Learners are usually provided with support devices because they find it difficult to learn from multimedia presentations. A key question, with no clear answer so far, is how best to present these support devices. One possibility is to insert them into the multimedia presentation (canned support), while another is to have a human agent provide them (human tutoring). Human tutoring poses potential advantages: it uses spoken modality, displays non-verbal cues and implies social interaction. However, there is mixed evidence regarding these supposed advantages, and prior research comparing human and computer support presents problems. Our goal was to explore whether the advantages of human tutoring actually exist while avoiding the problems of prior research. In one experiment, participants learned Geology from a multimedia presentation including one of three forms of support: human tutoring, canned support or no support. After viewing the presentation, participants solved retention and transfer tests. Results revealed that participants in the human tutoring condition outperformed those in the other two conditions, who did not differ from each other. This means that human tutoring is advantageous, a fact that has implications in the design of support devices in multimedia learning. (Contains 2 tables, 2 figures and 2 notes.)