Goldberg: Observations

Subject 2 liked the Music Processor interface.

• Subject 2 seemed to like working with the Rating and Choosing interfaces. She preferred having the names of variations available. She said that the visual representations in the Choosing interface were not useful to her, but standard music notation would have been.
• Once she started using the Music Processor, however, it was evident that it offered the interface she would really use.
• Subject 2 seemed confused by the pull-down menus in the Music Processor that offered access to Goldberg's operations. Instead, she preferred looking through the snippets of music notation we had prepared for each variation. When asked, she said she could read the music and get a very good idea of what each variation would sound like. She could also do this very quickly, sorting through the pile of variations rapidly.
• A "smart microphone" input device, which the user could sing into and get the resulting musical notation on-screen, would have been useful for this subject. At one point, she wanted to enter a new variation, but she couldn't play it at the keyboard. However, she said she could hear it in her head and thought she could sing it.
• We were doing lots of interpretation of Subject 2's requests when testing the Music Processor interface. Actually implementing a system that could do everything we were simulating would be extremely difficult.

Conclusion: For a trained musician and traditional composer like Subject 2, the Music Processor interface seems to offer the most intuitive and appealing metaphor. Although she did not dislike the other interfaces, she was most excited about the Music Processor.