The air gap technique is an old method for scatter rejection. It is still used in lung examinations and may be reconsidered for use in digital radiography. Using magnification techniques, for example in mammography, the air gap thereby introduced simultaneously yields scatter rejection. A Monte Carlo collision density method is exploited to investigate the physical parameters relevant to this technique. Radiation quantities of scattered photons at points behind a water slab both on and laterally displaced from the central axis are calculated and their dependence on field area, slab thickness, air gap length and detector type is derived. Values of 'scatter-to-primary' ratios of the plane energy fluence (the energy imparted to a totally absorbing detector) are given for perpendicularly incident 30, 70 and 130 kV energy spectra, slab thicknesses of 0.05 and 0.2 m (30 kV: 0.05 m), air gaps of length 0.002-1.0 m and field areas from 8 x 10(-5) to 0.3 m2. Contrast degradation factors are derived for both totally absorbing and thin detectors. The influence on the scatter-to-primary ratios of using divergent instead of parallel beams and of neglecting molecular interference in coherent scattering is analysed.