Gambler's fallacy in the classroom?

Does students' hand tremble after marking three consecutive identical answers in a multiple choice test? We design an experiment to study if the likelihood to change incorrectly to a different answer than the last one depends on the number of identical previous answers. We do not find a clear treatment effect, but observe that indeed the likelihood to change to an incorrect answer increases in the number of identical previous answers given by the student, even after controlling for how prepared (s)he was overall and how certain (s)he was that the answer to a given multiple choice question is correct. We claim that this behavior possibly is a reasonable reaction to previous exam experience.