The inverse method of Wunsch (1978) for determining steady state ocean circulation patterns is examined using a numerical model as the source of both data and observations. A simple, but commonly used, formulation of the technique is tested. The studies carried out have relevance to other descriptions, as they focus on features common to all formulations. The theme of the paper is a study of present methods rather than a comprehensive survey of parameter space. It is found that an optimizing principle of least perturbation from an initial velocity field can produce flows that are highly correlated with the observed fields. This is when station spacing is of the order of 1¿--2¿ and the number of layers is not greater than the number of vertical observations. Allowing flow through all faces of the boxes gives significantly more realistic velocities than merely having horizontal flow. However, three major problems arise with the box method. These are the inability to handle significant shears with moderate sampling, the lack of a uniformly successful way to decide on an initial state, and the necessity of using boxes of limited size. In particular, when the method uses boxes straddling an entire basin, the results are poor.