An acoustic study of German focused on voice frequency at sentence-, turn-, and discourse-end in conversations. The data were drawn from short dialogues in which the same word occurs at the ends of utterances, in the middle of a turn, at the end of a turn, and at the end of a discourse. The dialogues were read 10 times by a male and a female native speaker of German. Peak frequency, endpoint frequency, and the degree of the fall in frequency were computed. Results indicate that peak frequencies tend to vary greatly, while endpoint frequencies do not. Discourse-final frequency is characterized by a lower peak and smaller degree of fall, but not significantly lower frequency. In addition, the data show that the male speaker uses more sharply falling contours at turn-ends than the female speaker. (Author/MSE)