Brent,
Thanks for your response. It is a DAQCard, not a DAQPad. I was originally using a DAQPad but the first byte latency was killing me (plus 9 minutes to self-calibrate!!!) but that's another story.
I think your ground loop idea is probably right, I just thought the battery 0V would float to the AIGND level, but clearly not. I have broken the ground loop now, which was caused semi-accidentally via a link between the case of the linear regulator (at the battery 0V potential) and some custom signal conditioning electronics whose enclosure was at AIGND, but I would really like to know why the 0V of the battery would not just float to the AIGND level (I'm probably being dumb).
Cheers
Jamie Fraser