Abstract This article aims to disentangle three explanations that have been proposed for the increased explicitness of translated English, as reflected in the more frequent use of the complementiser that in translated English texts compared to non-translated English texts. These three explanations are designated as the cognitive complexity (or processing strain) hypothesis, the pragmatic risk-aversion hypothesis and the source-language transfer hypothesis. Four comparable register-controlled corpora are used for the analysis: a corpus of English translated from Afrikaans, a corpus of written Afrikaans, and corpora of written British and native South African English. A multivariate analysis of the factors conditioning complementiser omission across the four corpora is used to test the three hypotheses proposed. The transfer hypothesis is tested by investigating whether the translation corpus demonstrates overall omission preferences that are more similar to the omission preferences of Afrikaans than of English. The cognitive complexity hypothesis is tested by investigating whether translated English is more sensitive to the complexity-related factors that are known to condition omission than non-translated English. The risk-aversion hypothesis is tested by investigating whether translations opt for the communicatively and normatively “safer” choice of including the complementiser in contexts where non-translated writing would typically omit it, and therefore demonstrate less sensitivity to register and frequency effects than non-translated English. The findings of the study provide strong evidence against the transfer hypothesis and find stronger support for the pragmatic risk-aversion hypothesis although the cognitive complexity hypothesis cannot be ruled out.