Cross-linguistic evidence for probabliatic orthographic cues to lexical stress

Abstract

Reading a word requires converting symbols to a phonetic sequence but also determining the stress position of that word. In this chapter, we analysed corpora of English, Dutch, German, Italian, Spanish and Greek to determine whether sublexical probabilistic information from the very beginnings and endings of words provided sufficient information to determine stress patterns. We found that such information was sufficient to accurately determine stress positions. However, languages varied as to whether beginnings or endings were more informative. Furthermore, the extent to which stress patterns were regular within each language related to the reliability of the sublexical cues to stress position. The analyses show that stress does not have to be stored at the lexical level to support pronunciation.