In this paper we put forth some empirical evidence from the Brazilian Amazon that the relationship between roads and land clearing may be much more complex than the conventional wisdom implies. In particular we find that in areas that already at least partially cleared, improving the road network (i.e. decreasing transport costs) may actually decrease the rate of deforestation. We argue that our methodology of explicitly modeling the dynamics should be preferred to the more common static, contemporaneous analyses found in the literature. Furthermore, we endeavor to provide an encompassing explanation of our results. In other words, not only do we show that dynamic modeling yields different conclusions from the conventional wisdom, but using our dynamic approach we are able to explain why so many other studies came to (possibly erroneous) conclusions using more traditional methods.