Many students have great difficulty taking multiple-choice tests. Often they attribute this difficulty to ambiguity in the questions, which makes more than one of the choices correct. Occasionally this belief is correct --the questions are sometimes (rarely) ambiguous-- but more typically a student makes the question ambiguous by "answering the answers" rather than answering the question.
Many students (unconsciously) treat multiple choice questions as true-false questions. They read the question briefly, then turn to the choices and sequentially evaluate the truth of each one. Unfortunately, a statement can easily be true and still not be the right answer to the question. Indeed, sometimes the correct answer is a false statement. For example, suppose I ask what theory A predicts in situation X, and theory A happens to make a prediction that is inconsistent with the data. The correct answer will be a statement that is false.
When students dither between choice "a" and choice "b", it's often the case that both are true statements, but only one of them is relevant to the question asked. Then, the student's task becomes trying to decide which choice statement is "more true", rather than trying to figure out the answer to the question. Thus, the student's understanding of the question is colored by his/her reading of the answer choices. Additionally, the student may become so rattled by what has become essentially an insoluble problem that his/her raised anxiety level interferes with the interpretation of subsequent questions as well.
I suggest you treat multiple choice tests as completion tests. Literally cover up the answer choices while reading the question, and try to write an answer on the exam. If you don't know the answer to a completion question, you don't have the option of quickly turning to the choices (opening you to the problems noted in the last paragraph). Instead, you read the question again, and think about the question harder. Once you generate an answer, treat the question as a matching question, checking to see if your completion answer matches any of the choices. This strategy helps you answer the questions, rather than "answering the answers". Although at first glance it might seem impossible to use this strategy for many questions (e.g., questions that start "Which of the following..."), with a little ingenuity, most multiple choice questions can be answered without reference to the choices.
Many students make the same mistakes on going over a test as they did the first time through. The first time through the exam, they mark the questions they had problems with, and put stars next to the two choices they think they need to decide between. Then, when going over the test the second time, they immediately proceed to dithering between the two choices, rather than rethinking the question. I suggest that you adopt strategies that make this problem less likely, e.g., force yourself to reread the question with the choices covered, don't mark the "likely" alternatives on the exam sheet, and so forth.
Consider the results of a little experiment I conducted with Psych 11 students many years ago. Students seated at computer terminals read a brief passage on international banking and currency arbitrage. Then they were given a set of multiple-choice questions on the screen, one at a time. They could only see the question or the answer choices on the screen, not both at once. They could toggle back and forth as much as they wanted. I recorded their answers and the time they spent viewing the questions and the time they spent viewing the choices. Students who did well on the test spent a greater proportion of their time viewing the questions than the people who did poorly. Also, upperclassmen spent proportionally more time viewing the questions than underclassmen, and also did better on the test.