Most people try to determine the WTC collapse times by looking at video clips, but this doesn't seem to produce very accurate results. Resolution and smoke can make it hard to pick out the beginning of the collapse, dust clouds hide the end, and so estimates range from around 8 seconds to 16, 17, even more.
We can't provide a definitive answer to the collapse question, but maybe there's another way to approach the problem. There are audio clips available, made close to the collapse points, which record lengthy portions of what happened. If we can determine the first and last collapse sounds, then this will obviously give us an indication of how long the towers took to fall.
Problem #1: determining the start time
This is plainly not a perfect method. When estimating the start of the collapse, for instance, we need to bear in mind that sound will take time to reach the recording device. That is, the collapse will have begun before we hear it: the audio method will shorten the collapse time.
On the other hand, it could be argued that if the towers were demolished, then we may hear explosives before the collapse begins. In this case the true collapse could begin after the explosions: our audio method will lengthen the collapse time.
Problem #2: determining the finish time
Picking out when the collapse ends from audio alone is even more problematic. Echoes, reverberations, perhaps the sound of secondary collapses or explosions caused by flying debris, there are a number of reasons why sound might persist after the point when the tower collapse has actually ended.
Problem #3: questioning the source material
If someone edits the images in a video clip then it could be easy to spot, but the audio is a different matter. Are we really listening to an accurate account of what happened? It's difficult to tell for sure.
Problem #4: accounting for the speed of sound
Imagine a fixed microphone at the base of the towers, recording the collapse. The sound of the fall beginning has to travel from high up, perhaps taking close to a second to reach the microphone. The final sounds originate much closer, as rubble hits the ground, therefore an audio recording will tend to understate the collapse time. (Although if the microphone is moving as someone’s running away then this effect will be reduced).
But does it matter?
There's no clear way to resolve any of these issues, and in fairness they’re not that different to other commonly used methods. Many people use the seismic records to try and assess when the collapse started and ended, for instance, but there’s no clear indication of either event on the charts themselves. It’s all interpretation.
Still, if we just look at this as an experiment then maybe it doesn't matter. Let's just say we want to see if the audio evidence tends to support the 8, or 17 second end of the collapse time spectrum, and leave it at that.