Both theory and data suggest that that richer, more informal, and more interactive media should be better suited for handling the more complex, equivocal, and emotional aspects of collaborative tasks. To test this hypothesis, we constructed an experiment in which participants were required to make either written or spoken annotations to a document to help a fictional co-author revise it. We seeded relatively error-free texts with errors of different scope . The results provide strong evidence that a richer - in the sense of a more expressive --medium is especially valuable for the more complex, controversial, and social aspects of a collaborative task. Subjects stated that they preferred to use voice to comment on higher-level issues in a document and to use text to deal with lower-level problems of spelling and grammar. When subjects' annotation modalities were restricted, using written annotations led them to comment on more local problems in the text, while using speech led them to comment on higher level concerns. When they did use written annotations to comment on global problems, they were less successful than when they used spoken annotations. Finally, when they offered spoken annotations, they were more likely to add features, such as personal pronouns and explanation, that made their comments more equivocal and socially communicative. These results indicate the uses to which systems that provide voice annotation are likely to be put.