Patterns of incorrect answering or misconception-like responses to scientific concept questions have been well documented. Here we investigate both response choices and response times to gain insight into the nature of misconception-like responses. In a series of experiments involving questions on graphs in which participants must compare the slopes of two points, we find that students answering with misconception-like responses, namely comparing heights rather than the slopes, do so consistently and more rapidly than those answering correctly. We also find in a speeded experiment, that all students are able to compare slopes and heights, but comparing heights requires less time than comparing slopes. Finally, by imposing a delay in responding that is long enough for the responder to process both slopes and heights, we find a reduction in misconception-like responses. Thus the misconception-like responses can be explained in terms of speed-accuracy trade-off models in which responders place high priority on answering quickly.