High-speed photography (2000–6000 pictures/sec) was used to register the fracture propagation process in gypsum specimens (diameter 40 mm, length 14 mm) loaded via 2 mm wide, strip-shaped paddings and under direct contact with plane press platens. With paddings in 3 out of 4 experiments the fracture, as it appeared on the first film frame, was asymmetrically located on the loaded diameter. In two direct-contact experiments the initial fracture had its widest part centrally and lateral fractures, of the»triple-cleft«type, appeared subsequent to the central fracture.Smaller specimens (diameter 7, length 14 mm), manufactured in an investment compound (Ceramigold®) were tested by the use of plane and vaulted platens, giving the same width of the loaded area. With vaulted platens the fracture first appeared in a peripherical zone in 3 out of 5 experiments and with plane platens in 6 out of 11 experiments. In two thirds of all cases, where the primary fracture sign appeared peripherically, it appeared in an area immediately centrally of the compression zones. These fractures were probably peripherically initiated and thus non-valid. Specimen appearance after testing gave, however, no clear indication of a non-valid fracture.