Constituent family size and compound stress assignment in English

There have been claims in the literature (e.g. Marchand 1969, Bell 2008) that variable stress assignment to English compounds is influenced by the size of the constituent families, i.e. the number of compounds that share the same left or the same right constituent. This paper tests this claim empirically on the basis of a large amount of data from three different corpora. The expected effects can be found in some form (and to varying degrees) in all three sources, alongside of other effects that have been held to be responsible for compound stress assignment. The results can be interpreted as evidence against deterministic rule-based approaches to compound stress and lend independent evidence to a model in which compound stress assignment emerges from the lexicon.1