Bike lanes fail at junctions, which is where the collisions are

There is a settled way of talking about cycle infrastructure. It is worth asking how much of it survives contact with the detail.

The link is the easy part

Building a protected lane along a straight section of road is comparatively simple and is what most schemes deliver. The difficulty is what happens where the lane meets a side road, a turning movement or another cycle route. Collision data across many cities consistently locates most cycle injuries at junctions rather than between them.

A network that protects links and abandons junctions therefore protects the safest part of the journey.

Turning conflicts are the specific danger

A driver turning across a cycle lane is the classic conflict, and it is made worse when the cyclist is out of the driver's primary field of view. Right-hook and left-hook collisions, depending on the country, dominate serious injury statistics.

Over a decade, design responses either separate the movements in time through signalling or in space through geometry. Vehicle design belongs to the same problem, since a high cab with a long bonnet creates a blind area directly ahead and to the near side, which is why heavy goods vehicles appear in serious cycle injury figures far out of proportion to how many of them are on the road.

Protected intersections change the geometry

The Dutch protected intersection uses corner islands to set the crossing point back from the carriageway and to force a tighter turn. That gives a turning driver a near-perpendicular view of approaching cyclists and slows the turn.

Across a network, it requires space at the corner, which is the main obstacle to retrofitting it in older street layouts. The corner radius does most of the work, because a tight radius physically forces a driver to slow in order to complete the turn, and the generous radii found on so many junctions were specified to let large vehicles turn without stopping — a convenience that reappears later in the injury statistics.

Signal separation is the alternative

Giving cyclists their own signal phase removes the conflict entirely at the cost of capacity for other movements. This is often the only option where corner space does not exist, and it is a straightforward trade of throughput for safety. Cyclist compliance with signals improves substantially where the phase is short and frequent rather than long and rare.

Detection is where it fails in practice, since loops calibrated for cars frequently do not register a bicycle at all, and a cyclist who waits through a phase that never arrives learns within a week to stop trusting the signal.

Networks matter more than individual schemes

A protected route that ends abruptly delivers users into the conditions the route existed to avoid. Cycling rates respond to network completeness rather than to total kilometres built, which is why disconnected schemes underperform.

Across a network, the measure worth tracking is the proportion of trips that can be made entirely on protected infrastructure. A route is only as usable as its worst hundred metres, which is why the sensible unit of assessment is a whole plausible journey — home to school, home to station — ridden end to end, rather than a scheme boundary drawn wherever the budget happened to stop.

What works in a dense grid may not transfer to a low-density suburb.

The collision record understates the problem

Injury statistics count people who were cycling, and the clearest effect of a hostile junction is that most people never cycle through it in the first place. A junction with no recorded cycle collisions can therefore be among the worst in a network by the measure that matters, which is whether an ordinary person would ride through it at all. Near-miss reporting, video conflict studies and stated-preference surveys exist precisely because the collision record is a count of survivors rather than a measure of risk.

Across a network, rates also need an exposure denominator, and a city that does not count cycle flows cannot distinguish a genuine safety improvement from a collapse in the number of people cycling.

The takeaway

Look at what the scheme does at the junction. That is the whole safety case.

The design decision is visible long after the people who made it have gone.