Are there any studies on whether one will actually get more done during a days work if delays went down i. e. from 10 ms to 5 ms for each update. My point is that if we are already working at the speed of thought, the GUI is not the bottleneck.
While not talking about response times as low as the question, there are some very interesting results from tests carried out by Google and Bing here:
The bottom line is that users interact more with responsive web pages and the companies made more profit. I would think it's not too fanciful to extrapolate these and say that faster response times are nearly always desirable. Of course the gains will get increasingly small and you will run into the limitations of hardware.
It's probably the case as James Crook answered that fast UI responses act to increase comfort with the interface and so increase productivity rather than the time savings themselves allowing more to be done.
In the 10ms-5ms range you're running into the refresh rates and response times of pixels on your monitor. Many monitors are limited to 60Hz (17ms refresh). You're also getting close to limits of visual perception. We take around 100ms to direct our eyes to something new that has appeared on screen.
The difference between a compile taking 10 seconds and 0.5 seconds is very noticeable, and very demonstrably affects productivity. A change from 100ms to 50ms for a screen refresh makes a perceptible difference, but it is of a different kind. We're talking about how pleasant to use the system is rather than how fast it is. Just as most people think better in a noise-free environment (music is another topic), so too high resolution, smoothly moving flicker free images are less disruptive to thought than 10fps refresh rates and jerky scrolling and zooming.
If you want to test for whether a 100ms to 50ms change makes a difference, you're not testing whether the saved 50ms delay is saving you 50ms. You're testing whether the more comfortable environment is making people more productive. It's a different kind of test to a 10s to 5s test. You have to pay more attention to things that might make the environment less comfortable, like fan noise, flickering fluorescent and nearby conversations - they become relatively more important. At 10ms to 5ms you have to take that controlling of the environment to the next level - and you need to buy that 240fps monitor otherwise you'll see no change at all.
I learned, that in program usage, interaction delays below 0.2 s aren't recognised. A fast secretary makes 600 hits per minute, which is 10/s, most users won't reach 300/min or 5/s. A response faster than 0.2 s is felt as immediate response. You needn't be faster than that.
It depends on when the savings occur. CPM-GOMS studies (http://www.rpi.edu/~grayw/grayres/ernestine.html) showed that expected gains in efficiency of a new system may not be realized because many tasks may rely on other outside activities that will negate the savings. If your taks is highly sequential, you will see the gains, if not, it is less likely to be beneficial.
You could always argue that being a slave to the machine doesn't actually help you 'get more done during' in a day's work.
There are higher levels of brain activity than those involved in managing a software interface.
Not having 'instant gratification' can sometimes be a good aid to having to sit down and think a bit more deeply about something.