The eccentricity distribution of exoplanets is known from radial velocity surveys to be divergent from circular orbits beyond 0.1 au. This is particularly the case for large planets where the radial velocity technique is most sensitive. The eccentricity of planetary orbits can have a large effect on the transit probability and subsequently the planet yield of transit surveys. The Kepler mission is the first transit survey that probes deep enough into period space to allow this effect to be seen via the variation in transit durations. We use the Kepler planet candidates to show that the eccentricity distribution is consistent with that found from radial velocity surveys to a high degree of confidence. We further show that the mean eccentricity of the Kepler candidates decreases with decreasing planet size indicating that smaller planets are preferentially found in low-eccentricity orbits.