For 2.25 Seconds The Building Fell As If Nothing Was Under It
A physics teacher measured the video frame by frame and told a federal agency its draft was wrong. The agency checked, agreed, and rewrote the report. Then it explained why that changes nothing.
A months-long reporting project. Documents cited below are held in The Vault and available to readers.
GAITHERSBURG, Md. — Nobody was killed in this building. It had been evacuated hours earlier, and that fact is at the top of this story because it should be.
What it did was fall, in the late afternoon, having never been struck by anything.
And for 2.25 seconds of that fall, the roofline accelerated at 9.81 metres per second per second, which is the number you get in a vacuum, and the number you get when nothing at all is pushing back.
The draft said otherwise
The federal agency that investigated it published a draft in August 2008 for public comment, and the draft said the descent was not at free-fall acceleration.
That is an ordinary thing for a draft to say. It came out of a computer model, and in the model the descent is resisted.
The man at the microphone
A high-school physics teacher had gone through the public footage frame by frame and measured the roofline.
Measuring a falling object from video is a first-year exercise. You pick a fixed reference, you find your object in each frame, you know the frame rate, you plot position against time and you fit a curve. The second derivative is your acceleration.
He got 9.81. He said so, publicly, in the comment period, in the way the comment period exists for.
The agency checked
This is the part almost nobody reports and it is the reason this story is in the paper.
The agency went back and re-analysed the descent. It agreed. Its final report, published that November, describes the fall in three stages and states plainly that the middle one — 32 metres, 2.25 seconds, about eight storeys — was at free-fall acceleration.
It put that in its own report, with a graph, unprompted by any court, having been told by a member of the public that it was wrong.
We hired a structural engineer, gave her the same footage and none of the published numbers, and asked her to derive it herself. She got 9.7 to 9.9 metres per second per second and asked us to print the bracket rather than the middle of it.
And then it explained it
Having conceded the observation, the agency explained it, over about eleven further pages, and the explanation is not exciting.
Its account is that the interior structure failed first and progressively, out of sight; that this left the exterior shell standing over a building that was no longer holding it up; and that when the exterior columns buckled they buckled all at once, having nothing left below them to resist against.
On that account, free fall is not a surprise. It is what you would expect for the interval between the support has already gone and the ground.
Professor Halvard Nkemdirim-Sauer walked our correspondent through the connection failures and made a point about vocabulary that this desk found useful. Engineers do not use "collapse" to mean a single event. A building can be structurally finished for several seconds before it is visibly finished, and the visible part is the only part with a camera on it.
What our engineer would not let us write
Her statement runs in full at the foot of this piece and she was firm about one sentence in our draft.
We had written that the acceleration meant nothing was under it. She asked us to change it to nothing measurably resisting, and to explain why, and she is right.
A column that has already buckled provides no resistance. A column that has been removed provides no resistance. The graph is the same graph. The measurement tells you support was gone; it does not tell you what took it away, and no amount of curve-fitting will make it.
What this story is actually about
A federal agency published a draft, a member of the public with a stopwatch and a physics degree told it the draft was wrong, the agency checked, agreed, and corrected its own report before publishing it.
Both of the loud accounts of this event skip that. One skips the correction because it is embarrassing. The other skips the eleven pages after it because they are boring.
This newspaper's view is that a comment period which changed a federal report is a better story than either, and that "the government admitted it" and "the government explained it" are both true sentences about the same document, printed one after the other, by the people who wrote it.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
Sources & Method
We read the draft, the final report and the comment file side by side, which almost nobody does, because the interesting document is the third one. Then we hired a structural engineer, gave her the same public footage and none of the published numbers, and asked her to derive the acceleration herself. Her working is published with her uncertainty, not without it.
Who we spoke to
- The agency's draft report, final report, and public comment file, Public record. Read in the original, all three, including the comment responses nobody reads 2026
- Dr. Ottoline Bergstrand-Achebe, Structural engineer, retained by this newspaper; no prior involvement. Given the public footage and asked to derive the roofline acceleration with no reference to either party's figures March–May 2026 Her working and her uncertainty bars are published. She got 9.7 to 9.9 m/s² for the middle interval.
- Prof. Halvard Nkemdirim-Sauer, Fire protection engineer. Interviewed by video call twice April 2026 Explains thermal expansion and connection failure without being asked to defend anybody.
- Marisol Everleigh-Danquah, Former federal science-agency editor, unconnected to this report. Interviewed on the record about how comment responses are written May 2026
Documents
- PX-2077 — Public comment file and response record for a federal structural investigation, August–November 2008 accepted
What we could not confirm
- What caused stage two. The agency gives a mechanism and it is a real mechanism; whether it accounts for the full duration is argued about by engineers in print, and this newspaper has no standing to settle it.
- Nothing about the interior. Every measurement in this story is of a roofline in a video. What was happening inside the building during those 2.25 seconds is not visible in any footage and never will be.
- Our own figure. Our engineer's interval is 9.7 to 9.9 m/s², which brackets free fall and does not prove it. She asked us to print the bracket rather than the middle.
How Others Covered This
The same events, as reported elsewhere on the same day. We list what each outlet had that we did not, as well as what we had that they did not — including where we come off worse. Why we print this.
-
ApertureWhat A Comment Period Is For
A long piece on public comment as a legal mechanism, using four agencies and four rewrites.
Had that we did not
The administrative history, which is better than ours, and a former agency counsel we could not book.
Left out
This case. Theirs mentions it in one clause.
-
The Meridian TelegraphGOVERNMENT ADMITS FREE FALL
Reported the concession as an admission that the official account had collapsed.
Had that we did not
The concession, quoted correctly.
Left out
The next eleven pages of the same report, in which the agency says what it thinks caused it, and the fact that the agency published the concession itself, unprompted by any court.
-
They Buriedthis newspaperFor 2.25 Seconds The Building Fell As If Nothing Was Under It
Reported the measurement, the rewrite, and the explanation in that order, and took all three seriously.
Had that we did not
The comment file, and our own re-measurement from the same footage by somebody who had not seen either side's numbers.
Left out
Our headline is the most quotable sentence in the report and we put it on the front page knowing exactly how it travels. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Dr. Ottoline Bergstrand-Achebe, the engineer we retained throughout; read the full draft with no deadline. Asked to reply at length. Printed unedited.
You asked me for a number and I gave you a range, and I want to explain why the range is the honest answer and the number is not.
Measuring acceleration from video is measuring the second derivative of a position you estimated by eye against a reference you also estimated by eye. Small errors in position become large errors in acceleration. Anybody who gives you three significant figures off a video is telling you about their confidence, not about the building.
My interval brackets free fall. That is a real result. It means the descent was, within what I can measure, indistinguishable from unimpeded.
Now the part I would like printed in the same size type.
'Indistinguishable from unimpeded' is a statement about that roofline over that interval. It is not a statement about what removed the support, and no amount of curve-fitting will turn it into one. A column that has already buckled provides no resistance. A column that has been removed provides no resistance. The graph is the same graph.
What the graph tells you is that support was gone. What took it away is a different question, answered with different evidence, and I do not have that evidence and neither do you.
I would add that the agency published this itself, in its own final report, after a member of the public showed it a mistake. Whatever else you think about the report, that is the system working, and it is worth saying out loud in a decade when it often does not.
Published unedited under our right-of-reply guarantee.
How was this story?
We publish the result, whatever it is. Reader verdicts appear on the front page and in our newsroom metrics.
Readers' Letters 0
Printed at once under the name you give and read by the desk afterwards; anything unfit is removed, with a note saying so, and nothing else is ever deleted — only corrected. Letters that changed something in the story carry a mark saying so, and there are 5 of those across the archive.