OK. so I was on the right track (so to speak) in thinking it was change of direction that would trigger a recording, as it seemed to me that recording changes would be the most efficient in terms of points. However, as the reporting has a 2D bearing which often appeared to be the same for successive (and often short) legs, I was thinking my theory didn't fit the data I was seeing. However, it may have been that there was a sufficient change altitude or speed that was causing the recording of the point. That makes sense.
It's perhaps worth explaining why I am so interested in how the tracking is being done. We went on a month's holiday recently driving from Brisbane to Perth and back (about 12000 km) and we were using the Garmin to record waypoints and tracks in the hope we could automatically (or at least semi-automatically) geo-locate our photos. Now in the 2 years we've owned the Garmin, we've never had a problem with recording track data. All the track data we've ever recorded was still fitting in the memory, but with the benefit of hindsight, it was a large number of fairly short trips. So we set off to Perth with the tracking on, assuming we'd have a month of track data on our return. Sadly we found we had about 4 days of track data because the Garmin didn't have sufficient memory and was overwriting. So, I'm now very motivated to try to work out the best settings for trackings to make the most efficient use of the memory to avoid this problem in future (I'd rather not have to cart a laptop around just to upload track info). But of course the solution might lie in getting a camera with a builtin GPS (still a bit of an expensive option but probably the preferred option in terms of what we are actually trying to achieve) or getting another GPS with more track memory.
Thanks for everyone's thoughts.