Ever wondered what the "Base ISO" (or "Native ISO") of your camera is? No? Neither had I until today. If the question doesn't make sense then it may help if I describe how ISO is changed by the camera's electronics. This description will, no doubt, have the experts in the field gnashing their teeth in horror but I hope I have represented the essentials correctly.
Each of the pixels on a camera is a little photo-sensor which frees electrons as photons of light are absorbed. The process isn't anywhere near 100% efficient but the number of electrons freed is proportional to the number of photons seen and those electrons are stored at each pixel until they are read out by more electronics and an analogue to digital converter converts the resultant voltage into a number, usually with 12 or 14 bit resolution. The trick is to realise that the sensitivity/efficiency of the photo-sensor doesn't change when you change the ISO number. What does change is the amount of amplification given to the voltage from the photo-sensor which then goes to the A/D converter.
With that in mind it is obvious that, for a given light intensity on the sensor, regardless of the ISO number you dial in, the individual pixels will always take the same time get to a "full well" state, that is a state where they can't store any more electrons. Of course the resultant signal can't be properly used if you dial in a high ISO number because the subsequent voltage amplification will either clip the analogue signal or the resultant digital number will hit the 12 or 14 bit maximum (blown highlights).
So how does this help define "native ISO"? Unfortunately I haven't been able to find a proper definition so how about this? A sensor's native ISO is the same as the ISO number of a camera film which, when exposed to the same light intensity, reaches full saturation in the same time as the digital sensor reaches "full-well". By "full saturation" I mean that the photosensitive chemicals in the film have all been activated by incoming photons. A pretty sloppy definition, no doubt, but I hope it will do for the job in hand.
The reason why native ISO matters is that when you take a photograph at this ISO number the highlights will, if the picture is correctly exposed, have just resulted in an almost full-well number of electrons at each relevant pixel. This means that the statistical noise is lowest and the noise introduced by the read electronics is low in proportion to the signal. If you dial in a higher ISO than the native ISO the amplifiers have to work with an input signal which may, for really high ISO numbers, be much less than full-well (less photons equals less electrons) and so the underlying noise will be higher in proportion to the signal.
"So what," you yawn. We all know that using high ISO numbers results in more sensor noise. More surprising, perhaps, is that using an ISO number below the native ISO number also degrades the image. The reason is that you are effectively telling the camera to blow the highlights as the brightest elements of the picture will get to full-well before the exposure is over. It gets worse because the amplifiers are now set to a lower than optimal gain meaning that, as I understand it, shadow detail can also be lost into the noise threshold of the electronics.
In case you think this is academic have a look at this diagram for the Canon EOS 1Ds MkII, where halving the ISO number from 100 to 50 results in a S/N ratio equivalent to working at ISO 800!
Courtesy of The Luminous Landscape
One final thought concerns why so many cameras have an ISO setting of 100 or 200 as their lowest standard ISO number. Looking at this table it would seem that, to take the case of the Nikon D300, the native ISO of the sensor is actually about 140. That gives every reason for Nikon to warn against using ISO 100 but why have a lowest recommended ISO of 200? I have no idea.
Edit: Thread title changed to reflect use of "Base ISO" as well as "Native ISO"