Ground penetrating radar (GPR) surveys of unit and compound braid bars in the sandy South Saskatchewan River, Canada, are used to test the influential facies model for sandy braided alluvium presented by Cant & Walker (1978). Four main radar facies are identified: (1) high-angle (up to angle-of-repose) inclined reflections, interpreted as having formed at the margins of migrating bars; (2) discontinuous undular and/or trough-shaped reflections, interpreted as cross-strata associated with the migration of sinuous-crested dunes; (3) low-angle (< 6°) reflections, interpreted as formed by low-amplitude dunes or unit bars as they migrate onto bar surfaces; and (4) reflections of variable dip bounded by a concave reflection, interpreted as being formed by the filling of channel scours, cross-bar channels or depressions on the bar surface. The predominant vertical arrangement of facies is discontinuous trough-shaped reflections at the channel base overlain by discontinuous undular reflections, overlain by low-angle reflections that dominate the deposits near the bar surface. High-angle inclined reflections are only found near the surface of unit bars, and are of relatively small-scale (< 0·5 m), but can be found at a greater range of depths within compound bars. The GPR data show that a high spatial variability exists in the distribution of facies between different compound bars, with facies variability within a single bar being as pronounced as that between bars. Compound bars evolve as an amalgamation of unit bars and other compound bars, and comprise a facies distribution that is representative of the main bar types in the South Saskatchewan River. The GPR data are compared with the original model of Cant & Walker (1978) and reveal a much greater variability in the scale, proportion and distribution of facies than that presented by Cant & Walker (1978). Most notably, high-angle inclined strata are over-represented in the model of Cant and Walker, with many bars being dominated by the deposits of low- and high-amplitude dunes. It is suggested that further GPR studies from a range of braided river types are required to properly quantify the full range of deposits. Only by moving away from traditional, highly generalized facies models can a greater understanding of braided river deposits and their controls be established.