David Coveney

Anatomy of a Traffic Jam

I’ve just spent far far too long doing these two illustrations. The first shows a traffic jam that’s a problem, caused by people not using up all the road. The second shows what happens if the two self appointed guardians of the road move up a little, don’t block people, and if other road users don’t all try and keep to the left.

Basically, during busy periods, it’s much much better if people use all of the road. Yet in Britain it’s a common scene to see a mile of empty right-hand-lane prior to roadworks. Consequently the traffic jam is far longer than it needs to be and, in many cases, the jam will go far enough back that it blocks a junction – causing a lot of people extra delays and frustration.

So let’s all try and help others on not by obsessively queueing politely in traffic jams, but by using as much of the road as possible.

Anatomy of a traffic jam, part one of two
Anatomy of a traffic jam, part two of two


Since the page went up I’ve been made aware of the following two links which are very interesting and have animations of some traffic problems:

Wave motions and how to prevent them
How increased spacing helps improve merging

Thanks to brangdon @ cix for those.

I’ve resized the images slightly in the browser to help them fit better on this theme.  If you want to view them full size or in better quality, right click (or do whatever Mac owners do) on the image and click ‘view image’ to see the full unadulterated version.

I’ve also added this post to the Campaign for Thinking… just because


Posted: 1 August, 2007 at 9:07 AM

Parrot says:

Well said mate.


Posted: 1 August, 2007 at 11:49 AM

Jim says:

Nicely put. Also, it’s often the case that they advertise a lane closure but haven’t yet put out the cones. If you stay in the outside lane longer you will discover this and can merrily give the queuing traffic the finger as you speed on by.


Posted: 2 August, 2007 at 2:18 PM

teedyay says:


The highway code explicitly says you should queue in both lanes and merge at the last opportunity.



Posted: 5 August, 2008 at 11:22 PM

Danny Bentham says:

Ay up Dave. I did my Bachelors Dissertation on Traffic Management for the Future with the idea being to assess the feasibility of using wireless LAN and Kalman filtering/stereo vision/collision avoidance methods for cars to autonomously guide themselves by communicating with vehicles in a radius of X. It turned out to be one big lit review in the end and I never had any money/time to build a small scale vehicle. Whilst reading around, I found a chap who did some modelling work using Cellular Automata to simulate traffic flows. Might be worth Googling that on your lunch break one day and see what you can find.


Posted: 5 November, 2008 at 4:17 PM

Matt Hopley says:

Hi dave. Nice images. interesting read thanks for that. ive just come across your site. Brings back memories of the dissertation i did on the physics of traffic jams… heres the link if you want to add it to your site or just to have a read http://www.cactusconnects.co.uk/physics_of_traffic_jams.htm


Posted: 20 November, 2008 at 4:30 AM

nick says:

traffic in example 1 will actually get the traffic through the lane closure zone faster.

example 2, causes unnecessary traffic backups due to the fact that drivers must take turns entering the lane closure which adds extra time.


Posted: 10 December, 2008 at 7:03 AM

nathan says:

Well done bro.


Posted: 3 January, 2009 at 11:44 AM

Ruth says:

Hello, how are u? Thank’s for interesting website, which will help me to write Master dissertation. May u have more ideas? :) how eliminate traffic conventions? :)


Posted: 8 January, 2009 at 6:20 PM

Dave says:

Hi Ruth – whatever you use of mine in your masters, be careful! I’m not qualified in the subject and this is all just my own theory!


Posted: 18 April, 2009 at 1:16 AM

John says:

Agree with Nick. Sorry, this is not always correct. It depends on overall traffic rate of flow.

If traffic is not coming along that quickly, example 1 can begin to clear the jam, and if people heed “merge left” signs a mile back, the traffic can resume its normal flow, so NO ONE is at a standstill. Why else do many merges have signs put up in advance warning motorists that a lane will end? If the best policy is always to merge at the last possible moment, why warn anyone?? The reason is that merging early will prevent a jam, and if the input is slow enough, it will clear a jam more quickly.

On the other hand, the delay caused by “taking turns” at the merge in example 2 can cause the jam to propagate for hours.

Of course, this breaks down if the input flow gets too high. In that case, motorists would have to merge into one lane too far back to be feasible, and any rogue motorist who doesn’t merge early enough or causes unanticipated braking will result in a jam.

Overall, if the jam is GROWING backward steadily, yes, it’s better policy to use up the road as you suggest. If, however, the jam is SHRINKING or relatively stable, then early mergers can help to break up the jam more quickly.

In my experience, people behaving as in example 1 is only a problem in cities during rush hour with lots of on/off ramps close together. Otherwise, better to behave as in example 1. You, as someone always behaving as in example 2, are the one slowing down traffic in other cases.


Posted: 10 May, 2009 at 9:49 AM

Dave says:

Hi John, yes, there are times when early merging is better than late merging.

The problem is that it’s really hard to know when is a good time to merge early or late – it’s often counter-intuitive and most drivers simply won’t know when’s best.

I agree that if the jam is shrinking it’s best to start to merge early. My example is based on experience of living in a very densely populated part of the UK, which is why it’s biased towards locations with lots of on/off ramps – I also agree that in other locations there are probably better strategies.


Posted: 7 June, 2009 at 4:59 PM

Ruta says:

Thank You Dave and John. Yes, maybe in my city, where I live (Vilnius, LIthuania)…this simulation and theory won’t work..But still, I will check it.
Of course, I’ll use and other theories and after all investigation, I’ll choose the best one. :)


Posted: 10 June, 2009 at 9:13 AM

reddenna says:

find the solutions for traffic flow in Bangalore- India


Posted: 7 July, 2009 at 2:56 PM

Dave says:

You erm, just need to draw some pictures. Maybe! But yes, it looks like something of a tricky problem that one….


Posted: 17 October, 2009 at 1:35 AM

John Armstrong says:

Every person who looks at these two images and listens to them is yet another idiot who makes commuting hell. Please, just take them down.

In the first image, people are flying through the bottleneck. This means that the red cars will arrive at their exit faster. In the second image, the “road police” have moved out of the way, and the cheaters have clogged the right lane, made 5 red cars reach their exit faster — and are now taking turns at the bottleneck, slowing down the entire highway, and making the 1000 red cars not in the picture come to a standstill.

> it’s really hard to know when is a good time to merge early or late

When there are only 50 cars on the road, and you want to make 5 red cars get to their exit sooner, and make everyone else go slower. So the answer is NEVER. Don’t follow this advice. The only thing that matters is the speed of the cars going through the bottleneck. Turn-taking slows this down, don’t let it happen.

Read http://amasci.com/amateur/traffic/seatraf.html again.


Posted: 17 October, 2009 at 7:53 PM

Dave says:

John – you do understand that I agree with the contents of the link you’ve attached?

Sadly, you’ve largely invalidated your argument by claiming that anyone who thinks other than you do is an idiot and a cheater. They aren’t. They just have a different opinion.

The key principle is to make sure that people merge politely, in turn, and not to turn driving into a game where we all feel the need to ‘win’ or that we somehow lose out by letting someone come in ahead of us. I frequently happen to be in the left hand lane. If someone in the other lane happens to be going more quickly… well, lucky them. Next time it’ll probably be myself that’s going more quickly. Win some, lose some.

You quite rightly point out that the speed of cars through a bottle-neck is the critical thing in many jams. However, not all cars have to go through that bottleneck. Otherwise the speed of an entire motorway network would be set by its slowest bottleneck, which clearly isn’t the case. My point is that once flow reaches a certain point any folk who believe they need to block a lane are unwittingly making life worse for other people. If you can show cases where people policing the lane that needs to merge makes things better then do illustrate it.


Posted: 17 February, 2011 at 2:52 PM

ROG says:

Lots of comments about those that blast down the right lane and then bully their way into lane 1 but that is not what is being advocated here

It’s not the actual use of the clear lane that is the problem but the ‘manner’ in which it is used.

By gently cruising (not blasting) by the traffic in the other lane and then using the indicator in plenty of time to ‘ask’ to be let in is the correct method which I have used for years in car and truck without any confrontations at all.


Have your say

This site uses Akismet to reduce spam. Learn how your comment data is processed.