Three views of demonstrative reference are examined: ‘contextual’, ‘intentional’, and ‘quasi-intentional’. According to the first, such reference is determined entirely by certain publicly accessible features of the context. According to the second, speaker intentions are ‘criterial’ in demonstrative reference. And according to the third, both contextual features and intentions come into play in the determination of demonstrative reference. The first two views (both of which enjoy current popularity) are rejected as implausible; the third (originally proposed by Kaplan in ‘Dthat’) is argued to be highly plausible.