Overcorrection for Guessing in Multiple-Choice Test Scoring

Abstract

THERE ARE SEVERAL procedures for obtaining scores on multiple-choice tests. The number of cor rect responses, the number of correct r e sponses minus the number of wrong responses, the number of correct responses minus one-half the wrong re sponses, the number of correct responses minus one-quarter the wrong responses, are a few of the methods used in determining scores. Generally the number of correct responses is reduced by an amount equal to the number of wrong responses divided by one less than the number of options per test item. The formula is S=Rn-1, where R is the number of | correct responses, W is the number of wrong reI sponses, and N is the number of choices per test item. Doppelt, in Test Service Bulletin No. 46, January 1954, indicates that wrong responses are not all given in accordance with the laws of chance. Some may be given because of misinformation. He al so j points out that the formulas used are based on ques tionable assumptions. It is assumed that all the wrong answers plus some of the right ones are chance an swers or guesses. It is further assumed that the re duced score obtained by use of the correction for mula, gives a more accurate picture of the students actual ability. This reduced score is supposed to indicate the number of answers really known by the person taking the test. Doppelt further states that corrected scores are an attempt to rule out the ef fects of differential boldness in taking a test, rather than a method for getting a true picture of the exam inees ' knowledge. Some student are bold, and answer questions when they are not sure of the answers, while their more timid colleagues would rather omit those questions. Cronbach and others have found some evidence of acquiescence operating in taking tests, since do not-guess instructions placed ascendant students at an advantage over submissive students. It is argued that such instructions reduce the validity of achieve ment, since they become in some degree measures of personality traits. Some educators maintain that a student taking a true-false examination has a fifty percent chance of guessing the correct answer if he knows nothing a bout the question. However, L. L. Thurstone in his book: The Reliability and Validity of Tests indicates that the weighting which is given to errors by the ' Chance formulas is based on the theory of probability and for a given examination may be considerably in error.