Changing the camera mode to "frame" instead of "field", the camera combines two interlaced fields together. You can see the results of James waving his arms quickly, but you can also see the result of me moving the camera while taking the picture. This seems to suggest that, when hand-holding the camera, you should only use single-field mode.

We turned compression from "normal" to "high" for this image, too. However, the artifacts still look pretty bad. I'm assuming that most of the bandwidth of JPEG coefficients went into coding the differences between even and odd scanlines...