Floating Bus Stops: When a Protected Bike Lane Crosses the Bus Door

TL;DR;

• A true bus stop bypass gives passengers an island; a shared-use boarder puts boarding directly across the cycle track.
• London recorded six pedestrian casualties at 164 bypasses in 2020–22, but police data miss disability status and many non-injury events.¹
• Blind and mobility-impaired passengers report fear, wayfinding problems, and avoided journeys even where crashes are rare.²
• England paused only new direct-boarding shared-use boarders in November 2025—not all bypasses.³
• Wide islands, detectable edges, raised crossings, sightlines, slower cycling, and monitoring make the compromise better, not perfect.⁴

Two safety systems meet at one bus stop

A protected bike lane reaches a bus stop. If the lane simply ends, a person cycling must merge into traffic, pass a stopped bus, and then survive the bus pulling out again. If the lane continues beside the curb, passengers step across it to board. If it bends behind an island, passengers wait apart from cyclists—but must cross the track to reach the island.

There is no version in which geometry makes every conflict disappear. The design question is which conflicts remain, how severe they can be, and whether everyone can understand and navigate them independently.

That is why “floating bus stop” is an unhelpfully broad label. It is used for several layouts with materially different accessibility and safety consequences. In January 2026, England’s Department for Transport adopted it as a statutory umbrella term for places where a cycle track and bus stop intersect, while separately defining bus stop bypasses, boarding islands, and shared-use boarders in its floating bus-stop guidance.³

The evidence supports neither “they are death traps” nor “the crash count proves there is no problem.” Proper island-style bypasses have produced few recorded pedestrian injuries and can remove serious bus–bicycle conflicts. But recorded injury is only one outcome. A stop also fails when a blind passenger cannot tell whether a cyclist is coming, when a wheelchair user cannot turn off the ramp, or when fear makes someone abandon the bus trip.

The defensible position is therefore conditional: maintain bicycle protection through busy bus stops, but never make passengers board directly into a live cycle track; provide a real island and a legible crossing where space and demand permit; and treat disabled people’s independent access as a performance requirement, not a public-relations concern.

First, name the design correctly

The common names hide more than they reveal. The following distinctions combine the 2026 English statutory definitions with terminology used in the Living Streets inclusive-design study.³²

The first row is the layout many people picture: the bus stops at the curbside edge of an island, the cycle track runs behind it, and a passenger crosses the track between the island and the main sidewalk. The third row is the layout England paused in late 2025. Conflating them makes the debate almost impossible to resolve.

Living Streets found the British vocabulary so inconsistent that its researchers introduced “continued kerbside track” and “shared platform boarder” to distinguish a visibly uninterrupted track from a raised shared area. After mapping more than 600 British sites, they concluded that many hybrid layouts were suboptimal and only a small proportion appeared to meet their proposed minimum design quality.²

Why not just end the bike lane?

At a conventional bus stop, the bus and bicycle repeatedly trade places. A bus passes a rider and pulls to the curb; the rider leaves the lane to pass the stopped bus; then the bus pulls out and may overtake again. The speed and mass differences make a mistake far more consequential than a low-speed pedestrian–bicycle contact.

The 2026 English guidance states that once protected cycling is judged necessary on a bus route, it should be maintained through bus stops to reduce bicycle–bus conflict.³ Living Streets likewise concluded that the main purpose of continuity is protection from general traffic; avoiding direct interaction with buses is a related benefit, not merely a convenience upgrade.²

This is the same systems problem discussed in Why Your Bike Lane Ends at Every Intersection: a facility can feel safe between conflict points while abandoning its user exactly where the difficult movement occurs. At a bus stop the conflict is mid-block rather than at a corner, but “merge back into traffic here” is still not an all-ages design.

The alternative is not to declare one vulnerable group expendable. It is to separate the bus from the bicycle and give passengers a predictable, accessible place to wait, deploy a wheelchair ramp, and cross with priority.

What the collision and conflict evidence actually shows

Three kinds of evidence answer different questions:

• Police injury records tell us about reported crashes, but not near-misses, fear, or trips never taken.
• Video and trajectory studies reveal yielding and conflict behavior, but usually cover a few sites for short periods.
• Interviews and accompanied journeys reveal exclusion and cognitive burden, but do not estimate population crash rates.

The findings should be read together rather than used to cancel one another out.

London’s small number is reassuring—but incomplete

TfL’s 2024 review found a very low recorded injury count: five pedestrian casualties involving a cyclist and one involving an e-scooter across 164 bypasses over three years. For scale, London recorded more than 11,400 pedestrians injured in collisions with motor-vehicle drivers in the same period, according to the Mayor’s summary of TfL’s analysis.¹

That comparison establishes that motor vehicles remain the vastly larger source of pedestrian injury. It does not prove that a bypass crossing is safer per encounter than a roadway crossing: the numbers lack comparable exposure denominators. Nor does it answer whether the layout transfers a burden onto a small group of passengers.

TfL itself identified the limitations. Its review engaged more than 50 stakeholder groups and reported anxiety, access difficulty, inconsistent layouts, non-yielding cyclists, and likely under-reporting. Around one third of the 164 stops varied significantly from TfL’s best-practice design—through errors such as incorrect tactile paving, no zebra, or an island that was too narrow.¹ The national STATS19 injury database records age and sex but not disability, as the Mayor confirmed in 2024.⁶

TfL also found no overall reduction in older and disabled passengers boarding after bypasses were installed.¹ That is useful evidence against a large network-wide deterrent effect. Yet an aggregate boarding series cannot show whether a blind passenger changed stops, traveled with assistance, or stopped making a particular journey while other Freedom Pass users continued boarding.

Zebras help, but paint cannot guarantee yielding

TRL’s 2018 street observations found that zebra crossings drew more passengers to a predictable crossing point and made cyclists more likely to yield. But yielding increased only from 33% to 40%.⁵ TfL’s later video observations at eight sites similarly found that a significant proportion of cyclists did not yield when an interaction occurred, although the sample was too small for statistical certainty.¹

An earlier TRL off-street trial reached a compatible result: 98% of recorded interactions were minor, and zebra crossings produced the lowest observed interaction rates while shifting more avoidance action from pedestrians to cyclists.⁷ That supports a marked crossing—but also shows why a zebra should be part of a legible, speed-reducing design rather than a stripe expected to do all the work.

Under English law a cyclist must give precedence to a pedestrian on a zebra across a cycle track and must stop at a signal-controlled crossing. The current statutory guidance recommends enforcement and education where compliance is poor.³ But a design for independent mobility cannot depend on every approaching rider remembering a rule.

The crossing must announce itself physically: clear sightlines, tactile paving, a visible material change, a raised table, and geometry that makes an appropriate speed feel natural. Signs are reinforcement, not the safety system.

Why rare crashes can coexist with real exclusion

A sighted passenger can look left and right, estimate a bicycle’s speed, make eye contact, and decide whether the rider is yielding. A blind or partially sighted passenger may not receive any of those cues. Bicycles are quiet, a waiting bus masks sound, and a zebra does not tell a person who cannot see it whether the approaching rider has stopped.

In accompanied research, blind participants described flat, visually and tactually ambiguous layouts as unnerving. Living Streets concluded that busier British cycle tracks could be difficult or impossible for some blind and partially sighted pedestrians to cross, and that conventional zebras made little difference for those users as complexity and volume increased.² Its researchers found some stops that they considered unsafe to ask blind participants to cross without support.

The consequence is not limited to collision. It includes:

• waiting for a stranger to help;
• using a different stop or route;
• taking a short bus journey merely to bypass a confusing street obstacle;
• losing confidence in an otherwise familiar trip;
• avoiding the bus entirely.

The 2024 UCL research commissioned by Guide Dogs found that floating-island stops and shared-use boarders were not perceived as safe by participants, with vision-impaired people reporting psychological stress and avoidance of bus services; the Department for Transport summarizes those findings in its 2026 guidance.⁸

Mobility impairment creates a different set of constraints. A wheelchair ramp needs a level landing, enough length for deployment, and turning room after the chair leaves the bus. A narrow island can be technically present yet unusable once a shelter, pole, bench, waiting crowd, stroller, or second wheelchair occupies it. Excessive crossfall can tip or pull a wheeled mobility aid sideways. Raised speed treatments that are trivial on a two-wheeled bicycle may destabilize a handcycle, tricycle, or mobility scooter.

A 2026 rapid review of 19 studies and reviews found that bus-stop redesigns were consistently associated with reduced perceived accessibility, safety, and confidence among disabled people, while also documenting conflicts within accessibility itself: tactile surfaces can help blind travelers navigate but destabilize some mobility-aid users; kerb-free layouts help wheelchairs but remove orientation cues for cane users.⁹

This is why “nobody was hit” is necessary evidence, but not a complete accessibility audit. Avoidance suppresses the very encounters from which a crash rate would be calculated.

What England paused—and what it did not

On November 20, 2025, Roads and Buses Minister Simon Lightwood asked English local authorities to pause new floating bus-stop layouts where passengers board or alight directly from or into a cycle track. The pause remained in force when the Department for Transport and Active Travel England published statutory guidance on January 26, 2026.³

That is a pause on new shared-use bus boarders, not a nationwide moratorium on every bus stop bypass. The 2026 guidance continues to recommend two separated options in suitable contexts:

1. A bus stop bypass with the shelter and waiting area on a distinct island.
2. A cycle track with a smaller kerbside boarding island, for lower- to medium-flow locations.

The distinction matters. The government’s central design principle is that no passenger should board or alight directly into a cycle track. An existing shared-use boarder may be converted by adding a real boarding island; authorities may also conclude after site assessment that an ordinary bus stop is more suitable where motor traffic, cycling volume, and speed are low.³

The January 2026 document is explicitly a first version. Active Travel England is researching the safety, accessibility, suitability, and possible enhancements of different layouts, with completion due in 2027. Later guidance is expected to incorporate the results, including further work on activated warning signs.³

So, as of August 18, 2026, the policy is best described as restrict, redesign, test, and update—not “ban protected bike lanes at bus stops.”

A design checklist that treats access as measurable

England’s guidance now gives the full bypass a desired island width of 3.5 meters and a minimum of 2.5 meters, along with a 2-meter desired one-way track, a zebra, high material contrast, correct tactile paving, and horizontal and vertical speed deflection. It also requires designers to consider actual and latent walking, cycling, and bus-passenger flows rather than forcing the same template onto every site.³

Those dimensions are starting points. The broader research suggests a more useful test: can each feature solve a specific failure mode?

The 2024 MassDOT/UMass study is especially useful because it tested assumptions close to home. Researchers inventoried 56 floating stops in the MBTA area, conducted focus groups, and used video and LiDAR at five stops in Boston, Brookline, Everett, and Somerville. Participants preferred a full-width platform; partial platforms with the shelter left across the track encouraged passengers to wait until the bus arrived and then rush the crossing.⁴

The study recommended platform width, tactile wayfinding, aligned crossings and boarding areas, visible yield markings, audible information, shelter placement on the platform where feasible, and avoiding two-way or counterflow tracks at floating stops because they complicate detection for visually impaired passengers. It also warned that features often assumed to slow bicycles—curves and fencing—did not substantially reduce speed in its small field sample.⁴

Good infrastructure is not a rendering. It is a hypothesis that must survive observation after opening.

Measure the stop that people experience

Every installation should begin with a baseline and continue with a published after-study. At minimum, agencies should report:

• pedestrian, bicycle, and passenger volumes by time of day;
• boarding and alighting surges, including bus bunching;
• cyclist speeds before and at the crossing;
• yielding and crossing compliance;
• reported crashes, falls, near-misses, and complaints;
• ramp deployment, wheelchair maneuvering, and island crowding;
• accompanied audits with blind, partially sighted, mobility-impaired, neurodivergent, and disabled cycling participants;
• whether people changed stops, routes, travel companions, or modes;
• design dimensions, maintenance failures, and the changes made in response.

This is more demanding than checking a collision database. It is also more honest. A count of passengers says little about the passenger who no longer appears, while a fearful interview cannot establish an injury rate. The evaluation needs both.

Inman Square in Cambridge already shows that protected cycling, floating stops, constrained geometry, and emergency access can be designed together; it appears in our examination of fire trucks and safer streets. But the existence of a compromise does not make every compromise good. A successful detail must work on a dark winter evening, with salt covering the tactile contrast, two buses at the curb, and a wheelchair ramp deployed—not just on opening day.

The answer is not bikes versus disabled people

The political framing is often cruelly simple: either protect cyclists from buses or protect disabled passengers from cyclists. Street space is finite, but that does not make exclusion inevitable.

The evidence supports four conclusions.

First, maintaining protected cycling through a busy bus corridor solves a real and potentially severe safety problem. Sending riders into motor traffic every few blocks is not an acceptable default.

Second, a shared-use boarder that asks someone to step from a bus directly into a cycle track is materially different from a bypass with a substantial island. England was right to pause new direct-boarding designs while evidence and guidance develop.

Third, low recorded crash totals at island bypasses are encouraging, but they do not erase non-yielding, inconsistent construction, fear, or inaccessible navigation. Perceived safety is not imaginary when it changes whether a person can travel independently.

Fourth, design quality changes the answer. A full-width island, detectable boundaries, a level ramp area, predictable crossing, real speed control, sightlines, legal pedestrian priority, and continuous monitoring can turn a confused shared space into a legible sequence of separate spaces. At the busiest or most complex locations, a signal-controlled crossing may be warranted. At a site too narrow to provide accessible minimums, the honest answer may be a different stop location, a different cycle-route alignment, or much lower motor-traffic speed and volume.

A protected bike lane should not end at every bus stop. Neither should a disabled passenger’s independence.

FAQ

Q1. Are floating bus stops banned in England?
A. No. England paused new direct-boarding shared-use boarders in November 2025; bypasses with distinct passenger islands remain an option.³

Q2. How many people have been injured at London’s bus stop bypasses?
A. TfL identified five cyclist-involved pedestrian casualties and one involving an e-scooter at 164 bypasses during 2020–22; important data gaps remain.¹⁶

Q3. Can blind passengers safely use a bus stop bypass?
A. Some do, but inconsistent layouts impede independent travel; detectable boundaries, predictable tactile routes, controlled crossings, and consultation with blind users are essential.²

Q4. Do zebra crossings make cyclists yield at floating bus stops?
A. They help but are insufficient alone: one TRL study measured 33% yielding at uncontrolled crossings and 40% at zebras.⁵

Q5. What is the safest floating bus-stop design?
A. Evidence favors a spacious island, level ramp area, detectable separation, raised direct crossing, low cycling speeds, and good visibility.³⁴

References