Car drivers complain about the behaviour of cyclists: they ride two or three abreast and hinder passing, don't signal their intentions and run red lights. In return, cyclists complain that cars pass them too closely, open their doors without checking their mirrors and won't share the road.

And on the paths that cyclists share with pedestrians, walkers complain that cyclists come upon them without warning, shout abuse at them for not keeping to the side of the path and make walking with dogs difficult. Cyclists, for their part, complain that walkers wander across the path without concern for other users, dawdle in gaggles, and generally block the smooth flow on the path.

As a keen cyclist, walker, and, indeed, driver, I've seen all these displays of intolerance, rudeness and more. There is a lot of heat, but not much light. So I decided to collect some data to test the hypothesis that cyclists, drivers and walkers - all being members of the same community - are pretty much the same when it comes to their behaviour.

I thought I'd find that most folks - whether drivers, riders or walkers - are pretty civil and courteous, even though there are a few selfish and thoughtless individuals out there.

But I was greatly surprised.

I collected the data on my bike rides around the inner south on roads and shared paths. I homed in on two behaviours - one each for cyclists and drivers - that were easy to observe and, importantly, were formally similar.

On shared paths, cyclists are required to alert a walker (or indeed another cyclist) that they are overtaking by sounding their bell. On roads, drivers are required to give a cyclist that they are overtaking at least a metre's berth.

The rules are similar in the sense that they place the primary obligation on the faster, more energetic vehicle coming from behind to manage the overtaking manoeuvre safely. There's no comparable "rule" for walkers, so I couldn't include them in my experiment.

Over several weekends, I observed these overtaking interactions by both cars and bikes. I scored successive interactions until I had about 30 of each, needing this many for satisfactory analysis. What I scored was: "Did the bike/car observe their rule for walkers/bikes when they overtook?"

The results, to say the least, are interesting. I observed 33 successive events of cars overtaking bikes, of which 30 drivers gave the cyclists at least one metre's berth, while three passed too close. And I observed 24 successive events of cyclists overtaking walkers, of which two sounded their bell, and 22 just whizzed past.

I repeated the experiment some months later. I observed 38 successive events of cars overtaking bikes this time, of which 34 drivers gave the cyclists at least one metre's berth, while four passed too close. And I observed 33 successive events of cyclists overtaking walkers, of which two sounded their bell, and 31 didn't.

These data can be analysed with a statistical test called the "Chi-squared test" to calculate the probability that they could have been just the result of mere chance: that the behaviour of drivers was no different from that of cyclists and that our hypothesis was supported. But the probabilities reported by the test are vanishingly small for both sets of data. Cyclists and drivers behave astonishingly differently.

We must conclude from the data that cyclists treat walkers far, far worse than drivers treat cyclists.