We show that scientific production can be described by two variables: rate of production (rateof publications) and career duration. For 19th century physicists, we show that the time pattern ofproduction is random and Poisson distributed, contrary to the theory of cumulative advantage. Weshow that the exponential distribution provides excellent goodness-of-fit to rate of production andcareer duration. The good fits to these distributions can be explained naturally from the statisticsof exceedances. Thus, more powerful statistical tests and a better theoretical foundation isobtained for rate of production and career duration than has been the case for Lotka's Law.