The author sets out to investigate whether or not lexical normalization is a feature of translation in a specially compiled two-million-word German-English Parallel Corpus of Literary Texts (GEPCOLT). It represents an initial attempt to define a corpus-based methodology for the investigation of lexical normalization in translation. Frequency-ranked words form lists are used to identify potentially creative hapax legomena in source texts; and concordancing software is used to isolate creative collocations involving a selected source language node. The study finds that while around 44% of creative hapax legomena identified in the source texts in GEPCOLT are normalized in translation, the figure of creative collocations involving the German lemma AUGE is much lower at 16%.