The advantages and disadvantages of using true/false tests in learning assessments, as well as the insertion of penalization (negative marking) of wrong answers as mechanism of adjustment on the guessing success, have already been presented and discussed by several authors in the context of learning assessment. However, few studies has been done in Brazil to understand the extent of the influence of using penalization of wrong answers in the measure of student proficiency in the evaluated domain. Hence, this article presents a study carried out at the University of Brasília, with the students of the basic disciplines of Physics-1 and Physics-2 coming from 20 different courses, in the unified Physics disciplines model. The students were alternately submitted to assessments with penalty and without penalty. The effect of reducing the reliability of the score due to the penalty is statistically compared with the effect of increasing the discriminatory power of the test due to the reduction of the guessworking. A new magnitude called “statistical yield” is proposed to interpret the results of these combined effects. To conclude, it is demonstrated that the increase in the discriminatory power of the test outweighs the reduction of the reliability of the score when applying penalization of wrong answers in true/false tests.
Keywords:
Mass disciplines; true/false tests; penalization of answers; reliability of the score; discriminatory power of the test
