Machine reading research has reached the point where given a text and a reading comprehension question regarding that text, a trained model can specify the starting and ending point in the text that contains the answer to the question. However, there exists the problem of 'unanswerable questions', or questions that are designed to look like they contain an answer in a given text, but in reality cannot be answered by the text. Current models lack in this ability to detect whether or not a given question is actually answerable; this problem forms the basis of my research.