The instructor should compare the answer key to several of the trainees' responses to a
question. Some trainees may take a different approach from what the answer key
anticipated and still be correct. If so, an alternate correct answer will need to be added to
the answer key. If the key is changed, all tests should then be regraded using the revised
Periodically review the answer key. It is easy for an instructor's standard to change after
several tests are scored. Reviewing the answer key will help protect against distraction from
the standard. Also, by occasionally reviewing those items scored earlier, the instructor can
confirm that the standards are being applied consistently. Even when applying these
measures, some inconsistency is inevitable. One problem is how an item response is
graded following several good or several poor responses. The tendency is to score the item
low if it follows several high scores, or to score the item high if it follows several low ones.
Shuffling the tests between review of test questions, while not eliminating the problem,
allows these effects to be offset by random sequencing.
Score each item separately. Each test item should be scored for all tests before the next
item is scored. Scoring one item at a time allows the instructor to concentrate on just one
standard. This increases consistency when assigning points or categorizing items.
Avoid interruptions when scoring responses. The bias an instructor has toward an essay
item may change. If a bias exists it should be consistently applied to the responses of all
trainees. (For example, an instructor may be irritated one afternoon and calm the next
morning) By scoring all response sets at once, if a bias exists, its effects on trainee scores
will be consistent.
Provide comments and make corrections on the actual test. A trainee who does not receive
full credit for an answer will want to know why. Appropriate comments can explain the score
received. If trainees are to learn from their mistakes, they should be told what errors were
made and how to correct them. Another value in providing comments is the ability to tally
the various comments and analyze the results from test item improvement.