Neutron area survey instruments are designed to have an approximately isotropic response. In practice, the response cannot be perfectly isotropic for instruments that do not have spherical symmetry, and for all instruments it is modified by the inclusion of batteries, electronics, handles, etc. This affects the ability of the survey instrument to measure accurately an isotropic dose equivalent quantity. Measurements of the angle dependence of response for four of the most commonly used designs of survey instrument (Harwell 0949, Mark 7 NRM, NM2 and Studsvik 2202D) have been performed in a low-scatter room using radionuclide and monoenergetic neutron sources. The Monte Carlo code MCNP has been used to model the responses and to investigate their sensitivity to the polyethylene density, counting gas pressure and other manufacturing tolerances. Preliminary modelling results are presented here.