BP's examination of the Gulf of Mexico oil spill still hasn't determined precisely what caused the blowout, the company says. But industry investigators crawling over the Deepwater Horizon rig and its data are finding that an accident of this magnitude required an almost perfect chain of events: One unlikely malfunction needed to be followed precisely by another, then another, and so on, according to a preliminary investigative report I've obtained.

The report was produced earlier this month by one of the teams now examining every aspect of the catastrophic spill; I received it on the condition that I would disguise its provenance. Reproduced below is what is probably the most interesting piece of it: A diagram created by the investigators to illustrate just how improbable this event was.

Until now, public investigations into the April 20 spill have asked what went wrong. But what one gleans from this report -- and in particular this diagram -- is that that may be the wrong question. Instead we should be asking, how did everything that had to go wrong do so, and at the same time? For the well operators, the implications are not good: Unless one violated standard industry practice every step along the way, such a blowout may have been all but impossible.

The Wall Street Journal has provided the best investigative reporting on what went wrong on the rig, most crucially the failure of seals and the decisions made on the type and testing of cement. The New York Times did a fantastic job of examining the failure of the rig's blowout preventer, the five-story-tall device that is the ostensible last line of defense against such accidents. But until you see the sequence arrayed before your eyes -- a full account of what actually had to go wrong -- the enormity of the coincidence of events is extremely difficult to grasp. That's what this diagram -- labeled "High Level Root Cause Tree" in the original document -- does (click here for a more legible PDF):

You start from the bottom of the diagram, and move your way up. The key is the linking phrases AND, or alternately, OR. The ANDs mean that, in order for the described malfunction absolutely to take place, the one immediately below it must occur as well. The ORs mean that the described malfunction could take place, but wouldn't necessarily do so, if linked to the one immediately below it on the schematic. But if they do take place, there can be trouble.

As you see, the chain of events starts with many moving parts, then cascades out of control -- the seals on the head of the well; the tubular casing that's inserted into well as it's drilled; the cement used to seal the well and keep the natural gas under control; the "mud," a thick, complex, chemical-laced concoction used to lubricate and keep underground pressure from bursting to the surface.

Then comes a row of ORs: Do you have a failure to follow correct procedures when doubt is cast on your control of the pressure from the reservoir?  If the wellhead seal fails, do fluids from below reach the actual rig 5,000 feet above the ocean floor?

Then come some ANDs. If the fluids do reach the rig, and there is a source of ignition to cause an explosion, you get the blowout and fire.