The formation of dense water over the continental shelf and its descent along the continental slope have been investigated both theoretically and experimentally. Models have been developed for slope fronts and dense filaments, with emphasis on the role of the bottom boundary layer. An analytical, two-layer, two-dimensional model is first presented for the development of dense slope fronts near the shelf-break. The effects of vertical viscosity are explored and two behavioral regimes identified. The most relevant regime is determined by the parameter F = (UQ/gεs)(f3/v)1/2 where UQ is the flux of newly created dense water per unit length of coastline, g is the gravitational acceleration, ε is the density anomaly, s is the bottom slope, f is the Coriolis parameter and v is the vertical viscosity. In both cases, the alongslope velocity in the lower layer increases away from the coast during geostrophic adjustment, with an accompanying growth in the downslop Ekman flux. When F is small, dense water production near the coast can be balanced by transport within the boundary layer, which extends down the slope as a shallow intrusion with an alongslope speed of gεs/f. However, when F is large this type of flow cannot provide sufficient downslope transport. Dense water then accumulates, causing the front to steepen while diminishing the influence of the bottom slope. There is a corresponding increase in alongslope speed, which eventually plateaus at (2f/v)1/2/UQ where the Ekman flux balances the production of new dense water. These behaviors are strongly supported by results from laboratory experiments and are consistent with the limited available observations of the Antarctic Slope Front. After moving off the shelf, the dense water mass may continue to move down the slope within the bottom boundary layer, or alternatively, form an isolated filament with a front on both sides. Theoretical solutions are developed for dense filaments both with and without an active upper layer. In the latter case, the influence of dissipation is investigated beginning with a simple bulk parameterization. This produces a filament which broadens as it moves down the slope, while its mean alongslope velocity increases with bottom slope and its horizontal shear decreases. More realistic boundary layer dynamics have also been incorporated using a similar approach to that described for slope fronts. The solutions compare well with results from laboratory experiments on relatively stable filaments. Implications of the study for deep water formation around Antarctica are discussed briefly.