As any educator can tell you, many students believe that tests are the course-and focus nearly exclusively on studying for the exam rather than for acquiring their own knowledge. This leads professors to complain that students are more interested in grades than in education.

What should tests cover?

Focus on important course objectives that represent the major part of lectures and homework assignments-and share that information with your students so there are no surprises. If it's on a test, students will think it's important. If not, the professor must be wasting their time in lecture! This partly explains why students commonly believe that theory-which is often not tested-is a waste of time.

What types of questions?

An occasional comprehension question ensures that students understand the material in addition to being able to calculate. A comprehension problem requires more than quoting the textbook, but calculations are not required. For example, in a heat transfer course, ask students to explain how ice can form on a car windshield even though the air temperature never drops below zero degrees centigrade. Straightforward application problems are appropriate for short quizzes, while longer analysis problems with several parts are appropriate for longer tests. Don't wait until the senior year to include synthesis and evaluation problems.

How many tests?

More-frequent testing has some surprising pluses. By studying for tests, students spend more time on task and hence increase learning. Test anxiety is also reduced-particularly if you inform students that their lowest test score will be discarded. We suggest approximately seven tests (including a final) per semester for first-year students and sophomores, about five for juniors, and three or four for seniors. The downsides of increased testing, of course, are reduced lecture time and increased workloads for the professor and the TAs. An alternative strategy is to give more short quizzes and only a few tests, which will still keep students studying throughout the semester.

How hard?

Engineering professors often write tests that are too long and do not discriminate well between students who understand the material thoroughly and those whose knowledge is more superficial. We believe that many of the good students should earn a score in the 90s or even a 100. If no one can score that high, the test is too long, too hard, or too tricky. The counter-argument is that the professor wants to challenge students, but this argument forgets that in the workplace, our graduates will be expected to do essentially perfect work. They need some practice in being correct before they graduate.
If students who know the material earn high scores and students who don't know it receive low scores, the distribution will be bimodal. If several scores are in the 90s with a few 100s, then low-scoring students cannot rationalize their failure as the professor's fault-something that is easy to do if the test average falls in the 20s to 50s range. Also, there is little discrimination between poor and average students when the test average is low, whereas tests that discriminate are likely to result in more A's and more F's.

How long?

 A test that is a sprint also does not discriminate well, because it confuses speed with knowledge. The preferred bimodal distribution starts to look like a normal distribution, and students who work more slowly are unfairly punished. Remember that speed is much less important on the job than it is in school. By solving the test before using it, the professor can estimate the appropriate length: first-year students and sophomores need about five times longer, juniors about four times, and seniors three to four times.

Open or closed book?

 Engineering practice is mostly open book, and such tests essentially eliminate certain types of cheating. A reasonable compromise is allowing students to write anything they want on a 3 x 5" card that can be brought to the test room. The process of summarizing the material is also a very useful study exercise.

Phillip Wankat is a chemical engineering professor at Purdue University; Frank Oreovicz is an education communications specialist at Purdue's chemical engineering school. The authors welcome readers' feedback. You can reach them via e-mail at wankat@ecn.purdue.edu  and oreovicz@ecn.purdue.edu .