Abstract English, German, and Dutch show very similar word stress patterns, in that word stress is not fixed to a certain position within a word, but realized within the final three syllables. There is, however, no consensus on the actual stress-assigning algorithms and the role of quantity (e.g., Kiparsky 1982; Wiese 2000; Hayes 1995; Giegerich 1985, 1992; Trommelen and Zonneveld 1999a, b). Existing studies are methodologically problematic since they largely depend on convenience samples of existing words and do not test their claims with new words. Using mixed effects regression and classification trees as analytical tools, this paper presents the results of a production experiment with pseudowords and an analysis of large random samples as found in the CELEX lexical database. It is shown that stress assignment is sensitive to syllabic weight in all three languages, though in slightly different ways. The implications of these results for the metrical structure of the three languages are discussed.