Economists since Demsetz have viewed property rights as a way to internalize the external costs and benefits one party's action confers on another. They have thought this internalization desirable, reasoning that if a party didn't capture the full social value of her actions she wouldn't have optimal incentives to engage in those actions. Measured by this standard, IP rights are inefficiently weak. There is abundant evidence that the social value of innovations far exceeds the private value. But there is also good evidence that, contrary to what economists might assume, these spillovers actually encourage greater innovation. The result is a puzzle for Demsetzians. In this article, we offer three insights that help to explain the positive role of innovation spillovers. First, we note that in IP, unlike real property, a wide range of externalities matter, because IP rights are much less certain than property rights, and because the decision to create a legal entitlement will determine whether or not a transaction must occur. Second, we make the point that while society needs some ex ante incentive to innovate, it doesn't need (and doesn't particularly want) full internalization of the benefits of an invention. Third, we observe that even where internalizing externalities is desirable, property rights do not in fact do so perfectly, and they create problematic distortions in circumstances in which the buyer in a transaction makes productive reuse of the work. The result of combining these insights is that at least where innovation is concerned, we cannot rely on the easy equation of property rights with efficient internalization of externalities.