Vision plays an important role, both in language and in the external world. Almost everything in our lives is vision-oriented and this in turn has led to the spread of vision-oriented words into other cognitive realms. For instance, the English verb 'see' is used in such tags as "I see," to denote that understanding has taken place (this understanding may or may not be based on visual evidence), and a saying like "seeing is believing," which underlines the importance of vision for absolute proof (see Matlock 1989 for a discussion of this metaphor). In certain languages, vision forms a part of the evidential system. In these languages, it has been found desirable to have grammatical morphemes that denote that the action described in the sentence has been obtained visually. Surprisingly enough, it turns out that visual evidentials typically do not derive from vision words, such as the verb "to see." Rather, they tend to develop from demonstratives or tense/aspect markers. This paper gives a cognitive account for the deictic origin of visual evidentials. Some 160 languages in North and South America have been investigated for this study, as well as several languages from other parts of the world. The gauges from North and South America are data from a project which aims to map the spread of evidentiality in the Americas, while the data from other parts of the world come from the evidential literature. (Ferdinard de Hann)