Abstract: This paper examines patterns of knowledge diffusion from the U.S. and Japan to Korea and Taiwan using patent citations as an indicator of knowledge flow. We estimate a knowledge diffusion model using a data set of all patents granted in the U.S. to inventors residing in these four countries. Explicitly modeling the roles of technology proximity and knowledge decay and diffusion over time, we find that it is much more likely for Korean patents to cite Japanese patents than U.S. patents, whereas Taiwanese inventors tend to learn evenly from both.  We also find that both Korea and Taiwan are surprisingly reliant on relatively recent technology.