The variability of compound stress in English: structural, semantic, and analogical factors 1
It is generally assumed that noun–noun (NN) compounds in English are stressed on the left-hand member (e.g. cóurtroom, wátchmaker). However, there is a considerable amount of variation in stress assignment (e.g. silk tíe, Madison Ávenue, singer-sóngwriter), whose significance and sources are largely unaccounted for in the literature. This article presents an experimental study in which three competing hypotheses concerning NN stress assignment are tested. The stress patterns of novel and existing compounds, as obtained in a reading experiment with native speakers of American English, were acoustically measured and analyzed. The results show that there is indeed a considerable amount of variation in stress assignment, and that all three hypothesized factors, i.e. structure, semantics, and analogy, are relevant, though to different degrees. On a theoretical level, the findings strongly suggest that a categorical approach cannot be upheld and that probability and analogy need to be incorporated into an adequate account of stress assignment in noun–noun constructions. The article also makes a methodological contribution to the debate in showing that experimental studies using pitch measurements can shed new light on the issue of variable compound stress.(Received May 4 2005)
(Revised October 4 2005)
1 This article is dedicated to Günter Rohdenburg on the occasion of his 65th birthday. I would like to thank the two ELL reviewers for their remarks on the first version of this article. I am also grateful to the audiences at the DUtKöMarSie-Workshop 2005, the Sprachwissenschaftliches Kolloquium at Universität Siegen 2005, and the Jahrestagung der Deutschen Gesellschaft für Sprachwissenschaft 2005 in Cologne for providing comments and suggestions. I am also very grateful to Heinz Giegerich, Laurie Bauer, Maria Braun, Miriam Ernestus, Sabine Lappe, Hiromi Noda, Gero Kunter, and Mareile Schramm for commenting on earlier versions. Thanks are also due to Gero Kunter for his help with some of the acoustic analyses and Holger Mitterer for his help with the Praat scripts. A very special thanks goes to Harald Baayen for his critical support, detailed suggestions, and for sharing his expertise in statistics with me. I also thank my student Julia Albrecht for starting all this. Needless to say, the usual disclaimers apply. This work was supported by a research grant from the Deutsche Forschungsgemeinschaft.