To fully understand Anon's response I think we have to delve a bit deeper into the subject of exposure. I'm going to get a bit technical so hold on to your hat. I'm going to go into exhaustive detail a) because I've got some time to kill right now, and b) because the question that is vexing you simply won't go away unless you really understand these concepts. I'm an artsy fartsy painter-type, not a techie, so what I'm going to write here will almost certainly cause a five alarm fire to start at Techical Headquarters, but as we used to say after tuning our guitars: "Hmmm, well, I guess that's good enough for folk music." ;)
The imager in a digital camera (the CCD or CMOS chip) has millions of light sensitive spots called photodiodes. OK: I know that word photodiode is scary. But let's take it apart. The photo part is simply the Greek for light. The di part is the Greek for the number 2. And the ode part is simply a form of poetry originally invented by the ancient Greeks to woo handsome athletes. So a photodiode is simply a bit of electronics that converts light into two line poems (poems that only an engineer could love). Here's how that works:
Each photodiode reacts to in-coming light "particles" (photons) by generating electricity. The more light that reaches a photodiode, the more electrical energy the photodiode accumulates, just like a wool skirt building up static cling. The problem is that there is an upper limit and a lower limit to this process. Like a water glass, the photodiode can only hold so much electrical energy. If too many photons hit the photodiode it reaches that limit and if still more photons arrive it cannot deal with that - it gets stressed out and freaks. At the other extreme, the photodiode is never quite an empty glass - it always has a small amount of electrical charge in it; this is called noise. So if too few photons hit the photodiode there is no way to know whether the small amount of electrical charge it contains has got there from light striking it or not. This is called the noise floor. (We can tell that electronics engineers are sloppy house keepers because they would have called this a dirty floor instead of a noisy floor if they had ever used a mop or a broom.)
Now that you understand how a photodiode works, all you need to know is that each pixel in your image file is a number that corresponds to the amount of electrical charge in one photodiode (actually this part is a lie because I'm leaving out how the numbers get translated into colours). If only a few photons hit a certain photodiode, then the corresponding pixel will be a small number, which means a dark colour. If scads of photons hit that photodiode, then the corresponding pixel will be a large number, which means a bright colour. (So now we know the dark secret of digital photographers: all they do is paint by number.)
We can also see that there is a maximum brightness number that a given photodiode can generate. This is pure white. And there is a minimum brightness number that a given photodiode can generate. This is NOT pure black because of the noise floor. If you subtract the noise floor number from the maximum brightness number you get the exposure latitude of the imager, which is usually called dynamic range. The dynamic range of most good digital imagers is something around seven stops. Whether this is enough to capture both the highlights and the shadows of any given scene depends on how contrasty the scene is. If the dynamic range of the scene is greater than the dynamic range of your camera's photodiodes then something will get clipped. If you waste the dynamic range of your camera by setting the exposure too high or too low, then something will also get clipped.
To avoid setting the exposure too high or too low, we have two tools: the exposure meter and the histogram. A histogram of an image shows how many pixels are dark, how many are white, and how many fall in the large grey area in between those extremes. My Pentax DS, like most dSLRs, can show me a histogram of a picture after I've taken the picture, IOW, a dead histogram. I can see how the pixels spread from black to white and if they are bunched up at either end I know I need to re-take the picture at a different exposure. Of course, sometimes that's not possible. So if I have a camera with a live histogram I can see the pixel spread just before I take the picture and adjust my exposure accordingly. And again sometimes that's not possible: you can't follow the action in a ball game and fiddle with the exposure at one and the same time.
But as Anon points out, there is another issue with the in-camera histogram that affects its usefulness - it is based on the JPEG version of the image whether you are shooting JPEG or shooting RAW. To create a JPEG the camera has to make certain guesses as to how you want to interpret the scene. As I'm writing this Tim Gray has just posted a link to an example of why this can be a problem. I haven't seen this particular problem in non-Canon cameras, but what I have seen is similar. To create a JPEG in the camera most manufacturers cut out the darkest stop or so of photodiode data to eliminate a good chunk of shadow noise. They also don't try to extrapolate the brightest stop of data when one or two colour channels is clipped like a good RAW converter will. This knocks your dynamic range down from seven or eight stops to six. They also apply an aggressive tone curve to boost contrast, although some cameras allow you to control how aggressive that curve is.
Now, none of this matters if you are dealing with a low contrast scene. None of this matters if you are dealing with a static subject, bracket your exposures, then re-combine them in Photoshop. But all too often in outdoor photography, if you take JPEGs with a compact camera, you are going to find either your highlights have been clipped, which often equates to a pure white sky; or you are going to find your shadows are inky ... or both in the same shot.
A live histogram can be a useful tool to help you make sure you get the best exposure for a contrasty scene, but it can't overcome the built-in clipping limitations of an in-camera JPEG. What might be better here is if your camera can do automatic exposure bracketing without filling up the image buffer.