Which brings me to one of my favorite physicist stories ever. From Richard Rhodes' The Making of the Atomic Bomb, page 47, describing the 1907 work of Ernest Rutherford and Hans Geiger:
"There was a way to make individual alpha particles visible using zinc sulfide... A small glass plate coated with zinc sulfide and bombarded with alpha particles briefly fluoresced at the point where each particle struck, a phenomenon known as "scintillation" from the Greek word for spark. Under a microscope the faint scintillations in the zinc sulfide could be individually distinguished and counted. The method was tedious in the extreme. It required sitting for at least thirty minutes in a dark room to adapt the eyes, then taking counting turns of only a minute at a time -- the change signaled by a timer that rang a bell -- because focusing the eyes consistently on a small, dim screen was impossible for much longer than that. Even through the microscope the scintillations hovered at the edge of visibility; a counter who expected the experiment to produce a certain number of scintillations sometimes unintentionally saw imaginary flashes."
Thank god they invented the photomultiplier before I arrived on the scene.
Rhodopsin just like a photomultiplier detector has a certain noise level. The rate of noise counts decreases with temperature - so you would expect the performance of a person at this task to improve as you cooled their body.
This was actually done for toads. The experimenter would train a toad to lash its tongue at faint flashes of light and the performance of the toad at discriminating light from no light was measured as the toad's body temperature was cooled. As the toad was chilled the quality of its guesses improved. Interestingly it's error rate as a function of temperature matched the change in signal to noise ratio for the currents in the retinal cells. Bizzare! The toad's brain controlled it's tongue at the threshold of being able to statistically distinguish signal from noise.
Unfortunately this idea is impractical for warm-blooded creatures, whose body temperature is strictly regulated... at least until it isn't, at which point they're no longer much good for seeing photons because they're busy being treated for hypothermia.
Although I'm sure that Rutherford and Geiger, or their grad students, were quite happy to avoid the necessity of performing their human photomultiplier experiments while sitting in a chilled room.
How could they tell there were imaginary flashes? It seems like a private experiment, you can't check it. (Maybe there were not expected flashes at all and they saw some anyway?)
Well, I left out the next sentence of the book, in which Geiger builds an electronic counter (perhaps you've heard of it? :) and uses it to cross-check the humans.
According to the book, after using the counter to confirm that properly trained and rested humans were as accurate as the electronics, the next step was to put the electronic counter away because the humans could also give you spatial information: They could not only tell you that a flash had happened, but could tell you where it was in the visual field. Multichannel PMTs or (god bless 'em) chilled semiconductor-based multi-element low-noise photodetectors were still dozens of years in the future...
There really is no good way to tell if there were imaginary flashes. Once you get down to really dim (i.e. low photon count) flashes, you're in the realm of Poisson statistics. So, on average, a flash of light might have 4 photons in it, but it could easily have 1 or 7. Or 0!
"There was a way to make individual alpha particles visible using zinc sulfide... A small glass plate coated with zinc sulfide and bombarded with alpha particles briefly fluoresced at the point where each particle struck, a phenomenon known as "scintillation" from the Greek word for spark. Under a microscope the faint scintillations in the zinc sulfide could be individually distinguished and counted. The method was tedious in the extreme. It required sitting for at least thirty minutes in a dark room to adapt the eyes, then taking counting turns of only a minute at a time -- the change signaled by a timer that rang a bell -- because focusing the eyes consistently on a small, dim screen was impossible for much longer than that. Even through the microscope the scintillations hovered at the edge of visibility; a counter who expected the experiment to produce a certain number of scintillations sometimes unintentionally saw imaginary flashes."
Thank god they invented the photomultiplier before I arrived on the scene.