After studying this thread, I feel like I should weigh in.

The Massachusetts study is very interesting, but other studies (like the ongoing UNC study) show a significantly lower measure of overtake collisions (i.e., around 3%). Like Mark Twain says, "there are lies, damn lies, and statistics." All of these studies necessarily suffer sampling skew (like, only measuring emergency room visits, or police reports, etc.)

That being said, the risk of motorists failing to yield the right of way (left cross, pulling out from cross streets) is still the single greatest risk we face when riding after dark. I've given up riding on SR-8 between Beaverton and Hillsboro; even though it is well lit with a bike lane, the bright lighting and huge number of driveways for businesses work against me; I feel like I'm at risk about every hundred feet for a T-bone. (BTW, there are better routes out there anyway, but that's not the point I'm trying to make.)

So, shame on you folks who ride without a front light. You think that the front light is for you to be able to see? Hey, unless you're out in the middle of Yamhill County in the middle of the night, the most important purpose of a front light is for others to see you. There is a(n in)famous collision in Eugene Oregon in the 1970's. Night time: one bicyclist was cruising down the hill, the other pulled out from a cross street. Neither had lights. Both died.

No one has mentioned the specific advantage that a helmet mounted light gives. On almost every ride it seems I see a car pull up short when I lift my head slightly and look in their direction. When I'm very worried, I'll even shake my head as I look at them, thereby creating a flashing image.

Kudos for the anecdote about the dork who managed to drive full tilt into an emergency vehicle ensemble. To paraphrase, "there are none so blind as those who will not see." In spite of any attempts we make to become more visible, all of must--unfortunately--assume that we are invisible. Motorcyclists know this, and hey, they run a helluva lot more lighting than bicycles.

Next, there is a great benefit to redundancy. Lights fail. If you have only one light and it goes out, well, it sucks to be you. Bob Mionske points out in his book that having fully redundant light sources can even be part of a legal defense if you get hit, since it shows a heightened level of due care on your part.

There has been a discussion on bikeportland about the relative efficacy of blinking versus steady lights. Although blinking lights are arguably more visible, they can also be a distraction. To give you an idea of how serious this is, blinking lights attached to a bicycle are flat out illegal in Germany. International randonneur rules require that at least one of your lights, both front and rear, be steady--not blinking. YMMV, just give it some thought.

I'll conclude by sharing my lighting setup. I commute in a suburban area, with a mix of well-lit and very dark sections. I'll say that a January night in full rainfall with the cloud ceiling at about 150 feet is pretty adverse lighting conditions. On the front I run a MiNewt X2 on my helmet, another MiNewt on the handlebars, and a Trinewt on the handlebars. In full dark, I run all of these at full intensity, no blinking.

Rear facing I have a Planet Bike Superflash (wonderful!) doing its Annoying Strobe Thing and a Blackburn running as a steady point source. Finally, I have a second Blackburn attached to my left ankle in flashing mode. This adds the sinusoidal up-and-down motion characteristic of pedaling to the light.

P.S. -- Don't forget reflective clothing as well. I have reflective ankle bands and glo-glovs.