The Washington PostDemocracy Dies in Darkness

The weird, counter-intuitive science of traffic jams

Placeholder while article actions load

This is a fun 23-minute presentation from Tom Vanderbilt on the often-bizarre psychology of driving and why traffic jams happen:

Obviously one way congestion can happen is if there are too many cars and not enough road. But there's more to it than that. Eric Jaffe of Atlantic Cities has already done a nice job summarizing the talk, but I'll pull out a few more interesting tidbits:

— Self-driving cars steered by robots could do a lot to reduce traffic jams. That's because the human inability to maintain a steady, constant speed on the road is responsible for a lot of congestion. Japanese physicists discovered this when they had people try to drive around in steady speeds on a circular road. Jams materialized out of nowhere. People braked erratically and started responding uncertainly to people ahead of them.

So perhaps robots can do better one day. Vanderbilt describes sitting in one of Google's autonomous cars a few months ago and watching it maintain a reliably steady distance in front of it. In theory, autonomous cars could also communicate with each other — something that human drivers struggle to do.

— We're all basically idiots when it comes to merging. Whenever a road shrinks from two lanes to one, it would actually be most efficient for everyone to stay in their lane right until the point where the lanes converge and then execute a "zipper merge." But most drivers feel bad about doing this and tend to shift over to the open lane early. This causes congestion.

"My main takeaway from this is that the individual driver can often not understand the larger traffic system," says Vanderbilt.

— Cars tend to drive closer to bicyclists who are wearing helmets. That comes from Ian Walker, who set up a bicycle with sensors and drove around the city. Vehicles tend to crowd closer to him when he was wearing a helmet than when he wasn't. That's not necessarily surprising, but it's a reminder of all the weird unconscious tics we adopt while driving and making on-the-fly assumptions.

— There are all sorts of fun patterns in who honks their car horns. In what sounds like a exciting job, researchers sat at intersections and refused to move when the light turned green to see who honked at them. Men honk more quickly than women. People with expensive cars also honk more rapidly — although people in convertibles are less likely to honk.

— There are also fun patterns on driver courtesy. Older drivers are more likely to stop for others. Drivers are more likely to be courteous when the other car has extra passengers inside. People also are more likely to violate traffic rules the closer they are to home — a "familiarity effect."

—  There are all sorts of optical illusions that can trick human drivers. Fog makes objects seem like they're moving slower than they really are. And experiments show that humans are really bad at judging the speed of an oncoming train at a crossing until it's nearly arrived. Another point in favor of self-driving cars, perhaps.

—  Congestion often looks tantalizingly easy to clear up — in theory. One study that tracked drivers in Boston during rush hour found that if you could remove just 1 percent of people on the road (say, to mass transit), you could achieve a whopping 18 percent improvement in traffic flow. But  for whatever reason, cities haven't figured out how to do that just yet.

Anyway, a lot of these points are also in Vanderbilt's excellent 2008 book, "Traffic: Why We Drive the Way We Do (and What It Says About Us)." But if you want a quicker version, this talk is pretty entertaining.

Further reading: Seeing as how self-driving cars are a ways off, here's a look at other ways traffic experts have tried to harness science to improve the flow of cars and trucks on the roads.