When Is a Bicycle No Longer a Bicycle?

The category problem is back

In August 2025, the Dutch government had what seemed like a straightforward regulatory problem: define a fatbike. Ministers wanted to know whether fatbikes, or perhaps a broader class of heavy electric bicycles, could be separated from ordinary e-bikes for additional rules. But the research commissioned by the Ministry of Infrastructure and Water Management came back with an awkward answer: weight and motor torque overlapped too heavily with ordinary e-bikes, cargo bikes, tricycles, and adaptive cycles; torque was difficult to measure reliably at the roadside; mass was not visually obvious; and manufacturers could redesign products around a threshold. More importantly, the research did not identify a defensible safety threshold at which either variable cleanly separated a dangerous fatbike class from legitimate bicycles.

That might have ended the project. Instead, in April 2026 the cabinet returned to the objective of creating a separate fatbike category, now as the basis for a minimum age and a broader national route for municipal fatbike-free zones. The letter acknowledged that a watertight definition remained difficult and commissioned another search for pragmatic criteria. Some municipalities have already used local public-order powers for specific places, but those measures are legally different from a national vehicle category. As of 1 October, no later substantive national definition report supplying the promised new criteria had been located in the official record examined for this article.

The obvious reading is that Dutch traffic law has repeatedly allowed motorized two-wheelers onto bicycle infrastructure, only to discover later that the vehicles had outgrown their old categories. That was also my starting suspicion, but the research changed it. The bromfiets, or Dutch moped, is a real example of a later category-infrastructure mismatch; the modern ordinary pedelec is an important counterexample because Dutch and EU law deliberately construct it around pedal-dependent assistance, limited continuous motor power, and assistance ending at bicycle-like speed. The faster speed-pedelec was later moved into the moped regime.

The fatbike controversy contains at least three different problems that are easy to collapse into one. The first is whether the legal definition of an ordinary pedal-assisted bicycle is still conceptually sensible; the second is whether police and regulators can efficiently tell if the machine in front of them actually complies with that definition, including its software and configuration; the third is whether the injury data can tell us if fully compliant fatbikes are unusually dangerous, or whether a significant share of the harm is coming from modified, throttle-driven, or otherwise non-compliant vehicles.

The evidence gives three different answers: the legal boundary remains reasonably coherent, enforcing it is genuinely difficult, and the casualty data do not yet resolve the compliant-versus-non-compliant split.

That legal line changes the rules a rider and vehicle face. A compliant pedelec follows bicycle road-placement rules and does not acquire the ordinary moped package of type approval, registration plate, AM license, motor-vehicle insurance, and helmet requirements; a speed-pedelec does. Draw the line badly and the state can put very different vehicles into the same traffic stream; draw it well but make it too expensive to enforce and the same thing happens in practice. If injury data cannot distinguish the vehicles on either side, policymakers may respond to the wrong mechanism even after casualties rise.

The central problem, then, is how law keeps bicycle-like operation observable and enforceable as technology and markets change, and how user and infrastructure rules should respond when a legally compliant vehicle still creates different risks in different settings. Product definition, enforcement, and injury evidence are connected, but none can substitute for the others.

The bromfiets: a real mismatch, and a real correction

The safest place to begin the history is not with a tidy post-war origin story, because the primary material inspected for this article does not support one with enough confidence. By the 1990s, however, the mismatch was explicit in the lawmaking record. The assumptions behind older bromfiets rules included machines that could meaningfully be pedaled; the government noted in the explanatory material for the 1999 reform that contemporary mopeds generally no longer fit that premise.

A moped traveling among bicycles is not merely a bicycle with a little extra help. The speed difference changes overtaking, crossing conflicts, and the time other road users have to react, yet putting the same vehicle into motor traffic creates a different problem if cars are traveling much faster than the moped. Road placement is an attempt to choose the less dangerous mismatch, not a moral classification of vehicles into good and bad categories.

The 1999 Bromfiets op de Rijbaan reform moved ordinary urban mopeds from bicycle facilities toward the carriageway while retaining exceptions where the motor-traffic environment made that unattractive. The reform is useful precisely because it was not based on the idea that every powered two-wheeler always belongs with cars. It responded to a class whose real-world interaction with bicycle traffic had changed.

The subsequent 2001 evaluation found that police-registered injury crashes involving mopeds fell by 31 percent on sampled routes affected by the reform between 1999 and 2000, while injury crashes fell by 19 percent on comparison routes where mopeds remained on the cycle path. Using that background trend, the evaluators estimated a roughly 15 percent net effect on the affected routes.

Those numbers require more restraint than they usually receive. This was a one-year before-and-after evaluation in 19 municipalities, not a randomized experiment; other road-safety measures were changing at the same time, police data missed some crashes, compliance was incomplete, and the evaluation had limited travel-exposure data. The report’s separate 16 percent figure referred to the estimated share of the total urban decline attributed to the route effect. It was not a finding that the reform caused a 16 percent nationwide reduction in moped injury crashes.

Even with those limitations, the decline was larger where the placement rule changed, and the evaluators attributed a material portion of that difference to the reform. The lesson is narrower than “motorized vehicles do not belong on cycle paths”: a category can become mismatched with its environment, and moving the line can improve safety when the alternative traffic environment is suitable. Later Dutch experience makes any broader rule much harder to defend.

The pedelec is the counterexample

A lawful ordinary electric bicycle, or pedelec, is not legally just a bicycle to which somebody happened to attach a motor. Under the current Wegenverkeerswet 1994 definition, assistance must depend on the rider pedaling, the motor’s nominal continuous power may not exceed 250 watts, and assistance must progressively reduce and end by 25 km/h. EU Regulation 168/2013 uses essentially the same functional boundary to exclude these pedal-assisted cycles from the harmonized L-category type-approval system that covers mopeds and other more capable powered two-wheelers.

Pedal dependence is easy to overlook because public discussion often compresses the rule into “250 watts and 25 km/h,” but it does much of the conceptual work. The motor is supposed to assist cycling, not replace it with independent mechanical propulsion. A 25 km/h machine with a throttle is not necessarily the same regulatory object as a 25 km/h pedelec, just as a nominal speed number does not by itself tell us how a vehicle accelerates, how it is controlled, or how it interacts with surrounding traffic.

The behavioral evidence is consistent with that distinction. In a Dutch naturalistic study, corrected results put mean urban speed at 17.3 km/h for conventional bicycles, 20.1 km/h for ordinary pedelecs, and 26.9 km/h for speed-pedelecs. The samples were small and self-selected, so these are not universal population speeds. The useful observation is the spacing: ordinary pedelecs in this study were much closer to conventional bicycles than speed-pedelecs were.

Dutch regulation also changed when the faster vehicle stopped fitting the ordinary bicycle treatment. From 2017, speed-pedelecs were brought into the bromfiets regime, with the associated motor-vehicle rules and road placement. That is strong evidence against the broadest version of the “category drift” story: the state did not simply leave every bicycle-shaped powered vehicle on the cycle path forever.

This does not prove that the present pedelec definition is perfect. A 250-watt legal limit says nothing directly about rider judgment, and a 25 km/h assistance cutoff does not mean the vehicle will always be traveling at 25 km/h or that every cycle path is equally suitable for every legal e-bike. Geometry, loading, acceleration, and the local density of pedestrians and cyclists can still change how the vehicle interacts with its surroundings.

But the burden of argument changes. A new category inside the lawful pedelec class needs more than the observation that some e-bikes look more motorcycle-like than others. A workable criterion would have to identify a materially different problem, be observable or testable at tolerable cost, avoid sweeping legitimate cargo and adaptive cycles into the same bucket, and resist trivial redesign by manufacturers or users.

The 2025 weight-and-torque study tried to find such a line and failed. That does not prove weight or torque are irrelevant to safety; it shows that the evidence did not establish a policy-suitable threshold and that the variables created serious overlap, measurement, and gaming problems. Earlier DTV work likewise showed that appearance and geometry can be turned into candidate definitions, but draftability is not the same as safety relevance, low collateral capture, or resistance to redesign.

Fatbikes expose three different problems

The word “fatbike” now does too much work in Dutch debate: it can mean a fully compliant pedelec with wide tires and a bench-style seat, a legal bicycle being ridden aggressively by a young or inexperienced rider, or a machine with a throttle, altered controller, or high-speed assistance that no longer fits the pedelec definition. Those are not interchangeable safety or enforcement problems.

The product boundary

For current Dutch traffic law, a fully compliant fatbike is an ordinary pedelec and follows bicycle rules because it meets the same functional definition. There is no separate operative national legal category called “fatbike” as of 1 October 2026.

The case for separating lawful fatbikes has to be built on more than appearance. The commissioned 2025 research found extensive overlap in mass and torque with other lawful electric cycles, including cargo and adaptive vehicles; it also found that those variables performed poorly as roadside criteria and could be designed around. The current evidence does not establish that tire width, styling, bare vehicle mass, or nominal torque independently explains the observed youth injury burden once speed, age, rider selection, exposure, and modification are considered.

That is an evidentiary limit, not an acquittal of every fatbike design. Mass plausibly changes collision energy; acceleration plausibly changes conflict dynamics; geometry may affect handling. The problem is that plausible mechanisms are not yet a usable legal threshold. If a criterion is intended to determine who may use the bicycle network, “this characteristic probably contributes somehow” is not enough.

The enforcement boundary

A stronger case exists that the formal pedelec line can be difficult to police in practice. Even here, however, the problem is more precise than saying that Dutch law is blind to software or that an officer has to catch a rider traveling above 25 km/h.

The Openbaar Ministerie’s current testing guidance allows technical testing of apparent pedal-assisted bicycles. A 2025 first-instance case in the District Court of the Northern Netherlands involved a V20 fatbike that initially cut assistance at 25 km/h. Officers then changed a display setting from 25 to 50 km/h; a subsequent test showed assistance continuing to around 40 km/h. The court held that the machine did not satisfy the ordinary pedelec definition and treated its technical capability, rather than the rider’s observed road speed, as decisive to the classification issue.

A 2026 Gelderland case supplies the necessary counterweight. Police suspected a Phatfour fatbike of non-compliance, but expert testing later found the vehicle compliant in its tested state, and the court ordered it returned in a seizure-review proceeding. These are fact-specific first-instance decisions with different procedural postures, not a national appellate rule. Together they support a narrower proposition: classification can turn on demonstrated technical configuration, while appearance, suspicion, or the abstract possibility that a model might be modified is not enough.

The law has also moved against designs intended to frustrate testing. Since 1 July 2025, the Regeling voertuigen prohibits a pedal-assisted bicycle from having a provision whose apparent purpose is to make checks of motor power or pedal-assistance function more difficult or to influence them. That does not make every configurable display or controller illegal, nor does it remove the need to prove non-compliance in an individual case. It does show that the legal system has already recognized observability itself as part of the regulatory problem.

The enforcement problem is less about a missing legal concept than about throughput. A precise definition can still be administratively weak if proving the relevant facts requires equipment, staff time, and case-specific technical evidence. Expensive determinations reduce feasible coverage and the expected probability of detection.

Regulatory gaming also occurs at different points. Manufacturers can redesign products around a nominal threshold, while retailers or riders can alter software or hardware after sale. The first tests whether the criterion is durable; the second tests whether non-compliance can be detected and proved. Roadside enforcement cannot cover the whole chain: product design, imports, and retail require upstream supervision, while later modification still requires retailer or roadside enforcement.

From January through May 2026, police issued 1,642 fines under a broad offense covering vehicles used without required road approval, and police told the ministry that a majority involved fatbikes. The dataset itself could not separate vehicle types. That shows a real enforcement workload, not that most Dutch fatbikes are illegal.

The safety boundary

The injury signal is also real, and it should not be minimized merely because the causal decomposition is incomplete.

VeiligheidNL’s national emergency-department analysis estimated that, from 2020 through 2024, e-bike visits involving substantial injury increased sharply. Among 12- to 17-year-old e-bike riders, estimated emergency-department visits for MAIS2+ injuries rose 491 percent, and visits involving brain injury rose 522 percent. Increased riding distance appears to explain only part of the growth, but this is not a full exposure-adjusted causal estimate of crash risk per kilometer.

Fatbike-specific data are thinner. The 13 emergency departments participating in the Letsel Informatie Systeem recorded 301 fatbike-related cases in 2024, more than half involving riders aged 12 to 17. That is a useful measure of clinical burden in the participating hospitals, not a national census of fatbike casualties; without reliable fatbike exposure, it cannot tell us the risk per rider, trip, or kilometer.

The most important missing variable may be compliance itself. The main casualty datasets do not reliably tell us whether the fatbike involved in a crash was a lawful 250-watt pedal-assist bicycle, a modified high-speed machine, a throttle-driven vehicle, or something else outside the ordinary pedelec category. The Netherlands lacks the data needed to answer a question sitting at the center of the political debate: how much of the youth fatbike injury burden would remain if every vehicle actually complied with the existing pedelec rules?

Injury burden, injury severity, and exposure-adjusted risk answer different questions. A fatbike count without an exposure denominator cannot tell us risk per rider, trip, or kilometer, and casualty data cannot identify legal compliance unless that variable is recorded. Rising injury burden can justify policy attention even if risk per kilometer is unchanged; claiming that the lawful fatbike itself is unusually dangerous requires a comparison the current Dutch evidence does not supply.

That uncertainty cuts in both directions: it prevents an enforcement-first account from claiming that illegal modification explains most of the problem, but it also prevents a fatbike-specific account from claiming that the lawful design itself has been shown to be unusually dangerous per kilometer. Neither conclusion follows from the available data.

Crash coding and enforcement coding are part of policy capacity. If the state cannot distinguish lawful from unlawful vehicles in injury data, and cannot identify vehicle type in broad enforcement data, it cannot reliably learn whether a new rule is targeting the mechanism that produces the harm.

Helmets and age are not the same intervention

The youth injury data create understandable pressure for rules aimed directly at young riders, but a helmet requirement and a minimum age solve different problems and impose different costs. Treating them as one generic “youth safety package” obscures the mechanism of each.

Helmets address injury severity after a crash occurs, and the broader evidence that bicycle helmets reduce head injury among crash-involved riders is strong. Dutch experience with snorfiets riders gives a useful real-world comparison: after the national helmet requirement took effect in 2023, SWOV’s 2026 evaluation found that the share with head injury among seriously injured snorfiets riders fell from 55 percent in its selected pre-period to 32 percent in 2023-2024.

That is not a clean forecast for teenagers on e-bikes. The SWOV comparison is observational; other changes occurred, the post-policy period is short, and former snorfiets riders shifted among several travel modes. It demonstrates why “helmets reduce head injury in a crash” and “a helmet mandate has a known net social effect” are separate propositions. A mandate changes not only injury protection among people who continue riding, but potentially which trips occur and by which mode.

The cabinet’s proposed under-18 helmet rule is broader than an e-bike-only measure. The draft instrument covers under-18 use of pedal-assisted bicycles and relevant light electric vehicles (LEVs), and the consultation draft also addresses passengers within its covered scope. It was not in force as of 1 October 2026. Consultation materials identify 1 September 2027 as the expected start for the helmet element, subject to the remaining legislative procedure and a later commencement decision. A consultation summary published on 1 October records responses to the proposal; it is not enactment, commencement, or a government endorsement of those responses.

The evidentiary foundation for that proposal is asymmetric: crash-conditional helmet protection is well supported, while the Dutch population effect of the proposed mandate is not known in advance. Some teenagers may wear a helmet and keep riding; some may switch to an ordinary bicycle; some trips may move to a car, public transport, or another powered vehicle; and some may disappear. Those responses affect exercise, independence, household travel, and total road risk, and current evidence does not establish their proportions.

Age restrictions operate through a different mechanism. A minimum age removes motor assistance from younger riders rather than adding protective equipment to trips that continue. The youth injury burden supplies a reason to examine such a rule, but it does not reveal a scientifically correct birthday at which e-bike risk suddenly changes. Choosing 12, 14, 16, or another threshold necessarily combines evidence with judgments about acceptable risk, independent mobility, enforceability, and the alternatives available to affected riders.

Age 14 nevertheless has a distinctive administrative feature in Dutch law. The general duty under the Wet op de identificatieplicht to show valid identification on a lawful demand begins at 14. Police may demand identification only when there is a lawful basis and it is reasonably necessary for their task; the identification duty is not a free-standing power to check anyone at will. More importantly, the WID does not itself create a future traffic rule. A new minimum-age regime would still need a legally clear way to establish age and specify what happens when age cannot be established.

That distinction avoids an attractive but false shortcut. “People aged 14 and older must show ID on lawful demand” does not mean “anyone who cannot show ID is automatically forbidden to ride an e-bike.” Those are different legal propositions. If lawmakers wanted age 14 to function as an easily administered riding boundary, the new rule would need to make the proof-of-age and enforcement mechanism explicit rather than borrowing a consequence that current identification law does not contain.

Age 14 has an administrative attraction: it aligns with an existing identification framework, whereas 12- and 13-year-olds have no general ID duty. Moped rules offer a different model, with a minimum age of 16 and an AM license that itself proves eligibility; ordinary pedelecs have no equivalent credential. None of this makes 14 a biological safety threshold.

The incidence is also straightforward. A higher minimum age excludes more young riders from motor assistance; a lower minimum excludes fewer. Moving the threshold to 14 would impose a mobility cost on children under 14 who currently use lawful e-bikes, including some for whom distance, disability, terrain, or family circumstances make assistance genuinely useful. Any age rule has to count those losses rather than treating every displaced e-bike trip as a successful return to an ordinary bicycle.

The same counterfactual discipline applies to physical activity. Electrically assisted cycling requires less acute exertion than conventional cycling, but that has one implication if a teenager’s alternative is an ordinary bicycle and another if the alternative is a parent-driven car, public transport, or no independent trip at all. Dutch longitudinal evidence shows that e-bikes can substitute for both conventional bicycle and car travel in broader populations, while the youth-specific evidence remains incomplete. That record cannot support a blanket claim that adolescent e-bike adoption reduces total physical activity.

The upper edge of the proposed helmet rule raises a separate question. Helmet protection does not suddenly change at age 18, which is a legal boundary rather than a demonstrated biological threshold. A rule covering all riders in a defined class would also be more visually observable than one requiring an officer to distinguish a 17-year-old from an 18-year-old, although that is an administrative inference rather than a measured enforcement effect. Any extension would need to avoid duplicating existing helmet requirements for bromfietsen, snorfietsen, and speed-pedelecs, and would widen the compliance, enforcement, distributional, and mobility costs. Consultation objections show stakeholder resistance, not representative public opinion.

The helmet and age debates remain analytically separate. Evidence supports crash-conditional helmet protection, substantial youth injury burden, no biologically optimal minimum age, and an administrative connection between age 14 and the ID regime. Those facts do not uniquely determine how to weigh injury protection, mobility, simplicity, enforceability, and behavioral response.

What should the ordinary fietspad be for?

The strongest counterargument to an enforcement-centered account is simple: a coherent product definition is not necessarily a complete infrastructure rule.

A fully legal pedelec can still be relatively fast on a crowded urban cycle path. Rider behavior, acceleration, loading, and local geometry can change how it interacts with surrounding users. More importantly, moving a vehicle away from cyclists does not make risk disappear; it moves the rider into another traffic stream.

The speed-pedelec evidence makes this visible. In one naturalistic Dutch study after reclassification, riders averaged about 28.5 km/h on bicycle paths but only about 31.8 km/h on 50 km/h roadways. Another naturalistic study found more traffic conflicts when speed-pedelecs used bicycle facilities, while the authors warned that an actual crash with a motor vehicle on the roadway could have more severe consequences.

Speed compatibility is relative to the traffic environment: a vehicle can be too fast for one stream and too slow for another. The relevant comparison is the risk imposed on people using the cycle path against the risk faced by the powered-vehicle rider and others in the alternative environment.

A single national product label cannot do all the work. Product classification answers whether the machine is a bicycle or motor vehicle for regulatory purposes; user rules answer questions such as helmets and age; and road-placement rules decide which traffic stream presents the better fit. Enforcement determines whether the formal categories have practical force, while data systems determine whether the government can tell afterward which mechanism is failing.

The current evidence gives relatively strong guidance about some characteristics and weak guidance about others. Pedal dependence and assisted-speed limitation are meaningful because they are part of a deliberate bicycle-like functional line, and wider actual-speed differences create more overtaking and less predictable interaction with surrounding users. A rule that cannot be checked cheaply and consistently may exist more convincingly on paper than on the street. Weight and torque, by contrast, currently perform poorly as standalone fatbike discriminators because they overlap with legitimate vehicles, are difficult to measure or define cleanly, and invite redesign around the threshold.

Accessibility belongs inside the same analysis. A heavy cargo bike may be heavy because it carries children or goods instead of making a car trip; an adaptive cycle or tricycle may have unusual mass or geometry because those features make cycling possible for someone who otherwise could not use a conventional bicycle. That does not exempt such vehicles from every conceivable safety rule. It means that a crude physical threshold can capture precisely the vehicles whose unusual design serves a legitimate transport function rather than the risk the rule was intended to address.

A practical test for the fietspad is contextual: compare the vehicle’s interaction with bicycle traffic against the traffic environment it would enter instead. A lawful pedelec can remain inside the bicycle regime while particular road environments still raise a separate placement question. Any local or category-specific placement rule then has to solve target definition, route legibility, compliance, and alternative-road risk rather than assuming that removal from a cycle path ends the safety problem.

What the evidence allows now

The research does not produce one magic rule, but it does narrow the claims that can be defended now.

The ordinary pedelec definition remains a coherent product line built around pedal-dependent assistance, limited continuous power, and an assisted-speed cutoff. No separate fatbike category based on weight, torque, or appearance has yet been shown to outperform that line on the dimensions regulation would need: relevance to safety or interaction risk, observability, reasonable enforcement cost, resistance to trivial redesign, and limited spillover onto cargo and adaptive cycles. That is evidence that the proposed category is not yet technically established, not proof that no workable category could ever be designed.

The existing legal line leaves an implementation route that does not depend on redefining every lawful fatbike. Product conformity, market surveillance, customs, and seller enforcement can address problems before sale; post-sale modification still requires retailer or roadside detection. The evidence does not establish that one point is always cheapest or most effective, only that upstream controls and roadside testing solve different problems.

User rules can be assessed separately from product classification. A general minimum age does not require a fatbike definition if it applies to a broader, legally coherent universe of currently unlicensed motor-assisted or light motorized two-wheelers. Age 14 would align with the existing general identification threshold, but only if a new rule supplies its own workable proof-of-age and enforcement consequences. A higher threshold would exclude more young riders; a lower threshold would preserve more mobility. The evidence does not identify the correct point on that trade-off.

The proposed under-18 helmet rule rests on another kind of evidence. Helmet protection against head injury in a crash is well supported, while the behavioral and population-wide consequences of a mandate remain uncertain. Proceeding with the proposal would place greater weight on known crash-conditional protection than on the unresolved risk of substitution, suppressed trips, and lost independent mobility. Extending the same principle beyond age 18 would remove an age discontinuity that has no demonstrated biological basis, and it could create a more visually observable rule. It would also expand the compliance, distributional, and mobility costs. Neither choice can honestly be presented as the mechanical output of a study.

Any intervention also needs an evaluation design. Casualty and enforcement records should distinguish vehicle type and, where feasible, compliance status; exposure data are needed to separate rising injury counts from rising risk. Helmet rules additionally need compliance, head-injury, and travel-behavior measures, while age rules need evidence on displaced trips, enforcement encounters, and who loses access to motor assistance.

Failure criteria can be stated in advance. A vehicle category is weak if trivial redesign evades it or it sweeps cargo and adaptive cycles into an unrelated rule; enforcement is weak if testing costs prevent meaningful coverage. Helmet or age rules warrant reassessment if compliance stays low or mode substitution creates large offsetting harms. Persistent inability to distinguish compliant from non-compliant vehicles would still prevent learning why a reform succeeded or failed.

Road placement remains the least suitable problem for a single national slogan. The historical bromfiets reform and modern speed-pedelec evidence both show why moving a powered two-wheeler away from a cycle path cannot be evaluated without the alternative traffic environment. Speed differences, local geometry, motor-traffic speed, route consistency, and compliance all affect the comparison. The present research supports a two-sided risk test, but not one scientifically derived national speed or product cutoff that answers every local case.

So when is a bicycle no longer a bicycle? Legally, the ordinary pedelec answer is more coherent than the public argument often suggests: pedal assistance must remain auxiliary, stay within the defined power limit, and end by bicycle-like speed. Once a machine no longer satisfies that functional boundary, calling it a fatbike or e-bike does not make it a bicycle.

The broader policy answer is less tidy. The evidence does not establish that a visually recognizable fatbike category, or a technology-neutral-sounding weight or torque threshold, is a workable shortcut around the existing legal line. It does establish real enforcement and information problems, and it shows that product rules, age limits, helmets, and road placement operate through different mechanisms and impose different costs. Those findings narrow the choices considerably without turning the remaining value judgments into scientific conclusions.

The Netherlands has allowed vehicle categories to drift out of alignment with its infrastructure before, and it has also corrected them. The useful lesson from the fatbike controversy is not that every new machine needs a new label; it is that a durable system has to keep a functional bicycle category enforceable, add user rules only with clear mechanisms and costs, notice when actual traffic interaction no longer fits the infrastructure, and collect enough information to tell when a rule is failing.