When students were informed of the number of errors they had made on a 30-point true-false test and given their answer sheets on which they crossed out and changed the answers they thought might be the ones in error, they could not significantly raise their scores (corrected for chance) when the number of true and false statements were approximately equal. If the number of true statements were greatly in excess of the false statements, say two true to one false, they lowered the score by their changes. On the other hand, if the false statements greatly outnumbered the true ones, then they raised their scores by their changes, no doubt due to the fact that they tended to equalize the number of true and false answers.
Proceedings of the Iowa Academy of Science
©1936 Iowa Academy of Science, Inc.
Finkenbinder, E. O.
"A Measure of the Amount of Chance in True-False Tests,"
Proceedings of the Iowa Academy of Science, 43(1), 306-307.
Available at: https://scholarworks.uni.edu/pias/vol43/iss1/106