The frustrated pyrochlore antiferromagnet Gd$_{2}$Ti$_{2}$O$_{7}$ has an
unusual partially-ordered magnetic structure at the lowest measurable
temperatures. This structure is currently believed to involve four magnetic
propagation vectors $\mathbf{k}\in \langle \frac{1}{2} \frac{1}{2} \frac{1}{2}
\rangle^*$ in a cubic 4-$\mathbf{k}$ structure, based on analysis of magnetic
diffuse-scattering data [J. Phys.: Condens. Matter 16, L321 (2004)]. Here, we
present three pieces of evidence against the 4-$\mathbf{k}$ structure. First,
we report single-crystal neutron-diffraction measurements as a function of
applied magnetic field, which are consistent with the selective field-induced
population of non-cubic magnetic domains. Second, we present evidence from
high-resolution powder neutron-diffraction measurements that rhombohedral
strains exist within magnetic domains, which may be generated by
magneto-elastic coupling only for the alternative 1-$\mathbf{k}$ structure.
Finally, we show that the argument previously used to rule out the
1-$\mathbf{k}$ structure is flawed, and demonstrate that magnetic
diffuse-scattering data can actually be fitted quantitatively by a
1-$\mathbf{k}$ structure in which spin fluctuations on ordered and disordered
magnetic sites are strongly coupled. Our results provide an experimental
foundation on which to base theoretical descriptions of partially-ordered
states.