The computational and algorithmic role of striatal dopamine signaling has been one of the most investigated topics in systems neuroscience. Though there have been many hypotheses of dopamine function over the years, the reward prediction error (RPE) hypothesis has remained the most popular for decades. Despite substantial evidence consistent with this hypothesis, considerable recent evidence is also inconsistent with it. Accordingly, multiple alternate hypotheses of dopamine function have been recently proposed. Here, I review one such alternative from my lab stating that dopamine is a teaching signal for causal learning. I show that there are fundamental similarities between this and other recent hypotheses, and that this hypothesis is consistent with the core results supporting RPE coding. Therefore, RPE should be considered a provisional hypothesis of dopamine function requiring further evaluation, with a critical need for openly examining a host of recent experimental results inconsistent with the hypothesis.