An analytical approximation for the scatter to primary dose ratio (SPR) on the central axis was validated against Monte Carlo results and experimental measurements for homogeneous and inhomogeneous phantoms. The analytical approximation only included first-order Compton scatter. The contribution to the total SPR from first-order Compton scatter, multiply scattered photons and electron scatter was investigated using Monte Carlo simulation for homogeneous phantoms (up to 30 cm thick for 6 and 18 MV beams; source to detector distances from 150 to 230 cm) as well as for a neck, thorax and pelvis phantom. SPRs were measured on the central axis with an ionization chamber for water phantoms (up to 20 cm thick at 4 MV, 30 cm for 6 MV and 10 MV and 40 cm for 18 MV; source to detector distances of 185 and 200 cm) and for phantoms representing the neck, thorax and pelvis (for air gaps of 50 cm and larger). The mean difference between the experimental and analytical SPRs on the central axis for source to detector distances of 170 cm or greater was within: -0.003 (neck); -0.012 (thorax); -0.028 (pelvis, 10 MV) and 0.008 (pelvis, 18 MV) respectively.