Risk analysis can be applied to translation in several ways. One application concerns the specificity of translation, where risk is primarily the probability of the translator losing a translation-specific kind of credibility; it concerns relations between people and can be called 'credibility risk'. A second kind of risk ensues from the translator's uncertainty when making decisions about how to render an item; it involves cognitive processes and can be called 'uncertainty risk'. A third kind of risk then has to do with the way texts are interpreted and used in contexts, where some elements are high-risk because they are key to communicative success, while others are low-risk; this kind of risk applies to the different parts of texts and can be called 'communicative risk'. This third sense of risk then allows for a rationalist model of translators' decisions and effort distributions, positing that high effort should be invested in text items with high communicative risk. Although the differences between these levels of analysis can be confusing and require some careful definitions, the interactions between them offer a rich, non-essentialist view of translation as a social relation, as a product, and as a teachable mode of decision-making.