An obvious criticism of the multiple-choice question (MCQ) format is susceptibility to lucky guessing. A subtle criticism is its all or nothing scoring. MCQ makes no attempt to discover what unsuccessful candidates do know about an item. An alternative method for MCQs is to ask each examinee to mark all options that are definitely incorrect.
Scoring for "eliminate the incorrect" is simple. When an examinee marks the "correct" option as incorrect, the score is 0. Otherwise the score is the number of "incorrect" options marked as incorrect. For a 5 option item, possible scores are 0 (no options marked or "correct" option marked) to 4 (all "incorrect" options marked). This partial credit approach discourages guessing, and enables the examinee to gain credit for whatever partially correct solution is obtained. Here is an example with a score of 3:
Which cities are definitely not the capital of Peru? [x] A. Bogota [x] B. Caracas [ ] C. Colon [ ] D. Lima [x] E. Montevideo
In a 60 item verbal comprehension test the average point-biserial correlation for conventional scoring was .41. For revised scoring, the average item score-total score correlation was .51. Person separation reliability improved from .91 to .95 and separation from 3.3 to 4.5. This improvement is equivalent to adding 50 more items to the test.
The next stage is to construct a measurement model for this response structure and to explore alternative scoring strategies. I welcome comments.
MCQ - Oust the Incorrect, Y-Y Wu … Rasch Measurement Transactions, 1991, 5:2 p. 143
The URL of this page is www.rasch.org/rmt/rmt52d.htm
Website: www.rasch.org/rmt/contents.htm