B also.

That said, I try very hard to test on higher-level concepts rather than on minutia, so a cumulative final fits my teaching style. I expect my students to be able to speak intelligently to the main themes of the course. I've told students many times: “Come to class ready to learn, and be ready to engage with the material via the in-class discussions. That's the best way to prep for the exams.”

My goal is to structure my exams so that students who have paid attention in class and learned the material through other assignments should do quite well on the test – without the need to cram or commit the material to short-term memory. “By the last week of the term, you either know this stuff, or you don't,” I tell them. Many of my exam questions are essay questions that require students to analyze a scenario, and synthesize material from different parts of the course.

That might not always work, depending on the nature of the material; some material doesn't lend itself well to such questions. This approach may not be scalable, either. (It's not uncommon for me to have six or seven pages of essay questions per exam. Most of the time, I have between 15 and 30 students per course, which keeps grading manageable. If I had 50 or 60 students, however, I might have to rethink this approach.)

As a footnote, most of my courses are 400/600-level courses. I don't know if the style I've outlined would be a good fit for freshmen and 100-level material.

Answer 2

To address 1, I've never heard of any such policy on cumulative exams. In my experience, such pedagogical decisions are made solely by the course instructor. They make the decision as to what material appears on the final exam, or even whether to have a final exam at all.

In the case of large multi-section coordinated courses, this decision might be made by the course coordinator (especially if all sections give a common final exam) or by agreement of the various instructors. There might be some expectation that they will respect precedent unless they feel strongly that it should change, in which case they might be expected to discuss it at the department level.

Otherwise, I think any explicit rule about the content or format of exams would be seen as a serious infringement on the instructor's autonomy, which is a central aspect of university teaching.

Answer 3

At my large public research university (in the US):

In physics, exams are nearly always non-cumulative. The finals are treated essentially as a third midterm, none of which are cumulative.

In math, the it depends on the professor and the course. If it is one of the "common"/high-enrollment courses that students in e.g. biology, chemistry, also have to take, then the final is usually cumulative. In courses designed for upper-level math majors only (e.g. topology, graph theory) it is up to the professor, but usually cumulative (usually the courses are very linear; it would be impossible to pass a test over the final material without knowing the material from the midterms).

In humanities, usually not cumulative. Or if cumulative, trivially so (e.g., one or two basic questions about the early material).

edit: I should clarity, it's always up to the professor, I'm just outlining general departmental tendencies

Answer 4

I'm responding to #2 and 3, as there are no rules regarding #1 that I'm aware of.

Pedagogically, there is evidence that repeated testing improves learning. Most of the literature is via authors Karpicke and Roediger. Here's a list of resources from a research project at Washington University, and a rather overwhelming overview of student learning and good teaching by John Kihlstrom at Berkeley.

I don't see why the benefits of a comprehensive final would be lessened by a "two out of three" midterm organization or different for undergrad vs grad level students.