A couple of points:
1. NZ and the USA both have a requirement for ground visual during NVFR, and that is why they dont have NVFR LSALTs and those sorts of restrictions. If you can see the ground, you shouldn't hit it. Oz, as we know, has no such requirement. Nor, contrary to some of the posts here, is there a requirement for a visible horizon. Thus we have LSALT, ratings, and instrument proficiency requirements during training and ratings. As AOTW implies, this situation could certainly be revised.
2. I too disagree with McGowan's call on the instruments, though his other comments are spot on. NVFR is ALL about instruments, bith helo and fixed wing. Ideally, all changes should be made using the instruments, backed up by looking outside. The blacker it is, the more time you spend on instruments. Sorry Terry.
3. Contrary to imabell, some of the NVFR helo accidents have involved pilots with significant IFR experience, and have been unrelated to the weather.
4. Somatographic illusions are far less relevant in the helo, though far from irrelevant. Helicopters accelerate only to about 60 to 80 kias on climb out, and they rarely accelerate level to a positive rotate. Generally they increase speed and altitude simultaneuosly.
5. Centaraus: I think UA training in the actual aircraft must be part of the training to ensure that vestibular inputs are experienced: critical in my view.
6. Centaraus: you alledge that CASA claim the Helicopter Association of Australasia said a standby AI was not required. Where do I substantiate that claim? The HAA drafted a response to the intial recommendations of the Mackay accident last year, which has yet to be answered by the ATSB, and I can post that if required. To my knowledge, the HAA has made no formal position known to CASA (or the ATSB for that matter) on standby AIs.
7. The NVFR rating is currently the only rating I can think of that is perpetual. Renewals could be created, and a defined requirement for instrument profficiency introduced. Personnly, I think that until we change NVFR to require a visible horizon and or ground visual, we must insist on instrument profficiency.
Why dont we look at those issues too?