The multiaccess interference and the array gain have been modeled to evaluate the average probability of error in direct-sequence code-division multiple-access receivers with adaptive antenna arrays. The angular gain of the spatial filter (array beampattern) is modeled by a piece-line function that approximate the passband (or in-beam) with a fixed gain and the stopband (or out-beam) with an equivalent attenuation. The users are partitioned into in-beam/out-beam interferers and are counted differently for the evaluation of the probability of error. This simple way to account for the multiaccess interference can be exploited to evaluate the average probability of error when the users are randomly distributed within an angular sector. The analytical model can be employed to evaluate the average performance for two-dimensional RAKE and linear minimum mean square error multiuser detection receivers over frequency selective Rayleigh fading channels.