The dataset consists mostly of speech recordings. At the center are 1766 elicited Dinka utterances (wav format), collected through controlled elicitation from 14 speakers of the Bor dialects of Dinka. They are accompanied by annotations and f0-related files. In addition, the dataset includes a perception study carried out with 6 Dinka people. The dataset was collected in order to investigate the phonological contrast between early- vs. late-aligned falling contours in Dinka. This motivation is laid out in detail in a published paper: Remijsen, Bert (2013). Tonal alignment is contrastive in falling contours in Dinka. Language 89(2), 297-327.