The effects of using spatially selective and nonuniform radiofrequency pulses on the apparent T2 of a train of multiple echoes have been calculated by computer simulation. It is demonstrated that the use of selective pulses produces errors in estimates of T2 in imaging experiments because regions of the slice are effectively off resonance and therefore receive spin rotations less than in an ideal case. Components of the transverse magnetization that are incompletely rotated by 180 degrees refocusing pulses produce longitudinal components that are later returned to the transverse plane and add to subsequent echoes. These so-called ghost components influence the precise value of T2 estimated, but even when completely removed the resultant estimate is too short. The apparent T2 is influenced by the magnitude of the nutation angle inhomogeneity across the slice and paradoxically is not always more accurate with larger numbers of echoes. The reduction of apparent T2 caused by increasing the slice-selection gradient across a thin slab of material replicates the effects previously ascribed to diffusion and affords an alternative explanation for these without recourse to motion of the spins.