Accurate air interface traffic forecasting and dimensioning is of importance in any cellular network for achieving cost and quality requirements. Previous studies of traffic modeling in cellular networks have tended to derive distributions to fit the measured data for the arrival rate and call holding processes or derive expressions for call blocking on the air interface for different handover and channel assignment procedures. In most cases it is assumed that the Erlang B model is not sufficiently accurate and some other call blocking model is required. However, there have not been a large number of studies published on how accurate or otherwise (and under what circumstances) the Erlang B model is in modeling air interface call blocking in practical cellular networks. In this paper call blocking measurements of the air interface of a “real” cellular network are presented. A statistical analysis is undertaken which shows that the measured data is correctly modeled by Erlang B at a level of significance of 0.05 when the number of channels are greater than 12 and the blocking experienced is greater than 1%. For available channels less than 12 and blocking less than 1% the Erlang B model overestimates the blocking.