This paper adds to the literature on the Sarbanes-Oxley Act's net effects by looking at whether its passage was associated with a change in innovation and patenting. Its effects are separated into temporary uncertainty and changes in long term investment incentives in a dynamic programming problem faced by innovators who learn over time about SOX's effect. Innovation is found to fall under uncertainty for potential losses that are low relative to the potential profits. As companies learn, innovation rates readjust to SOX's long term persistent effect. We examine US patenting in stem cell technologies from 2001 to 2009 for SOX related changes. To reduce the dependence of our estimates on timing assumptions, we look for changes over the whole period. We firstly use a rolling break test with a single break point with Monte Carlo correction to p-values for search process endogeneity and MLE bias. Secondly, we run a hidden Markov model allowing for multiple states in the patent process and transitions between the states. We find a large and statistically significant change at a date consistent with a SOX effect under both testing methods. A three state hidden Markov model finds subsequent correction consistent with the theoretical model. Four competing explanations are found to account incompletely for the observed data.