"The country is not a one-size-fits-all, a one-description-fits-all. It's much more a mosaic," the U.S. official in charge of training Iraqi security forces, Lieutenant General James Dubik, told military analysts today on a conference call. And he's got a point. So maybe it's fitting that the Pentagon's last two quarterly reports show all sorts of unexplained shifts -- even on the exact same pieces of data.

Brian Katulis, a senior fellow at the liberal Center for American Progress, told reporters at a National Security Network briefing today that the Pentagon reports can't keep their stories straight when it comes to the incidences of sectarian attacks and murders. Take two most recent reports, from March (pdf) and June (pdf).

On page 17 of both reports is a graph entitled "Sectarian Murders and Incidents" that tallies sectarian attacks by month. The March report lists that, for instance, December 2006 hosted over 900 sectarian "incidents" resulting in just under 1300 murders. But in the June report, the numbers shade up: December 2006 hosted over 1000 incidents yielding over 1600 murders.

Similarly, the March report listed a decline of about 150 sectarian murders from September to October 2006. But the June report changes that to an increase of nearly 400 murders during that same time period. Speaking generally, the June report makes 2006 look like a more deadly, sectarian year than did its March predecessor, but there are exceptions: April 2006 had 700 sectarian murders in the March report, but somehow, that figure drops to under 400 murders in the June report.

So what accounts for the difference? Katulis says he doesn't know, and hasn't been able to get an answer from the Pentagon. Two possibilities present themselves: either Multinational Corps-Iraq, from whose database the statistics emerged, changed its definition of "sectarian" incidents and murders; or new information became available after March. Whatever the answer, a reader of the June report doesn't have any way of knowing that the March report gave different statistics on sectarianism.

Similarly, the November 2006 report (pdf) contains seven mentions of the term "death squads." ("Armed groups that conduct extra-judicial killings" is the definition given by the November report.) Yet in March, the term only appears once, with no given definition -- it appears in scare quotes, even -- and in June, the term vanishes from the report's lexicon. Again, there's no stated reason why the term vanishes: it's not as if Iraq is suddenly free from death squads. In many cases, as the draft GAO report leaked to The Washington Post today reveals, they've embedded within the Iraqi security forces we're building.

The question immediately raised by the retroactive changes in statistics on sectarianism and other key measures is whether the next quarterly report, due in September, will similarly juggle the data without notifying readers to the change. "We have to be able to understand the models, and have an apples-to-apples comparison," Katulis says, in order to make sense of what Dubik called the "mosaic" of Iraq. Right now, the quarterly Pentagon report looks less like a mosaic than it does a Rorshach test.