The Red Myth
Unaided night vision even now in the 21st century is still the subject of some controversy.
For those just looking for an executive answer as to what supplemental lighting should be used to reduced the recovery time back to night vision (dark adapted or scotopic) here it is: a fully dimmable white light! This of course is a very incomplete answer but so are the answers red or blue-green and you should know why.
Lets start with red, specifically what I will call the red light myth.
I believe the myth started in the photographic darkroom.
Until about 1906 most photosensitive material (plate, film, and paper) was not very sensitive to red. Some of these orthochromatic materials are still used. This allowed these materials to be dealt with for a short time under a relative bright red light because the human eye can see red if the level is bright enough. The fact that L.E.D.s (having a number of advantages over other light sources) were economically only available in red for some time has also help to perpetuate this myth.
As more research about the eye was done it was found that the structure responsible for very low light vision, the rods, were also not very sensitive to red.
It was assumed then that like film you could use red light, which is seen by the red sensitive cones (there are also blue and green sensitive cones to give color vision), without affecting the rods.
It takes a while for true night vision to be recovered. About 10 minutes for 10%, 30-45 minutes for 80%, the rest may take hours, days, or a week. The issue is the chemical in the eye, rhodopsin - commonly called visual purple, is broken down quickly by light. The main issue then is intensity; color is only an issue because the rods (responsible for night vision) are most sensitive at a particular color. That color is a blue-green (507nm) similar to traffic light green (which is this color for a entirely different reason). It would seem that using the lowest brightness (using this color) additional light needed for a task is the best bet to retain this dark adaptation because it allows rods to function at their best.
Unfortunately there are a number of drawbacks using only night vision.Among these are:
If you need to see directly in front of you or see detail you need red. Like many myths the red light myth has some basis in fact.  The red truth?
Why red? The center 1.5% of your retina (the fovea) which provides you with most detailed vision is packed almost exclusively with red sensitive cones.
This is the same area that has no rods and is responsible for the night blind spot. There are fewer total green sensitive cones than red. The number of blue sensitive cones is very small compared to green and red.
Which is just as well since the lens in the human eye cannot focus red and blue at the same time. And using green really only changes perceived brightness because of the way the signals are processed in our neural pathways. Unlike a digital camera, more pixels, in this case, doesn't give us more detail.
Chart showing the distribution of rods vs cones. Note the absence of rods in the center and the absence of both about 15° away from the the center toward the nose where the optic nerve passes.
At first glance the tendency would be to pick the hue of red at which we are most sensitive (566nm) which would make sense except for the real reason: we don't want to involve the rods. The reason is the rods share the neural pathways with the cones so that you have this fuzzy image overriding the detailed one. This effect disappears at slightly higher mesopic levels which is why white is a good choice for most tasks. Many people look at the numbers for sensitivity for rods and cones and forget that in most cases the numbers have been adjusted so that rod peek sensitive matches cone peak. Rods are in fact sensitive well into the infrared (not too useful except to know that light you can barely sense can adversely impact your night vision). The key then is finding a hue that we can have at a high enough intensity that we can see the detail we need without activating our rods to the point were they obscure that detail. Most source say this should be nothing shorter than 650nm. Experimentation shows a L.E.D. with a peek around 700nm seems to work best (perceived as a deep red). Note that red may be fatiguing to the eyes.
True night blindness is rare. Most of what people call night blindness is either a lack of vitamin A in the diet or a failure to understand the night blind spot.
Cataracts, even minor ones, increase the effects of glare at night and the eye's lens does yellow and passes less light as we age which may contribute to what some call night blindness.
Note: The red filtered light at the intensity most people use is likely decreasing night vision much more than a properly dimmed white or blue-green light would!
Note: There are day blind spots also but are in a different position in each eye so are less of a problem.
Note: Blue-green (also called cyan, turquoise, teal and other names) as used here is NOT the combination of two colors but is a single particular hue. I use the most common name for that hue.
This is intended only as an overview, no warranty of this information is expressed or implied
[Update 17 Nov 2003] I find new myths are springing up. Such as blue-green L.E.D.s are emitting two colors of light. This is a mis-understanding of the color name and that this is the most accepted name for this one color. Another is that blue improves night vision. While at somewhat higher levels it, of course, is stimulating the rods. It is not an optimum color. Another long standing myth is that human visual perception is based on three colors when it is really based on four. The rods are usually ignored because many people believe, wrongly, that at the brightness at which we perceive color the rods are no longer providing our brains with any information. In fact the perception of brightness is highly influenced by the rods well into the photopic (bright light) range of vision. Fluorescent lamp manufacturers have used this knowledge for a long time. "Cool White" lamps have an additional amount of green phosphor added to make us "see" them as being brighter! Of course the whole subject of color vision and the variances thereof (wrongly called "color blindness") will require a number of new pages even in synopsis form.
A point I forgot to cover is that to help preserve night vision in one eye the other may be closed or covered if you know your are about to be exposed to a brighter light, such as from a oncoming vehicle. For normal observation both eyes should be kept open. If it is difficult to concentrate on the desired image the eye not being used may be covered but not closed. Closing affects focus and possibly acuity.
[Update 14 Dec 2003] A very important point barely mentioned in the original is that human peripheral vision is almost completely rod based! The implication then is that we cannot see color at the edges of our vision. If you think we can, try this simple experiment. You will need a small assortment of color cards (try sheets of construction paper) and someone to assist you. Sit looking straight ahead while your assistant, about 6 to 10 feet away, slowly moves a random color card into the margin of your vision. Now, while still looking straight ahead, what color is is the card?
This is the second most important factor that has been ignored in the design of outdoor lighting, the first being glare! However this study (in pdf), at the U. S. Dept. of Transportation, is a subjective study of blue tinted headlamps.
[Update 23 Jan 2004] A few random notes to be better integrated into this document later.
Luminances are approximate and will vary with the individual and conditions.
Vision luminance rage 1 * 10-6 to 1 * 106 cd/M2
Rods luminance rage 1 * 10-6 to 1 * 103 cd/M2 (may still play a roll above this range)
Cones luminance rage 1 * 10-3 to 1 * 106 cd/M2
Explain "Purinke shift"
20/20 vision is the ability to resolve 1 minute of arc at 20 feet.
Discuss Ricco's Law.
Saw the yahoo search string: "Do color blind people have better night vision?", which is an interesting question. Those with the genetic factor shifting the L (red) cones toward green are accepted as usually having better twilight vision, however the only place I've seen a suggestion of improved night vision is in the Wikipedia!
References (all external links open in new window - Not responsible for the content of any outside links)