Hi Roy,
Reach is the amount of detail you can resolve from a distant small object (such as a bird). The primary determinant of reach is pixel density. (Assuming, of course, that we hold other factors equal: we use the same lens for example, and shoot under similar lighting conditions.)
To a first approximation, then, reach is directly proportional to pixel density.
Reach has nothing at all to do with crop factor. The crop factor of a camera describes how much of the image collected by the lens you throw away because of having a small sensor. When you put a 400/5.6 lens on a 40D, for example, the sensor can "see" 39% of the light the lens collects and throws away the remaining 61% of the light, which falls outside the boundaries of the 40D's relatively small 337mm2 sensor. Put the same lens on a 5D Mark 1 and none of the light is thrown away.
This is why the full-frame 5D is regarded as an excellent camera for use in situations where heavy cropping is not a requirement: it uses more of the light, and captures it with bigger sensor elements which produce a more accurate, less noisy image. Used for (e.g.) a wedding, you can easily compensate for the wider field of view by using a longer lens. Bigger sensors, in short, are better sensors ..... provided that you are not focal length limited and can easily use a longer lens.
But when you consider the situation of the bird photographer, very often it simply isn't possible to use a longer lens and you have to crop hard before you print. Here we see the 5D Mark 1 at its worst. (Or the Nikon D3 and D700 twins - they are very similar in this regard.)
Imagine a bird that nicely fills the frame on a 350D with a 400/5.6. On a 350D, that gives you 8.2MP worth of usable picture. We threw away 61% of the light the lens gathered, but that doesn't matter - we were going to crop that part out anyway. Now, from the same place, take a picture of the same bird with the same lens, but use a 5D Mark 1. Initially we get to keep all the light (throw nothing away) but then we have to crop the shot back to the area of interest, which leaves us with just 5.0MP worth of picture. We have less reach. (In fact, we claw back a certain amount of our loss because those bigger pixels have less noise and better quality, but we are still well down on the overall result we got with the 350D.
Does this mean smaller sensors are better for bird photography? Absolutely not! Reach has got nothing at all to do with sensor size. It's all about pixel density: the number of pixels containing information about the bird. The 5D Mark 1 is a poor choice as a birding camera not because it has a large sensor - that is irrelevant as we crop away most of it anyway - it's a poor choice because it has low pixel density.
Now, let's try the same shot with a 5D Mark II. Once again we use the same lens and the same bird (by now we have nailed it to the perch, all in the name of scientific objectivity). Once again we crop away the part of the picture that does not contain any information about the bird, and we wind up with 8.2MP worth of useful information ...... exactly the same as if we had shot with the 350D. (Or with a 20D or a 30D, all three share an almost identical sensor.)
(Actually, if we want to nit-pick, it isn't exactly the same, the 5D II has marginally more detail: it returns 8,213,400 pixels where the 350D returns 8,185,344 pixels - but that's a difference too small to bother about.)
So far as reach is concerned, pixel density is the name of the game. The 50D is the current champion with 44,605 pixels per mm2; next comes the 450D with 36,049 px/mm2; the 40D, 400D, and 1000D with 29,860 px/mm2; the 1Ds III and the 5D II have 24,336 px/mm2; the 20D, 30D, and 350D all have 24,253 px/mm2, and the 1D III has 19,178 px/mm2.
Now, let's consider the secondary factors. I said earlier on that pixel density is the primary determinant of reach, but there are other factors. For example, under identical conditions (same lens and same long-suffering bird as before) a 20D will resolve detail about as well as a 40D - although it has fewer pixels, each one works more effectively. (I'm not sure why, but careful comparison between my two 20Ds and my two 40Ds confirms this: there is nothing to chose in the overall result. I suspect that the 40D went a fraction beyond what was sensibly possible so far as sensor design went at that time, where the 20D/350D/30D sensor did, and still does, punch harder than its weight.) Similarly, the 1D III, despite having significantly lower pixel density than a 20D or 40D, produces almost the same amount of detail, and does it with better colour and contrast.
A year or two ago, we all though that pixel density had reached its limit. Because the 40D was unable to improve on the 20D/30D despite having more pixels, we all thought that 25,000-30,000 odd px/mm2 was the limit. Beyond that, we thought, we were into a game of diminishing returns.
Then the 36,000 px/mm2 450D came along and made everybody wonder, and now the 44,600 px/mm2 50D has decisively demonstrated that it can return more detail per mm2 than any previous SLR. Clearly, the sensor manufacturers have learned another trick or two. This leaves the 50D as the current "reach champion". I haven't used one but I imagine that the 450D comes next, and on the face of things we have an 8-way tie for third place: 350D, 20D, 30D, 40D, 400D, 100D, 5D II, 1Ds III. All eight are around the 25-30,000 px/mm2 mark. You could argue that the 10MP 40D, 400D and 1000D are in front, but many observers (I count myself amongst them) do not think there is any genuine difference in reach between (e.g.) a 20D and a 40D. Or you could throw another, as yet unmentioned but very, very important variable into the mix: focus accuracy. Of those eight just mentioned, the 1Ds III has by far the best autofocus system, and a perfectly focused 24,000px/mm2 shot will beat a slightly blurry 29,000px/mm2 shot every time. A 1Ds III, in other words, has a better chance of producing an acceptably sharp image of a distant bird than a 40D, or indeed a 350D. And that - producing an acceptably sharp image of a distant bird - is what reach is.
As a footnote, I think it's this last reason that keeps my 1D III in service as my front-line birding camera. In theory and in formal testing it resolves no more detail than my 20D or 40Ds, and it certainly resolves less than the 50D does - but in real-life conditions, the 1D III focuses faster and gets an accurate focus more often, so I mostly go to it as my first choice. Sometimes, however, I resort to the 50D for extra reach, especially when I have a static subject with good contrast for it to "grip" on.