Split-beam positional estimates of fish detected in a river at close range often do not correspond to the actual position of the target. These inaccuracies create problems in determining whether a fish is moving upstream or downstream. We hypothesize that these positional estimates are degraded by two factors: size of target relative to beam diameter, and the complex scattering of the fish. These parameters create a near-field effect, within which the phase measurements of the returning echoes are corrupted. Examples of fish tracks from near and far range fish detected by a split-beam echo sounder are provided to illustrate these inaccuracies. Experimental data from tethered spheres and complex targets show increasing distortion with target complexity and proximity to the transducer.