If you are a professor on the semester system, you have probably just finished administering your first midterm for the spring, or are about to.
I used to go for long evening midterms and a long final (2 hrs each), and decided these were a nightmare to schedule and a nightmare to grade. After every exam, I would just lose the will to live with all the grading, and I would fall behind on all my other work, plus there is always someone who had to skip the exam for whatever reason, so I often had to administer and proctor make-up exams.
I have recently switched to more frequent exams — three in-class midterms and a final, 50 min each — and my will to live has been holding fairly steady during grading. The students love the higher exam frequency, as they get info on how they are doing more often and their final grade is not so strongly affected by any individual midterm, which alleviates stress. Also, as most professors at my school like to give two midterms, students get swamped during certain weeks as all exams fall at the same time; with three instead of two midterms, we are off-cycle with the other exams and students have a little more time to prepare.
I used to do weekly quizzes, but that simply ends up being too much grading for comparatively little gain. If the students know the quizzes carry very little weight, they are usually not overly compelled to study just for the quizzes. Those who study for the quizzes tend to study and do well anyway. I could make quizzes carry more weight, but then they become like exams, so why have full midterms and not just quizzes in that case?
Anyway, so far an exam about every 4 weeks (or every 3 HW assignments) seems to work well; it provides a balance between the exams being short enough to be done in class and not be too painful to grade, frequent enough to give plenty of feedback to students, yet each is long enough to give meaningful info on the student’s mastery of the material.
I have a confession — I actually quite enjoy creating problem sets, especially for exams and projects. That’s a pretty cool and creative part of the job. For undergrads, I don’t go too crazy with creating original homework because undergrads like to use the textbook in order to feel they have gotten their money’s worth (also to copy the solutions of others); when I assign textbook problems, I explain to students that the homework carries very little weight and is primarily for their benefit, to help them practice, and then the HW does get graded pretty liberally. In some classes it makes sense to have a project component or a programming assignment, which are fun for me to create and for the students to do, and these carry more weight. For graduate students, I have original homework problems and projects, and more often take-home exams over in-class ones. Grad courses are fun as you get to work with examples that are not quite well-behaved mathematically, and with a little programming you can get great insight into some nontrivial but highly instructive problems… But I can really get creative with exam problems, make them instructive and interesting as well as an assessment tool.
I would say a test is of optimal difficulty if the average is 70-80%. There are variations from class to class, but this has generally held, with perhaps an occasional one in the 60-70% range. When I design, I plan for the 1:3 ratio of the times needed to take the test by me and the students, i.e. I need to be able to do it about 15-16 min for a 50 min exam. In an occasional class, I have had to go with a 1:5 ratio, but luckily not too often.
How do you test your students? What are some aspects of exam creation/administration/grading that you particularly like or dislike? How are the tests in the biomedical sciences? How about the social sciences or humanities?
xykademiqz 
Re: quizzes – I have found that students often don’t understand what I am looking for in a test question. It’s not that they don’t know the material [though some don’t!] but that they don’t know how to write a short answer to demonstrate that they know the material. So I started doing weekly quizzes to give them regular feedback about what grading would look like. I find that exams in classes with quizzes do a better job of actually identifying who knows the material and who doesn’t, rather than who knows how to take my tests and who doesn’t.
Physical Sciences here, all undergrad classes. Giving exams outside of the normal 60-minute class period is not really feasible here, and the standard final exam times are 120 minutes. In most of my classes, I had migrated to 3 “mid”-terms and a final, which I prefer, but my this year I’m up over 100 total students per semester, so I’ve dropped back to 2 mid-terms. We have to submit deficiency reports for students during the 7th week, so I always have an exam by the 5th or 6th week. Luckily, it looks like I will be back to my normal 60-80 students in the coming years. Not coincidentally, I’ll be teaching more upper-division classes, which will allow me to be more creative on exams.
A lot of my 120-minute final exams are basically a 60-minute exam over the new material covered since the previous exam, and a usually less than 60-minute part covering material from the previous exams. I usually pick problems that didn’t go very well the first time.
I weight the HW fairly low (around 15-25%), but I also do an in-class review session before every exam based on review problems I post a couple days earlier. In turn, those problems are often based on HW problems. For grading reasons and “extra time for them to think about my sometimes clever problems”, I generally limit my classes to 10 homework sets per semester.
I taught a 150-minute once a week course a couple times and found that giving 30-minute in-class quizzes every other week and one 120-minute midterm worked out pretty well.
Engineering here (bioinformatics and applied circuits, mainly). I’ve gotten away from giving exams—for bioinformatics I have weekly programs or papers and for circuits I have weekly lab design reports. Grading programs and papers is more effort than grading quizzes and exams, but tells me a lot more about what students can really do. I’m not interested in their ability “to write a short answer to demonstrate that they know the material”, but in how well they can create something out of the knowledge they supposedly have.
Incidentally, in the few times I’ve given exams, I prefer to have a mean score around 50% and a standard deviation of around 20%, so that I get the maximum information out of the short exam time. If students are averaging 80%, then most of the questions are telling me nothing about the students—those exams are optimized for separating the failures from the complete failures, rather than giving information about the middle of the class. Sometimes I do mis-estimate the skill level of a class, and make an exam that is too hard (if the average is only 30%, then again most of the test tells me little about the class).
I teach undergraduate biochemistry. I typically do 4 exams with the option to drop the lowest score (i.e. completely miss the boat once) and a comprehensive final worth 2 exams in points. The average on the hourlies is usually 70ish (60 early in the semester, 80 later) and the comprehensive 65ish, so I’d say it’s going ok. These are all essay tests. Draw things, write short answers, think critically.
We also do lots of lab reports and projects each semester. I’m at a PUI, so the workload is largely teaching, but when my enrollment gets above 60 students/semester, I suffer.