The interaction between eccentricity and an external forcing fluctuation in gear rattle response is investigated experimentally. The experimental rig consists of a 1:1 ratio steel spur gear pair, the input gear being controlled in displacement and the output gear being under no load. Gear transmission errors recorded using high accuracy encoders are presented. Large variations in backlash oscillation amplitude are observed as the relative phase of the input forcing and the sinusoidal static transmission error due to eccentricity is varied. A simplified mathematical model incorporating eccentricity is developed. It is compared with experimental findings for three different gear eccentricity alignments by way of plots relating backlash oscillation amplitude to forcing amplitude and phase relative to eccentricity sinusoid. It is shown that eccentricity does not fully account for the experimentally observed large variations in amplitude. Through analysis of the experimental data, it is suggested that further tooth profiling errors may explain the discrepancies.