Hmmm... this has changed from when I knew it in the past. Many gaming web sites in the past were really shook up on this issue and they concluded that we can only really see differences in fps up to 60fps (rounded of course). I know for a fact that NTSC standard for television is 32 FPS. That's the standard for your typical shadow mask rounded TV screen. Other things like HDTV I know nothing about.
I didn't know movies double frame, although it makes sense. I don't know if this 60fps thing is true, all I know is that I read about it on many respected online gaming sites and one gaming magazine as well. As for high framerates, I always assumed that people bragged about it because it determined the life expectancy of their system. Higher framerates, while not useful, were justified because it meant the system would most likely be able to run future games for the next 2 years fairly well. Another thing I think people are forgetting is individual perception. Our eyes are constantly changing, from when we're born to when we get old and die. When we are first born, babies can only see within about 30 degrees straight ahead, and not to the side. This allows them to focus on things quickly and easily and not be overwhelmed from stress (that could potentially damage the developing brain prematurely). Also, a baby's vision is extremely blurry. As time passes, things become clearer, but then the older we get, things start to fade again. Vision is best in males (for women it's different) in their mid/early twenties, like 24 for instance. That's the age right before it starts degrading.
The argument that higher framerates are necessary for games is a compelling one. TV can get away with 32 fps, because it displays real-life pictures that from a distance look closer to something we see in real life (of course it's never exactly as we see it but you get the point). Games, on the other hand, have cartoonish qualities to them. While they strive to be realistic and in rare opportunities seize this chance to look very close to real life, most of the time they fail to be this way. One more failing is that in games, objects move very fast, which means that a stuttering effect is a possible danger even with v-sync turned on, because we begin to see flashes of individual frames. That's why from my own personal experience running games at 32fps is less than ideal. I can notice frame shifts and stutters. I tend to find that games become smooth at around 37fps, although that's just a rough guess based on my playing half-life with the netgraph turned on. 24-28fps is just very stuttery and ruins my enjoyment of the game.
It definitely is also important to note not just the average framerate but also the minimum and maximum framerates. If the minimum dips to 8 fps then I don't think that would be very playable, even if the average was 35.
Another thing I also noted in another forum post is that anti-aliasing is much more important than running games in high resolutions, and if the Matrox Parhelia's 16xFAA turns out to be what it claims to be, then the performance will be better as well. Keep in mind that running 1024x768 with 4x FSAA is the same as running 2048x1536 as they are both the same number of pixels.
1024x768x4=3145728
2048x1536=3145728
And since anti-alising framerate performance will improve over time due to better math routines, while running at higher resolutions will not, it is makes more sense to measure our framerates at lower resolutions with some sort of anti-aliasing turned on rather than measuring our framerates at gigantic resolutions. Add to this also the fact that many monitors people use today are not very high end and do not have adequate horizontal frequency to display such high resolutions as well as the cost of the monitors that can. Some monitors also tend to develop moire at higher resolutions and some geometry problems as well, making it even more impractical. If we take into account the future advances in anti-aliasing technology and optimizing further the programming behind the GPU and memory chips on the graphics cards themselves, we will start to see that in benchmarks on all the web sites, no one will be testing 1600x1200 and other high resolutions anymore as we see them doing right now, simply because 1600x1200 will be slower than anti-aliasing and may even have less quality. At most, 1600x1200 might be tested by hardware sites as a "guide" but in general no one will ever even use this resolution.