A mingogram reading experiment was carried out in which an expert in Swedish prosody was presented with computer simulated mingograms of unknown Swedish sentences and asked to identify the following categories: stressed and unstressed syllables, grave and acute word accents, focal accent, and terminal juncture. Out of a total of 178 occurrences of the different categories, 151 were correctly identified (85%). The categories were identified by using the F0 contour, energy envelope and a duplex oscillogram. On the basis of this experiment, a set of preliminary, hierarchically ordered automatic analysis rules have been formulated using F0 movement patterns synchronized with energy envelope peaks to define the prosodic categories. These rules have been tested by using two non-expert mingogram readers and are being implemented on an automatic prosodic analysis system.