While folklore in finance holds that split valuation effects are due to dividend increases associated with splits, little is known about magnitudes of dividend and nondividend components of split announcement effects. We find that splits and dividends are indeed informational substitutes, a notion we characterize more precisely, but a significant portion of split valuation effects, 46% according to our estimates, cannot be attributed to dividend information in splits. Our techniques extend the literature on conditional event-study methods and we illustrate their practical value in testing hypotheses and analyzing data not amenable to analysis by standard procedures.