College instructors--questions about exams and grading

[EKS]

College instructors, particularly those in science, math, and engineering, I have a few questions.

 

How many of you grade on a curve?  I know when I was in college many moons ago, all of the science (and, I think, math) professors graded on a curve.  This meant that sometimes 20% was passing and 30% was an A.  Is this still done?  If not, why not?  If grading on a straight scale is mandated by your institution or department, what is the reasoning?  

 

Also, how many items--problems, prompts, etc.--do you include on your exams?  I've studied a bit about psychometrics, and, generally speaking, the accepted wisdom is that the more items that are included on an exam, the more validity it has--meaning that the grade/score obtained gets closer to the truth about the examinee's true level of knowledge/understanding.  

 

Would you consider a two item exam to be a valid instrument?  How about a five item exam?  Would it make a difference if partial credit was granted or not?  

 

Have you ever received any training in how to develop a valid exam?

 

Note that by "exam" I mean something that makes up something like 25+ percent of a student's grade.

 

Thanks!

 

 

[regentrude]

We are not grading on a curve in our introductory classes.

1. With 600 students per class, it is impossible to find natural breaks where to put grade cuts, since points present themselves on a continuum.

2. Students appreciate knowing up front what the expectations are. They appreciate having certain numbers of points and fixed grade cuts, because that allows them to know where they stand.

 

I do not know what you mean by a "two item" exam. Our monthly exams consist of three worked problems and five multiple choice questions. Partial credit is given on the fully worked problems. The full problems test the three major concepts covered during the month.This allows a very nuanced judging of students' performance. Students have 3 exams, and each counts 20% of the overall grade.

I could construct a one problem exam that incorporates all the main concepts taught during the month; it is just harder to write such a  problem.

 

I have learned to develop good exams by learning from experienced instructors. It is very difficult to construct good exam questions that test precisely what you want to test and are phrased in a way that are absolutely unambiguous and state all assumptions.

 

Our final exam is multiple choice and has 20 questions. Students don't do particularly well on those because buy its very nature an mc exam does not offer partial credit.

 

 

[kiana]

I do not grade on a straight 90-80-70-60 scale in more advanced classes. I like being able to ask more challenging questions so I use a scale with lower cutoffs. But the scale is still available for the students to see and listed in the syllabus. 

 

I will also calculate everyone's grade twice (excel makes this so easy), once as listed in the syllabus and once with the comprehensive final weighted more heavily, and give the higher of the two grades. I don't announce this though, because when I did they figured they'd make up grades on the final and so slacked during the semester, and the failure rate was astronomical. But I do this because I believe that a student who learns the material well enough to get an A/B on a comprehensive final but did not do well on the midterm for whatever reason should not need to repeat the class. 

 

Items -- It depends on the class. In a lower-division, primarily computational class such as college algebra, there will be far more items, usually a bit under 20. In an upper-division, proof-based class, there will be fewer items (like 12-ish) and frequently I will present several proof problems and tell them to omit one or two of their choice. 

 

I did not ever receive any training but as a graduate student I observed and graded other people's tests for a long time before I wrote any of my own.

[G5052]

I've been an adjunct for 18 years, and both colleges that I've worked for strongly discourage grading on a curve.

 

The one I used to work for (smaller community college) actually required that you meet with your dean if you needed to curve a test. They had mandatory training for adjuncts, and a full-time faculty member met with you during the first semester to review your assignments and quizzes/exams.

 

The one I work for now is very large, and in the online component where I am, we use standardized quizzes and tests, so no curve.

 

My classes are very application oriented, so the bulk of their grades are projects. Currently I teach basic web design and multimedia, so they build websites all semester. One class has chapter quizzes, and both have 3 exams and a comprehensive final. Those run about 30 questions, a mix of multiple choice and short answer.

 

My background in math and computer science, and there just a few questions on test was not unheard of in the upper levels. Sometimes I had 2-pages proofs in graduate school, and there were also long essay questions where we outlined a design and security approach for a system. I had several math professors who gave no partial credit at all. 

 

 

Edited March 9, 2017 by G5052

[regentrude]

I want to add: the ideal exam that gets the most thorough evaluation of a student's performance would be a comprehensive oral examination by a fair examiner.

in an oral exam, the examiner can move between topics, and recognize where a student has gaps and can probe other areas to discover whether this is a singularity or the student's overall knowledge is lacking.

This is standard back home. My college transcript has only three grades from the comprehensive orals over 5 semester of math and 8 semesters each in theoretical and experimental physics, respectively.

[EKS]

I do not know what you mean by a "two item" exam. 

 

Sorry, that was probably ambiguous.

 

By two item, I mean two question/problem/prompt exam.

[regentrude]

Sorry, that was probably ambiguous.

 

By two item, I mean two question/problem/prompt exam.

 

Whether a two item test is a good one depends entirely on the problems.

I can write one single problem that combines multiple important concepts taught during my course, and I can get a thorough understanding of the student's level of mastery from this single problem if the problem is well designed.

 

One single multiple choice question would, of course, not allow any insight into the student's performance, since a grave conceptual error and a minor algebra mistake would both cause the student to get the problem wrong.

Edited March 9, 2017 by regentrude

[EKS]

Whether a two item test is a good one depends entirely on the problems.

I can write one single problem that combines multiple important concepts taught during my course, and I can get a thorough understanding of the student's level of mastery from this single problem if the problem is well designed.

 

One single multiple choice question would, of course, not allow any insight into the student's performance, since a grave conceptual error and a minor algebra mistake would both cause the student to get the problem wrong.

 

I assume the same would be true of a single problem that required working out but was only graded as right or wrong.  So it's a combination of writing good problems and scoring those problems in a manner that allows the instructor to determine whether the student's errors were minor or not.

[regentrude]

I assume the same would be true of a single problem that required working out but was only graded as right or wrong.  So it's a combination of writing good problems and scoring those problems in a manner that allows the instructor to determine whether the student's errors were minor or not.

 

absolutely.

[EKS]

Thanks all!

 

This is interesting--it appears that curving grades has gone out of style.  While it makes sense on one level--that there is, at least superficially, a more uniform standard that students are being judged by, it also puts pressure on the instructor to write a test that actually tests what he or she thinks it's going to.  Unfortunately, the students are the ones that pay if the instructor's test or grading scheme are lacking, and, as far as I can tell, there is really no recourse.

[regentrude]

This is interesting--it appears that curving grades has gone out of style.  While it makes sense on one level--that there is, at least superficially, a more uniform standard that students are being judged by, it also puts pressure on the instructor to write a test that actually tests what he or she thinks it's going to.  Unfortunately, the students are the ones that pay if the instructor's test or grading scheme are lacking, and, as far as I can tell, there is really no recourse.

 

Curving grades still happens at many institutions.

There are different ways of, and motivations for, curving.

An instructor may use exams that really distinguish at the top end, but have low averages, and she may curve the grades so that the best people - who may still have gotten only 40% of the exam correct - receive an A. That is a fair way of doing things. The only issue I have with this is that it does not allow the students to judge how they are performing compared to the grade standards until the end of the semester, which can be frustrating and stressful.

Then there is curving to ensure that only a certain percentage of students can receive a grade of A. This I find highly problematic, because it pits students against each other and in extreme cases provides an incentive to sabotage another student's work, since students benefit from another student's poor performance. I object to this on principle. I encourage cooperation in my courses and would consider any setup that creates a competition for grades to the detriment of collaboration undesirable.

 

In any scenario, whether curving or absolute grade cuts, students are to some degree at the mercy of the instructor constructing a fair test that tests what was taught and grading sensibly. It is, however,  not true that there is no recourse. A student can appeal the course instructor, and if he feels there is grounds for it, can submit a formal complaint to the department chair or the committee on capricious grading (yes, there is actually such a  thing) of the institution. Capricious grading charges can be brought against an instructor if he uses different standards for different students or punishes minor mistakes exceedingly harshly. Instructors should be advised by their departments to have clear, documented, grading standards and apply them uniformly across all students.

Edited March 9, 2017 by regentrude

[Bootsie]

I have been teaching at the college level for 30 years.  I do not curve individual exams.  At times I have allowed students to gain extra credit toward their exam grade by completing additional problems that the material covers.  At times I have given credit for a problem to everyone (this is more likely to happen with multiple choice questions) if I determine that there was an issue with the question itself.  I now teach a class which has a requirement that the class GPA fall between 2.75 and 3.10.  I may have to curve final grades to ensure that I fall within that requirement.  

 

As far as how many exam questions, that varies a good deal according to the time allotted for the exam (50 minute class versus 1 hour 15 minute class), the material being covered, and the length of the individual questions.  If I have a few multi-part questions, I try to grade such that a mistake on one part does not cause someone to lose points on the remaining parts of the question.

 

More questions can lead to a more valid test score, but not if the extra length results in the test measuring more about endurance or reading speed (unless that is part of what is being tested.)  

 

I have attended seminars about testing (none that have been required), but most has been learned from mentors.  I have also been involved in the development and grading of a professional designation exam and worked with the psychometricians.  

 

I have heard things like, "if only 10% of the test takers get a problem correct, it is invalid."  I do not agree with that.  If only 10% get a problem correct--and those 10% are among the lowest overall exam scorers, there is a problem.  If it is the top 10% it is a good question to distinguish those who are really at the top of  the distribution.   These questions can be just as valid as a question that only 10% of the class misses. 

 

[Reefgazer]

Biology instructor at a community college here.  I do not grade on a curve; no one grading method is mandated by my school, so how I assign grades is my choice.  I can think of no co-workers grade on a curve, although I am sure there are a few.  I personally don't like the grade on a curve practice.  If I give a test or assignment with standards that meet an A, then that is what I expect for an A.

 

I have roughly20 written questions on an exam, along with an essay.  I do not give multiple choice tests, although I do give multiple choice quizzes.

 

I would consider a 2 item exam to be valid if it required a longer, multipart answer that encompassed many aspects of the unit.  In such a case, though, I would expect it to be graded as partial credit, but I could also accept it if it were pass/fail.

 

I have not had any formal training in developing an exam.

College instructors, particularly those in science, math, and engineering, I have a few questions.

 

How many of you grade on a curve?  I know when I was in college many moons ago, all of the science (and, I think, math) professors graded on a curve.  This meant that sometimes 20% was passing and 30% was an A.  Is this still done?  If not, why not?  If grading on a straight scale is mandated by your institution or department, what is the reasoning?  

 

Also, how many items--problems, prompts, etc.--do you include on your exams?  I've studied a bit about psychometrics, and, generally speaking, the accepted wisdom is that the more items that are included on an exam, the more validity it has--meaning that the grade/score obtained gets closer to the truth about the examinee's true level of knowledge/understanding.  

 

Would you consider a two item exam to be a valid instrument?  How about a five item exam?  Would it make a difference if partial credit was granted or not?  

 

Have you ever received any training in how to develop a valid exam?

 

Note that by "exam" I mean something that makes up something like 25+ percent of a student's grade.

 

Thanks!

 

[Bootsie]

Thanks all!

 

This is interesting--it appears that curving grades has gone out of style.  While it makes sense on one level--that there is, at least superficially, a more uniform standard that students are being judged by, it also puts pressure on the instructor to write a test that actually tests what he or she thinks it's going to.  Unfortunately, the students are the ones that pay if the instructor's test or grading scheme are lacking, and, as far as I can tell, there is really no recourse.

There is not necessarily more pressure to write a test that test what I think it is going to because I don't curve.  Curves should not be used to mediate poor test design.  For example, I could write a multiple choice test with nonsensical answers and have the average student receive a 25--but there would be a distribution around the 25 even though students were simply guessing.  I could then curve the grades, but the grades would be meaningless.  

 

I think curves are more likely to be used when a philosophy of grades should separate the top students from the average and the below average.  % correct cutoffs are more likely to be used when the philosophy is that students who clear a particular hurdle should be rewarded with a particular grade.  If all students clear the hurdle they all get the grade; if none clear the hurdle no one gets the grade.  

 

I always find students want a curve to raise their grade but are never in favor of a curve that might lower their grade.

[kiana]

I always find students want a curve to raise their grade but are never in favor of a curve that might lower their grade.

 

Yeah. What they want is a ramp, not a curve. 

[Bootsie]

One of the reasons I do not like grading on a curve is that it makes a student's standing based upon that student's cohorts.  For a number of reasons, student ability is not normally distributed across sections.  One example is when a honors class is offered at the same time as a non-honors required class one semester.  None of the honors students are in that section, but they are overly represented in a section the next semester.  

 

In many classes I find that I have a bi-modal distribution--It is hard to devise a curving scheme that fits that well.  Do I add the same number of points to every student's exam so that the highest is a 100?  So that the average is 75?  So that the minimum is 50?  Do I fit to a bell shaped curve?  How do I account for a low average grade (although several students did well) when the low average is due to several students receiving severely low grades because they copied a number of answers from another student's paper (not realizing that they had different numbers in their problems)?  In a relatively small class, the poor performance of one or two students can significantly impact the average--but should that lead to a curve?  

 

 

[FaithManor]

When I was teaching music history and theory, I never graded on a curve. The grading scale was part of the syllabus with a detailed explanation of the points awarded. It was simpler to make expectations known up front, and stick to it.