The aim of this thesis is to address a gap within many recent accounts of causation that take the truth-values of causal statements to be sensitive to normative considerations such as whether or not an event is normal. These accounts are intended to provide a solution to the problem of the apparent arbitrariness of our selecting some events as causes while others are relegated to being mere background conditions. This problem arises when we consider the intuitions we have about certain puzzle cases, which seem like they can be explained if we adopt a normative account of causation by holding that the truth-values of our causal statements are sensitive to what is and is not normal (Hitchcock and Knobe 2009) (Halpern and Hitchcock 2015) (Menzies 2007) (McGrath 2005). There are other ways of explaining these intuitions that do not commit us to the view that the truth-values of our causal statements are sensitive to facts about normality, however those alternative explanations generally do not succeed at providing a better explanation than the one offered by the normative account. We therefore have good reason to prefer a normative account of causation. Unfortunately, although these normative accounts of causation do provide some explanation for our intuitions about the puzzle cases, they also suffer from a number of problematic unclarities that result from those accounts not containing any kind of conceptual analysis of normality. I will address this issue by developing my own original analysis of normality that I then go on to apply to causation. The end product of this will be a new normative constraint on accounts of causation.