This paper examines the well-known literal translation hypothesis and discusses its significance for translation theory. The hypothesis claims that as translators process a given text chunk, they tend to start from a literal version of the target text, and then work towards a freer version. The idea has been implied or explicitly studied by many scholars, and does not seem to have a single source. After some preliminary conceptual analysis an optimal formulation of the hypothesis is proposed. The paper then assesses the hypothesis in terms of the kinds of wider significance any hypothesis can have. The criteria discussed are testability, relations with other hypotheses, applicability, surprise value and explanatory power. Some of Englund Dimitrova’s research (2005) on the hypothesis is discussed. A rather different study by Nordman (2009) is argued to have implications for the broader contextualization of the hypothesis.