I spotted that some universities do not have a comprehensive exam for their PhD students while it is standard at most universities (specifically for computer science). For what reasons do some universities drop the requirement of comprehensive exam? What is the impression towards someone who has graduated from no-comprehensive department?
The primary reasons for having a comprehensive or qualifying examination are as a measure of quality control, and as a means of "population control."
The first ensures that students meet minimal criteria for promotion to PhD candidacy. This allows, in part, PhD programs to take a few candidates who they feel could be successful, but are not 100% confident thereof.
The second is used if departments admit more incoming students than they have funding to support as PhD candidates. Then the exam is used to "cull the herd," and keep the candidates the faculty wants to continue to fund.
My field used to be the kind of program where everybody had to do a qualifying examination. But, in recent years, this trend has started to swing away from comprehensive exams to research prospectuses. There is no stigma associated with either approach, though; both are considered acceptable.
Practically, departments choose not to have a comprehensive exam for a variety of reasons:
- They may feel that they need neither quality nor population control.
- They may not want to spend the time required to prepare, administer, and grade such exams (particularly if there are large numbers of students as well as time-intensive components, such as oral sessions).
- They may feel other means of assessing performance are superior (for instance, a research dossier and oral proposal).
My department does not have a comprehensive exam. The most frequent reason I have heard is that we don't need it. On the other hand, some other departments in our university have comprehensive exams and I have heard that they are primarily used to weed out students after the first year.
I will venture to say that there is absolutely no impression towards graduating from a no-comprehensive program. Research job committees look at publications, letters of recommendation and research experience. Graduates from my program have found jobs wherever graduates from similar programs find jobs.
Of course there are many advantages to having a comprehensive exam as are outlined in many of the other answers here, but there are also many reasons why departments don't have them.
A comprehensive exam creates more work for professors (they have to grade it, write it, and often even hold study courses for it) and potentially, if the program only admits good students, or if weaker students drop out due to poor performance in coursework, or unsatisfactory progress towards research, some professors consider the exam to not be worth the extra work (i.e. an unnecessary piece of bureaucracy).
In addition for some fields, research topics are very diverse, and course requirements are effectively all electives. In such cases, it would be unclear what course work a comprehensive exam would pull from.
Even if this is not the case, not all professors will agree with the definition of what a "core subject" in the field is. This is a bigger problem than you might think. Say as a professor I am really excited that I finally got a great student in my subfield and we are doing great research. Now say he fails the qualifying exam in some other topic that is viewed as a "core" topic, when my subfield is not considered as important. You can bet I will be bitter about that. If the professors can't settle on what topics are core (a political not just academic process), then sometimes the only solution that can gain enough approval from most of the professors is to not have a qualifying exam at all.
At some programs that do not have comprehensive exams, the advancement to candidacy exam can act as one, by including questions about topics from coursework that are only loosely related to the proposed research (in addition to questions more directly related to your thesis proposal or research presentation). These advancement to candidacy exams can require just as much studying as one would typically do for a qualifying/comprehensive exam. There are advantages and disadvantages to this approach. One advantage is that having the comprehensive exam material in the advancement to candidacy exam makes you go back through your course work and think critically about how each topic relates to your area of research. Your research allows you to review the material from a new, perspective that allows you to draw new interesting pieces of understanding.
In other fields, a comprehensive exam is viewed as a waste of time because the final exam in a set of core courses effectively act as a comprehensive exam. Usually, some students do not make the required grade and drop or retake the class (similar to a comprehensive exam). The big advantage here is that the professor who taught the material writes the exam, this saves time on the professor's end and also prevents needlessly failing students who might get hung up on inconsistencies between two different exam writers.
I got my PhD many many years ago in CS, and in my case the department had a 4 part/4 day written + a 2-3 hour oral a year or so later. The written was the same for everyone, 1 part CS theory (computability, formal languages etc), 1 part math (formal logic, combinatorics etc.), 1 part software (compilers, OS), 1 part hardware (mostly the content of the Hennessy and Patterson book, 1st edition at that point). All at the graduate level (they subsequently dumbed-down the exam because the pass rate was too low and some faculty were loosing favored students who could not pass)
I spent a good 6 months with my head in books and in study groups studying for that exam, and though I was happy to be done with it, I didn't see the great value at the time.
I do now. Even now, with my "background information" 20 years out of date (not really), I find that I'm often using some piece of what I learned in that half-year to explain something to my students or to conceive of how I might implement some research or project idea--not all of it, mind you, but I have used all 4 parts. Students ask my questions expecting to get answers that draw from various parts of the discipline, and I'm grateful that I have that knowledge.
In a recent committee meeting to vet faculty candidates, I was discussing courses we'd want then to teach and one member of a committee [a dean, not a CS faculty member] asked me about the common knowledge that all PhDs in CS would have, and I was lamenting the fact that many schools do not have rigorous comps, and so we can't make assumptions about what background knowledge they have. That said, I usually don't bother looking at the comps page on candidate's universities, rather we ask a bunch of technical questions in the phone interviews
As a side note about "admissions committees" and why they might do a good job and still need comps, many CS departments take a large number of people with degrees in related disciplines (Math, EE etc), and while I think they make great computer scientists, they lack something if they don't have the opportunity to study the foundations of CS
I simply would not have done a PhD if I had to take comps. To me it was extremely important to finish as quickly as possible. With starting salaries in Engineering at a pretty high level, every year I was in school I felt was like leaving money on the table. I suspect that many universities also find that students in high demand areas feel the same and there is a scarcity of grad students, particularly local students.