I hope Owain Glyndwr can cut in and correct where i´m wrong or to add where necessary.
Franzl,
There is not a lot I can add to your remarks, but I can't resist an invitation
The BEA wording is rather loose, but I think the context indicates that they are concerned with static "speed stability" rather than classical short period mode.
It is fundamental that you can change/control either speed or flight path by the use of elevator alone, but not both at the same time.
So JT's description of "pull to reduce speed, push to increase it" is fine for classical airplanes or the A330 in direct law, but it rather loses meaning when stick movement commands a flight path change. Even more so when the implied pilot's command on releasing the stick is "Hold this flight path". With the flight path being held by the FCS any pitching moments are taken care of by the automatics so the CG to aerodynamic centre moment arm isn't really a great concern for speed control. For what it's worth, the aircraft would have to be statically stable in pitch to get certification. Moreover (and I am not sure of my ground here) I think they might have had to show that it was flyable for extended periods in failed states to get EROPS clearance.
Speed stability on the A330 in normal or alternate laws is therefore a rather different animal to what we were all brought up to understand.
As I said previously, it looks as if the aircraft under normal or alternate laws would be more or less neutrally speed stable over a fair range of airspeed - say from 200 kts to 260 kts - and increasingly speed stable above that (Mach 'tuck' effects excepted - but they seem to be modest and it doesn't affect AF447 anyway). Below about 200 kts the aircraft would be increasingly speed unstable because it would be flying up the back side of the drag curve.
The problem, as I see it, is that if the FCS is trying to maintain 1g flight when the airspeed is falling it will apply NU pitch. But even though the aircraft is descending rapidly the measured 'g' (without pilot input) may not be all that far away from 1.0 so that NU command may not be all that big (I'm not specifically relating to AF447 here because there was considerable pilot input on those traces). In such a case the attitude might not change much even though the airplane was decelerating and the AoA increasing rapidly due to the increasingly negative flight path angle. I think this is what the BEA are getting at.