We have modeled the adsorption of ammonia on the Au(111) surface at coverages of 1/4 and 1/9 of a monolayer using density-functional theory employing the pseudopotential method, periodic imaging, a plane-wave basis set, and the PW91 density functional. The geometries of the adsorbate and the surface are fully optimized. The adsorption is found to be highly favored on top of a surface atom. Adsorption energies of 26 and 32 kJ mol(-1) are obtained for the 1/4 and 1/9 of a monolayer coverage, respectively, extrapolating to 34 kJ mol(-1) at zero coverage; the experimental estimate is 32-42 kJ mol(-1). Small changes in the work function are predicted and interpreted as arising from a surface layer whose effective dipole moment is 2.15 D, 0.77 D larger than the calculated value of isolated ammonia. Examination of the calculated charge density and the local electric field strengths indicate that the change in dipole moment is due to polarization effects and that ammonia to gold charge transfer is minimal, at most 0.01 e in magnitude. Qualitatively, the local densities of states and the charge distribution provide little indication of covalent bonding between the gold and ammonia, and quantitatively the adsorption is interpreted as arising from dispersive interactions with some contribution from polarization. This picture is in contrast with common notions of gold to ammonia binding which depict weak chemisorption rather than physisorption, but the usefulness of PW91 in distinguishing between these processes is questioned through examination of the calculated potential energy surface of Ne-2. PW91 is shown only to mimic dispersive interactions using modified covalent terms.