The O'Connell effect --the presence of unequal maxima in eclipsing binaries-- remains an unsolved riddle in the study of close binary systems. The Kepler space telescope produced high-precision photometry of nearly 3000 eclipsing binary systems, providing a unique opportunity to study the O'Connell effect in a large sample and in greater detail than in previous studies. We have characterized the observational properties-including temperature, luminosity, and eclipse depth-of a set of 212 systems (7.3% of Kepler eclipsing binaries) that display a maxima flux difference of at least 1%, representing the largest sample of O'Connell effect systems yet studied. We explored how these characteristics correlate with each other to help understand the O'Connell effect's underlying causes. We also describe some system classes with peculiar light-curve features aside from the O'Connell effect (~24% of our sample), including temporal variation and asymmetric minima. We found that the O'Connell effect size's correlations with period and temperature are inconsistent with Kouzuma's starspot study. Up to 20% of systems display the parabolic eclipse timing variation signal expected for binaries undergoing mass transfer. Most systems displaying the O'Connell effect have the brighter maximum following the primary eclipse, suggesting a fundamental link between which maximum is brighter and the O'Connell effect's physical causes. Most importantly, we find that the O'Connell effect occurs exclusively in systems where the components are close enough to significantly affect each other, suggesting that the interaction between the components is ultimately responsible for causing the O'Connell effect.