It is evident that speech is the most distracting sound in offices. Room acoustical design should aim to the reduction of speech noise. There is a great need for an empirical prediction model that would explain the attenuation of speech noise both at short and long distances from the speaker, the "noise source". The model should be usable also for, e.g. architectural or interior designers. Therefore, the use of complex room acoustical models, e.g. ray-tracing, is not possible. The aim of this study was to evaluate the spatial decay of STI and speech SPL in a group of 20 open-plan offices. Large variation of main room parameters was desired including e.g. room height and width, ceiling, wall and floor absorption, screen height and material. In each office, the STI, reverberation time, speech SPL and background noise level were measured at every workstation along a straight line starting from the source, omnidirectional loudspeaker. The differences between open-plan offices were unexpectedly large. It seems that control of horizontal reverberation, along with strong ceiling absorption, is necessary, if good attenuation of speech is desired. The benefit of masking sound was evident especially at short distances from the speaker. The new way of characterizing open-plan offices facilitates the acoustical design of open-plan offices. It is easy to demonstrate the effects of different room acoustical remedies on STI at short and large distances from the speaker. It is also possible to compare open-plan offices and room offices.