First, a prediction of the sound absorption of seated audiences in four concert halls was made by three methods. Employed were: reverberation chamber data obtained at the laboratory by the Kath and Kuhl method [Acustica 15, 127 (1965)] and the ISO-354 method on similar seating, and average absorption coefficients derived from RT data in 15 halls with three levels of upholstered seating [Beranek, Concert and Opera Halls, Acoustical Society of America, Woodbury, NY, 1996]. For each of the four halls, the residual absorption coefficients were determined from actual measurements (``residual`` meaning that due to all surfaces other than the seating). When the coefficients from the three methods and the measured residual coefficients were substituted in the Sabine equation, the K&K method gave the best prediction above 500 Hz and the Beranek method below 1000 Hz. In order to explain the large differences between K&K/ISO and actual hall measurements at the lower frequencies and to construct a method for predicting RTs in halls from reverberation chamber data, the difference in diffusivity in large concert halls and a reverberation chamber was examined. The incident angle distribution of sound rays onto a seated audience was determined by computer simulation and the oblique incident absorption coefficients of seated audiences were measured in an anechoic room. These findings were then employed to explain the large discrepancy.