This work is part of a project to develop numerical simulation methods for predicting the e!ect of animals and pen partitions on the air#ow in the animal-occupied zone. As a reference case for this, it was desired to create a steady two-dimensional #ow near the #oor in an empty test room (8)5 m long, 3 m high and 5 m wide). A plane wall jet is usually expected to generate a two-dimensional air#ow and therefore the test room was equipped with a slot inlet beneath the ceiling on one entire wall of the room. A slot outlet was placed in the #oor near the inlet wall. Numerical simulation and measurements of air velocity and air#ow direction revealed that the #ow was highly three-dimensional. The three-dimensional e!ects were analysed in a number of additional numerical simulations and it was found that the relation between room width and room height had a crucial in#uence on the development of the three-dimensional e!ects of the air#ow. The three-dimensional e!ects were small unless the room width to height ratio was larger than one, and in rooms with width to height ratios larger than about three, simulations showed a number of similar air#ow patterns repeated transversely to the main #ow direction. 1999 Silsoe Research Institute Notation H height of room, m h inlet slot height, m k turbulent kinetic energy, m/s ยธlength of room, m R C Reynolds number ; longitudinal component of velocity, m/s ; longitudinal component of inlet velocity, m/s < inlet air speed, m/s = width of room, m x, y, z coordinates dissipation of turbulent kinetic energy laminar viscosity, N s/m density, kg/m