The perception of lexical stress in German: effects of segmental duration and vowel quality in different prosodic patterns.

Several decades of research, focusing on English, Dutch and German, have set up a hierarchy of acoustic properties for cueing lexical stress. It attributes the strongest cue to criterial-level f0 change, followed by duration, but low weight to energy and to stressed-vowel spectra. This paper re-examines the established view with new data from German. In the natural productions of the German word pair Kaffee 'coffee' - Café 'locality' (with initial vs. final stress in a North German pronunciation), vowel duration was manipulated in a complementary fashion across the two syllables in five steps, spanning the continuum from initial to final stress on each word. The two base words provided different vowel qualities as the second variable, the intervocalic fricative was varied in two values, long and short, taken from Café and Kaffee, and the generated test words were inserted in a low f0 tail and in a high f0 hat-pattern plateau, which both eliminated f0 change as a cue to lexical stress. The sentence stimuli were judged in two listening experiments by 16 listeners in each as to whether the first or the second syllable of the test word was stressed. The results show highly significant effects of vowel duration, vowel quality and fricative duration. The combined vowel-quality and fricative variable can outweigh vowel duration as a cue to lexical stress. The effect of the prosodic frame is only marginal, especially related to a rhythmic factor. The paper concludes that there is no general hierarchy with a fixed ranking of the variables traditionally adduced to signal lexical stress. Every prosodic embedding of segmental sequences defines the hierarchy afresh.