Laboratory column experiments for colloidal transport and retention are often carried out with flow direction oriented against gravity (up-flow) to minimize retention of trapped air. However, the models that underlie colloidal filtration theory (e.g., unit cell models such as the Happel sphere-in-cell and hemispheres-in-cell) typically set flow in the same direction as gravity (down-flow). We performed unit model simulations and experimental observations of retention of colloids with different size and density in porous media in the absence of energy barriers under both up-flow and down-flow conditions. Unit cell models predicted very different deposition (e.g., for large or dense colloids with gravity number N(G) > 0.01 at pore water velocity of 4 m/day) under down-flow versus up-flow conditions, which reflect underlying influences of gravity and flow on simulated colloid trajectories that resulted in very different distributions of attached colloids over the model surfaces. The Happel sphere-in-cell model showed greater sensitivity to flow orientation relative to gravity than the hemispheres-in-cell model. In contrast, experimental results were relatively insensitive to orientation of flow with respect to gravity, as a result of the variety of orientations of flow relative to gravity and to the porous media surface that exist in actual porous media. Notably, the down-flow simulations corresponded most closely to the experimental results (for near neutrally buoyant colloids); which justifies the common practice of comparing up-flow experiments to theoretical predictions developed for down-flow conditions.