This paper tests Tirkkonen-Condit's (2004) Unique Item (UI) Hypothesis, which claims that UI are under-represented in translated texts and, on the other hand, Baker's (1993) Simplification Hypothesis and Halverson's (2003) Gravitational Pull Hypothesis, which predict over-representation of UI in translated texts. These hypotheses are contrasted by comparing the presence of English self-directed motion in English texts translated from Spanish from the Translational English Corpus (TEC, Baker 2003) and texts spontaneously produced in English from the Corpus of Contemporary American English (COCA, Davies 2008). Self-directed motion expressions are employed because of their linguistic divergences in English and Spanish. Twenty-eight English manner-of-motion verbs and eight English path-denoting satellites were selected to compare the number of self-directed motion expressions in the TEC and the COCA. This study yielded a total of 41,852 tokens from both corpora, that is, 209.2 expressions per million words in the TEC and 395.5 expressions per million words in the COCA. An independent samples t-test revealed that the number of expressions is significantly higher in the COCA (M = 3.32) than in the TEC (M = 1.76). A two-way ANOVA revealed significant main effects for Corpus and Lexical Frequency, but no Corpus*Lexical Frequency interaction effect was found. These results support Tirkkonen-Condit's UI Hypothesis and confirm that non-translated English is significantly richer in self-directed motion expressions than translated (from Spanish) English.