Systematic Review

What Actually Changes Passenger Behaviour: A Systematic Review of Behavioural Mechanisms and Seatbelt Use on Buses

Evidence on fifteen ways to change behaviour, drawn from thirteen fields and assessed for practical use on buses

17 August 2026 · Prepared by AITSC · Commercial disclosure follows the conclusion

1Introduction

1.1Background

Seatbelts are highly effective safety devices, yet reported use among Australian bus passengers is low.

In the 2023 Greta coach rollover, 6 of 35 passengers were belted. Ten people died. Every one of those deaths involved the person being thrown from the vehicle. No belted passenger was thrown from their seat.¹

This review considers ways to increase seatbelt use beyond better messaging. Research across several fields has measured the effects of different approaches to changing behaviour. We assess how that evidence could apply to bus passengers.

1.2Objectives

The review answers three questions, in order.

1. Across every field where it has been measured, which mechanisms change human behaviour, and by how much?
2. Which could work under normal bus operating conditions?
3. Which practical approaches should guide system design, and could they improve on existing systems?

The first question brings together published findings. The second assesses their practical application on buses. That assessment informs the design recommendations in the third question.

1.3Scope

Included. Studies of seatbelt use by drivers and passengers in cars, taxis, rideshare vehicles, buses and coaches. Behavioural evidence from aviation, occupational safety, healthcare, drowning prevention and public health, where a study isolates a single mechanism. Meta-analyses of behavioural interventions. The rules governing seatbelt reminders in Australia, Europe and the United States.

Excluded. Crashworthiness engineering, seat and anchorage design, glazing, and the biomechanics of ejection. These mattered a great deal at Greta. They are not behavioural questions. Also excluded are autonomous vehicles and child restraints below the adult seatbelt threshold, both of which have their own separate literatures.

2Methods

2.1How the literature was searched

The literature search had two stages.

The first stage covered a broad range of research. Thirty searches covered thirteen fields: behavioural economics and nudge theory, applied behaviour analysis, safety engineering, occupational safety, road safety, criminology and deterrence, aviation cabin safety, clinical alarm management, infection control, drowning prevention, public health and public transport compliance, environmental and consumer behaviour, and education. The aim was to find every mechanism with a measured effect, not to find seatbelt studies.

In the second stage, for each mechanism identified, we searched for the strongest available effect estimate, preferring meta-analyses and randomised field trials over single studies, and for any study applying that mechanism to seatbelt use.

Two primary sources were retrieved in full and read directly: the Swedish bus observation study⁴ and the Greta investigation report.¹ Others were read through abstracts, published summaries and secondary citation.

2.2What was included and what was not

A study was included if it met three criteria. It reported a behavioural outcome, not an attitude or an intention. It isolated an identifiable mechanism. It reported a quantity, whether an effect size, a rate or a change in a rate.

Where a mechanism had both an original finding and a later replication or bias correction, we report the corrected figure. Re-analysis has substantially reduced several well-known estimates of behavioural effects, and section 5.2 sets out which ones.

2.3How to read the numbers in this review

2.4How the mechanisms were compared

The effect estimates in this review come from different study designs, populations and outcome measures. They cannot be ranked precisely against each other. Table 2 does not support a single combined effect estimate.

The evidence supports a general pattern: changing the effort or consequences of an action tends to produce larger effects than information alone. This conclusion draws on findings across several fields rather than on a single effect estimate.

2.5How practical use on buses was assessed

Section 6 assesses five bus operating conditions that differ from the settings of many of the studies. Each mechanism in Table 2 was assessed against those conditions and classified as practical, impractical, or requiring modification. Table 3 records the reasons. These classifications are the review’s assessment, not findings from bus trials.

3Results: the size of the problem

3.1Belt use falls the further you sit from the driver

Seatbelt studies consistently report lower use in seats farther from the driver. Compliance drops as the occupant moves away from the driving position. It drops again when the vehicle belongs to somebody else.

A systematic review and meta-analysis of 68 studies found pooled belt use of 43.94 per cent among drivers, 38.47 per cent among front seat passengers and 15.32 per cent among rear seat passengers.² These are pooled global figures across countries with very different laws. They show the difference between seating positions, but do not establish wearing rates for any particular country.

Lower rear-seat use is also reported in taxis and rideshare vehicles. Observed rear seat belt use was 26 per cent in San Francisco taxis and 28 per cent in Las Vegas taxis. In rideshare vehicles it was 52 per cent and 18 per cent respectively.³ Front seat passengers in the same cities were belted at 89 to 100 per cent.

3.2Buses

The best bus dataset available is Swedish. Anund, Forward and Sjörs Dahlman combined 328 bus observations across ten cities with seven focus groups and a survey of 1,737 respondents.⁴

Two points are important when interpreting these rates.

First, announcements and belt use rise and fall together. Charter services announce the most and have the highest belt use. Regional services announce the least and have the lowest. The authors report this association directly. It is also confounded, because charter and regional services differ in vehicle type, road speed, trip length and who is on board. The pattern is consistent with reminders helping, but does not establish that reminders caused the difference.

Second, what passengers say about their own belt use is unreliable. Charter passengers reported 44 per cent when observation showed 92 per cent. Regional passengers reported 31 per cent when observation showed 27 per cent. The error runs in both directions. Survey estimates should not be treated as direct observations of belt use.

3.3Australia

Reported Australian bus passenger wearing rates range from 7 to 45 per cent. Car occupant wearing rates are about 97 per cent. At Greta, 17 per cent of passengers were belted.¹

Reported enforcement is too infrequent to provide a strong deterrent. One passenger fine was issued across the whole of New South Wales in the three years to 30 June 2025.¹ Section 6 explains why deterrence is unlikely to be effective when the chance of being fined is so low.

4Results: why passengers do not buckle

4.1Passengers report problems with belt condition, fit and convenience

There is a common assumption that not buckling reflects a considered view about risk. The evidence does not support it.

In the Swedish focus groups and survey, the reasons given were: the belt was dirty, the belt would not pull out or would not latch, the belt was too short or too long to fit, they wanted to sleep, and on short trips they were worried about not being ready to get off in time.⁴ One participant described a belt that had left dirt on their clothes. About 18 per cent of respondents across all service types said the belt was simply uncomfortable.

The same pattern shows up in recreational boating, where the barriers to wearing a lifejacket are discomfort and a worry that wearing one signals inexperience.⁵ Discomfort and inconvenience can discourage use even when people know the safety benefits.

4.2Immediate discomfort can outweigh concern about a possible crash

Compare the two options as a passenger experiences them.

Not buckling has a consequence that is delayed, very unlikely, and severe. Buckling has a consequence that is immediate, certain, and trivial: a few seconds of effort and some discomfort.

Present bias means giving more weight to immediate effects than to possible future effects. It can explain non-use without assuming that passengers misunderstand crash risk. Workers who skip protective equipment face a similar choice, choosing the certain immediate comfort over the remote possibility of injury.⁶

This suggests an intervention should give passengers an immediate reason to buckle up, as well as explaining the safety benefit.

4.3Bus seats may not trigger the habit of buckling up

For habitual car users, the familiar seat and belt can trigger buckling up automatically, without a conscious decision each time.⁷

A bus may have a different seat, sitting position and belt arrangement. Those differences can mean that a familiar car habit does not transfer automatically.

A regular car seatbelt user may therefore forget to buckle up on a coach. This is a possible explanation for non-use even among passengers who value their safety.

4.4Passengers behave differently from drivers

Belt use is correlated between people travelling in the same vehicle, and the driver’s own behaviour predicts the passenger’s.⁸

On a coach, most passengers cannot see the driver’s belt use. The driver cannot see most of the cabin. Swedish drivers said this directly. Even when they check that everyone is buckled before departure, they have no way of knowing what happens after that.⁴

5Results: fifteen behavioural mechanisms

5.1The fifteen mechanisms

5.2What the studies suggest

Three things stand out.

The approaches differ in whether they change an action’s consequences or rely on information. Rows 1 to 4 all work by changing what happens to the person, straight away, whether or not they are persuaded. Rows 10 to 15 all work by trying to change what the person believes, notices or intends. Section 8 describes similar findings in other fields, while section 2.4 explains why the effect estimates cannot be compared directly.

Re-analysis has reduced several reported effect estimates. The headline nudge effect of d = 0.43 is reduced after adjustment for publication bias. Two separate re-analyses published in the same journal estimate the corrected effect at between zero and d = 0.08.^25,26 The watching-eyes effect has largely failed to replicate. The organ donation analysis reports a reduction in living donors under an opt-out policy. Conclusions based on these mechanisms should account for the later, smaller estimates.

Reducing effort produced a large standardised effect in the field-trial meta-analysis. In the biggest meta-analysis of behavioural field trials we found, interventions that changed how much effort something took produced d = 0.52. Interventions aimed at attention produced 0.20. Interventions aimed at memory produced 0.20.¹¹ Convenience on its own reached d = 1.18. In this analysis of more than two million participants, reducing effort was more effective than directing attention or helping people remember.

5.3Why immediate consequences may have stronger effects

Three areas of research offer explanations for this pattern.

Timing, likelihood and severity of consequences. Applied behaviour analysis holds that a consequence controls behaviour according to three properties: how soon it arrives, how certain it is, and how big it is. The first two matter most. A small, immediate and certain consequence can influence behaviour more strongly than a large but delayed and unlikely one. Industrial safety has used this framing for over forty years, and it is the standard explanation for why trained, well-informed people still take shortcuts.³⁰

The hierarchy of controls. Safety engineering ranks controls by how directly they reduce a hazard. The hierarchy starts with removing the hazard (elimination), then replacing it with a safer alternative (substitution), changing equipment or physical conditions (engineering controls), using rules and procedures (administrative controls), and providing personal protective equipment. The related Safe Systems Pyramid applies this principle to road safety: an intervention gets more effective as the effort it demands from the individual goes down.³¹ Education sits at the very top of that pyramid, which is the position of least effect and most individual effort.

A sign, a briefing and an announcement are all administrative controls. So is a law nobody enforces. The measures used to encourage bus passenger belt use rely on administrative controls, which rank below engineering controls in this hierarchy.

Control devices and warning devices. Shigeo Shingo’s distinction from the Toyota Production System explains the difference. A control device stops the process until the error is fixed. A warning device notices the error and leaves it to the operator to act. Warning devices are cheaper, more acceptable, and weaker.³²

Signs and timed announcements provide reminders without detecting whether belts are fastened. They cannot respond to an individual passenger’s belt use.

6Synthesis: which approaches are practical on buses

Section 5 reviews evidence from several settings. The following five bus operating conditions limit how some approaches can be applied.

6.1Five conditions

1. Passengers may be anonymous and travel only once. Without passenger accounts or repeat journeys, operators cannot readily track individual progress, award repeat-use incentives or provide later feedback.
2. Drivers cannot continuously supervise the passenger area. The driver is driving. Swedish drivers said plainly that they cannot monitor belt use once the bus is moving, and asked for technical help to do it.⁴
3. Delaying the whole vehicle is impractical. A coach carrying 57 people cannot be immobilised because one of them is unbuckled. The concern is the effect on safe and practical operation, rather than whether an interlock can be built.
4. Passengers may unbuckle during the trip. Passengers unbuckle to sleep, to reach a bag, to move.⁴ Being buckled at departure is not the same as being buckled at the moment of a rollover.
5. Almost nobody is ever fined. One passenger fine in three years, across an entire Australian state.¹

6.2Assessing each approach against bus operating conditions

6.3Design requirements for alerts linked to belt use

These requirements describe sensors at each seat linked to an alert at that seat. The review identifies this as a promising approach under the operating conditions considered.

6.4Passengers may bypass more restrictive systems

Evidence on bypassing systems provides another reason to consider less restrictive designs.

In the Kidd and Singer trial, part-time belt users were observed defeating the systems in their vehicles. 44 per cent got around the speed-limiting interlock. 21 per cent got around a persistent reminder.¹³ The two produced gains in belt use that were statistically indistinguishable. Bypassing was about twice as common with the interlock, without a statistically detectable additional improvement in belt use.

The interlock literature shows the same thing on a longer timescale. Alcohol interlocks work while they are fitted, and no lasting benefit was found after removal.⁹ These findings do not establish lasting behaviour change after removal. Effects measured while a device is installed should be distinguished from behaviour after it is removed.

7Results: evidence for alerts linked to buckle sensors

7.1Cars

Persistent alerts tied to the buckle have been tested repeatedly in cars, and they work.

Kidd and Singer gave 49 part-time belt users two vehicles for a week each. Compared with an intermittent reminder, belt use rose 30 and 34 per cent, and the persistent reminder performed on a par with a speed-limiting interlock.¹³ A survey of 1,218 drivers and passengers found that people who do not always buckle up said they were more likely to respond to a sound or a vibration than to a light.³³

Reducing the effort needed to comply, or allowing compliance to avoid a delay, can also increase belt use. Van Houten and colleagues fitted 101 commercial vehicles with a device that stopped the driver shifting out of park for up to eight seconds unless the belt was fastened. Belt use rose from 48 to 67 per cent in the United States and from 54 to 74 per cent in Canada. Once fastened, the belt tended to stay fastened for the whole trip.¹²

The important detail is that the driver could avoid the delay entirely by buckling first. Both designs give the driver an immediate reason to buckle up: avoiding a delay or stopping an alert.

7.2The one bus reminder trial located

We located one trial of a per-seat reminder on a bus. KATRI fitted an interurban bus with an occupancy sensor and a buckle sensor at every passenger seat, a warning light at each seat, and a monitoring display at the driver’s position.

Observed belt use was 9.6 per cent without the system and 59.0 per cent with it.¹⁴

Three features of that result deserve attention.

• The baseline was measured in bus passengers. 9.6 per cent is a bus baseline, not a car baseline. The intervention was tested on the population that actually has the problem.
• The warning was a light only. Section 6.3 does not specify sound, vibration or light. The bus trial used light, which surveyed car occupants rated as less likely to prompt a response than sound or vibration.³³
• Occupancy was sensed separately from the buckle. Section 8.3 explains why separate occupancy sensing is needed to avoid alerts at empty seats.

7.3What has never been tested

We found no published trial of a per-seat buckle-sensed audible alert on a bus. The entire bus evidence base is one trial, using a light. This limited bus evidence is the main limitation discussed in section 13.

Separately, the Swedish authors reached the same design conclusion from their own data, and said so plainly:

“Ideally, the announcements could be replaced by seat belt reminders as used in passenger cars.”

Anund, Forward & Sjörs Dahlman (2023), recommendations⁴

The researchers recommend replacing announcements with reminders that respond to belt use.

8Results: the same pattern in other industries

Each case below was chosen because it isolates one mechanism cleanly.

8.1Drink driving: interlocks work while installed

The Cochrane review of alcohol ignition interlocks found the programmes effective while the device was installed, and found no lasting benefit after removal.⁹ The Community Guide review reported re-arrest reductions of 15 to 69 per cent across the five studies showing a significant effect.¹⁰

The benefit depended on keeping the device installed. Preventing an action through an interlock does not establish that people will continue to comply after the device is removed.

8.2Aviation: limits of briefings and announcements

Aviation is the most useful comparison available. Commercial aviation uses briefings, signs and crew supervision, yet some passengers still do not wear belts.

Aircraft have a mandatory briefing, a lit sign at every row, repeated announcements, uniformed crew walking the aisle, and legal powers to detain and prosecute. Even so, more than 90 per cent of turbulence injuries involve people who were not wearing a belt.³⁴ About 37 per cent of surveyed passengers say they did not watch the pre-flight briefing video at all, and recall of the key safety messages averages around half.³⁵

A relevant finding comes from Cranfield’s evacuation research programme for the UK Civil Aviation Authority. Comparing crew behaviour across full-scale evacuation trials, evacuation rates with non-assertive cabin crew were about the same as with no cabin crew present at all. Assertive crew produced the fastest evacuations.³⁶

8.3Hospitals: why sensing the seat is occupied is not optional

Hospital research has measured how staff respond when frequent alarms do not indicate a problem requiring action.

Between 72 and 99 per cent of physiological monitor alarms are technically false or clinically irrelevant. Staff become less responsive to the alarms, a process called desensitisation.³⁷ Staff delay their responses, dismiss alarms, and adjust the thresholds to make them stop.

This provides a reason to include occupancy sensing on buses. Repeated alerts at empty seats may lead drivers and passengers to ignore valid alerts as well. Over a coach’s service life, which may approach 26 years, limiting false alerts is important for maintaining trust in the system.

8.4Hand hygiene: the warning about habituation

Healthcare studies provide evidence on reminders used routinely over time. Electronic hand hygiene monitoring with instant prompts, including audible ones, does improve compliance. The systematic reviews question whether those benefits continue indefinitely:

“For most automated systems, the visual or auditory cue was an integral part of the system and there appeared to be a naive assumption that simple reminders would be sufficient to boost hand hygiene compliance indefinitely.”

Systematic review of electronic hand hygiene monitoring³⁸

This concern also applies to the seatbelt reminders recommended in this review. It is addressed in section 10.2 and again in section 13.

8.5Boating: reported use can exceed observed use

A cluster randomised trial of peer-led training among boaters on Lake Albert, Uganda, measured lifejacket wear two ways.

Self-reported wear in the intervention group rose from 30.8 to 65.1 per cent. Observed wear rose from 1.0 to 26.8 per cent. In the control group, self-report went from 29.9 to 43.2 per cent while observation went from 0.6 to 8.8 per cent.⁵

Self-report overstated the outcome by more than double, in both groups. This matters directly for bus policy, because the Australian evidence base is thin and several figures come from surveys and crash samples rather than systematic roadside observation.

8.6Public transport during COVID-19

Mask mandates on public transport were the closest recent test of announcements plus a legal requirement, in exactly the setting of interest.

Announcements and regulations did raise correct mask wearing. Observed compliance still ran well below self-report. In one European study, 41.6 per cent of passengers wore masks correctly, against 74.7 per cent of transport workers.³⁹ A rule, plus signs, plus repeated announcements, plus a live public health emergency, produced correct compliance in a minority of passengers.

9Results: what the rules require

9.1Cars are covered. Bus passengers are not.

9.2What ADR 68/01 actually does

ADR 68/01 requires three things: a label on the back of each seat, an illuminated sign at the front of the cabin lit for at least 60 seconds after the last door closes, and one recorded announcement before the vehicle travels 500 metres or exceeds 25 km/h.

Put the two systems side by side.

A car chimes because it has detected that your specific belt is undone. It stops when you fasten it. The warning responds to your unfastened belt.

A bus under ADR 68/01 plays a recording on a timer. It is identical whether every belt is fastened or none are. It broadcasts the same message to everyone in the cabin.

9.3This is not only an Australian gap

UN Regulation No. 16 has required reminders on all front and rear seats of cars since 2021, and on the front seats only of M2 and M3 buses.^4,40 The rear passenger area of a European coach has no sensor-based reminder requirement either.

The gap in sensor-based reminder requirements also exists outside Australia.

9.4The cost-benefit case for bus reminders needs assessment

Driver seatbelt reminders were made mandatory in light vehicles on the finding that the safety benefit over a vehicle’s life outweighs the cost of fitting them. That reasoning does not depend on the type of vehicle, and the following factors support examining the case for buses.

1. More room to improve. Reminders were mandated for car drivers already at about 97 per cent, and withheld from bus passengers at 7 to 45 per cent.
2. Lower cost per protected person. One car fitment protects one seat. One coach fitment protects up to 57, over a service life approaching 26 years.
3. A coach crash can expose many unbelted passengers to harm. One rollover produced 29 unbelted passengers, 14 people fully thrown from the vehicle, 3 partly thrown, and 10 deaths.¹

We found no published analysis rebutting that reasoning for heavy vehicles. It has simply not been applied to them.

10Discussion: what should be built

10.1Five principles that follow from the evidence

1. Detect whether each seat’s belt is fastened. The system needs each seat’s buckle state to alert the passenger whose belt is unfastened.
2. Only alert on seats that are occupied. Hospital studies reporting false or irrelevant alarm rates of 72 to 99 per cent show why avoiding unnecessary alerts matters.³⁷
3. Stop the alert when the belt is fastened. In rows 3 and 4 of Table 2, compliance ends the delay or alert. A one-off prompt in row 11 does not respond to compliance.
4. Monitor belt use throughout the trip. A passenger who unbuckles after departure loses the belt’s protection if a crash occurs later.
5. Reduce the effort of buckling at the same time. Row 2 is the highest-effect category in Table 2, and routine maintenance can address it directly.

10.2Two proposed designs without published bus trials

Make a seat amenity available only when the belt is fastened. Condition 3 makes a vehicle interlock impractical. Restricting an individual seat amenity would avoid delaying the whole bus. Two possible designs are:

• Lock the recline until the belt is fastened. Bus drivers suggested this unprompted in the Swedish focus groups, asking for “seats that cannot be reclined unless the belt is used”.⁴
• Enable seat-back power, wi-fi or entertainment only when the belt is fastened. Deployed once as a taxi marketing campaign. Never trialled properly.

Both designs make access to a seat amenity depend on belt use, applying the interlock principle without preventing the vehicle from moving. They also avoid adding audible noise.

Varying the alert instead of holding it constant. The in-vehicle warning literature reports that responses to predictable signals can decline with repeated exposure, and that alerts that increase in intensity can convey urgency without adding constant noise. It also reports that how effective an alert is perceived to be and how annoying it is perceived to be are strongly correlated.⁴² An alert that starts quietly and escalates, or that varies, may help maintain attention better than an unchanging tone, although the long-term effect on buses remains untested. This addresses the limitation in section 13.2 directly.

10.3Belt maintenance may be the cheapest way to improve use

Two facts point at the same conclusion. The highest-effect category in Table 2 is reducing effort. The most common reasons passengers give for not buckling are dirty belts, belts that will not extend or latch, and belts that do not fit.⁴

The intervention that follows needs no electronics at all: clean and inspect the belts on a schedule, and specify belt geometry that fits both large and small passengers. The Swedish authors recommend exactly this.

Operators fitting sensors should also address belt condition and fit, since passengers report these as reasons for non-use.

10.4An applied example

AITSC manufactures a system that implements the specification in section 6.3.

Two passive switch sensors sit at each seat, one in the cushion and one inside the buckle. Together they establish which seats are occupied and which belts are fastened. After a 20-second grace period once the doors close, any seat that is occupied and unbelted triggers an 80 dB alert in that row, and the alert continues until the belt is latched. Monitoring runs for the whole trip, so the system also detects a passenger who unbuckles later. An optional seat-mounted unit adds a flashing indicator that makes the location of the unbelted seat visible from the aisle.

Measured against the five principles in section 10.1: it senses each seat, it alerts only occupied seats, the alert ends when the belt is fastened, and it monitors belt use throughout the journey. It does not reduce the effort of buckling. That fifth principle requires the operator to maintain the belts and address fit under section 10.3.

Measured against section 10.2, neither untested design is implemented. The aisle indicator may draw attention to the unbuckled seat and create social pressure. Evidence for that effect is limited, as rows 10 and 13 of Table 2 show, so it should be treated as a supporting feature.

11Discussion: what this review does not show

Listed separately from the limitations, because these are conclusions a reader might reach that the evidence does not support.

• It does not show that alerts are the best behavioural mechanism. Interlocks and reductions in effort have produced larger measured effects in other settings. The review identifies alerts as a promising approach that can operate under the bus conditions considered.
• It does not show that a reminder would have prevented the deaths at Greta. Several factors contributed to the risk of being thrown from the vehicle. Glazing, seat design, the road environment and driver health all contributed.¹ Raising belt use addresses one factor among several.
• It does not show that information is worthless. Prompts at the point of decision produce real, measurable, short-lived effects. They may supplement other measures, but have limited support as the main way to increase belt use.
• It does not establish an effect size for an audible per-seat alert on a bus. That trial has not been run.

12Discussion: implications

Three implications follow, in order of how directly the evidence supports them.

For regulators. ADR 68/01 requires reminders, whose effectiveness in increasing bus passenger belt use has limited direct evidence. Car reminder rules increasingly require sensing at individual seats. The cost-benefit reasoning used for car reminders warrants assessment for buses as well.

For operators. Two measures are available now and neither requires a rule change. Repair and maintain the belts, because that addresses the reason passengers actually give. Fit per-seat detection, because it can monitor buckle state when the driver cannot see the passenger area. Chain of Responsibility duties apply now, on vehicles that ADR 68/01 will never reach.

For researchers. Three areas need further trials. A per-seat audible alert on a bus, which has never been tested anywhere. Making seat amenities conditional on belt use, as described in section 10.2. And measuring whether responses to a fixed cabin alert decline over years rather than weeks.

13Limitations

13.1Direct evidence from buses is limited

The bus-specific evidence is one observational study and one pilot trial. The remaining evidence comes from cars, aviation, healthcare, industry and public health. Applying those findings to buses requires inference. The strongest effect-size evidence in this review comes from populations that are not bus passengers.

The estimates in Table 2 come from different study designs and support a broad qualitative assessment, rather than a precise ranking of effect sizes.

13.2Habituation has not been measured

The controlled trials of persistent reminders ran for weeks. A coach runs for about 26 years. We found no study measuring the response to a fixed in-cabin alert over anything approaching that period, and the hand hygiene literature names the assumption that reminders work indefinitely as unexamined.³⁸

Long-term response is a major uncertainty for per-seat alerts. It supports testing the varying and escalating alerts proposed in section 10.2.

13.3A latched belt may not be worn correctly

A buckle switch detects the latch, not the person. Participants in the Kidd and Singer trial sat on fastened belts and routed belts behind their backs, and 21 per cent circumvented a persistent reminder.¹³ Devices that defeat buckle sensors are sold openly. A wearing rate calculated from buckle sensors may overestimate how many passengers are correctly wearing their belts.

13.4Seat alerts do not replace broader policy measures

Chater and Loewenstein argue that measures aimed at individual behaviour can be promoted in place of broader policy changes, benefiting those who oppose those changes.⁴³ A per-seat device fitted by an operator targets individual passenger behaviour, and this review is published by a company that sells one. This concern is relevant to the review’s recommendations and commercial disclosure.

Broader policy options include mandatory retrofitting and consistent enforcement. Neither exists. Australian Design Rules are not retrospective, fleet turnover is estimated to take at least 25 years, and the investigating regulator has asked for a study into whether its own new rule works rather than treating the question as settled.¹ Operators can introduce measures on existing vehicles while broader policy changes are considered and the fleet is replaced. These measures do not remove the need to assess those policy options.

13.5Search method

This review followed a systematic two-stage search strategy, set out in section 2.1, but it did not have access to subscription bibliographic databases. Searching was done through a general index rather than Medline, Embase, Scopus, TRID or PsycINFO. The review does not provide exact search queries that others can repeat, a count of duplicate records removed, or a PRISMA flow diagram showing how studies were selected. It was not registered in advance.

The index used is weighted towards English-language and United States sources. Australian, Korean and Scandinavian material was found only through searches using specific names. Several sources were read through abstracts and published summaries rather than full text. Readers should treat the coverage as broad rather than exhaustive.

14Conclusion

Across the fields reviewed, changing the effort or immediate consequences of an action generally produces larger effects than information alone. The studies differ in design and population, so this pattern does not establish a precise effect size for bus passengers.

Bus operating conditions limit approaches that require repeat contact with identified passengers, continuous supervision, delaying departure or frequent enforcement. Passengers may also unbuckle during the journey. Per-seat sensors linked to alerts address these conditions: an alert responds to an occupied seat with an unfastened belt, stops when the belt is fastened, and remains available throughout the trip.

Australia’s response to the Greta rollover does not meet that specification. ADR 68/01 requires a label, a lit sign and one recorded announcement on a timer, with no sensing of any kind. These measures provide information to a population with reported wearing rates of 7 to 45 per cent; their effects on bus passenger belt use remain poorly measured. Europe stops at the front seats of a bus. The rules reviewed do not require buckle-sensing reminders throughout a bus’s passenger area.

The one bus reminder trial located reported wearing rates of 9.6 per cent without the system and 59.0 per cent with it, using visual alerts. Car studies provide further support for buckle-linked reminders. Direct trials are still needed to establish the effect and long-term acceptability of audible per-seat alerts on buses.

References

1. Office of Transport Safety Investigations. Bus safety investigation report: Rollover of coach 4666MO, Greta, NSW, 11 June 2023. NSW, 2026.
2. Kargar S, Ansari-Moghaddam A, Ansari H. The prevalence of seat belt use among drivers and passengers: a systematic review and meta-analysis. Journal of the Egyptian Public Health Association, 2023;98:14.
3. Nemire K. Seat belt use by adult rear seat passengers in private passenger, taxi, and rideshare vehicles. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 2017; and Nemire K. Warning signs to fasten seat belts result in higher rates of rear seat belt use in rideshare vehicles. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 2019.
4. Anund A, Forward S, Sjörs Dahlman A. Seat belt usage in buses: an observation study of usage and travellers’ perspectives. Accident Analysis and Prevention, 2023;190:107138.
5. Oporia F, Kibira SPS, Jagnoor J, Kobusingye O, Makumbi FE, Isunju JB, Nuwaha F. Peer-led training improves lifejacket wear among occupational boaters: evidence from a cluster randomized controlled trial on Lake Albert, Uganda. PLOS ONE, 2023;18(10):e0292754.
6. A systematic review of delay discounting among workers: framing effects, resource availability, and health. Frontiers in Psychology, 2026;17:1801136.
7. Neal DT, Wood W, Quinn JM. Habits: a repeat performance. Current Directions in Psychological Science, 2006;15(4):198–202.
8. Chakraborty M, Singh H, Savolainen PT, Gates TJ. Examining correlation and trends in seatbelt use among occupants of the same vehicle using a bivariate probit model. Transportation Research Record, 2021.
9. Willis C, Lybrand S, Bellamy N. Alcohol ignition interlock programmes for reducing drink driving recidivism. Cochrane Database of Systematic Reviews, 2004, Issue 4, CD004168.
10. Elder RW, et al. Effectiveness of ignition interlocks for preventing alcohol-impaired driving and alcohol-related crashes: a Community Guide systematic review. American Journal of Preventive Medicine, 2011;40(3):362–376.
11. Luo Y, Li A, Soman D, Zhao J. A meta-analytic cognitive framework of nudge and sludge. Royal Society Open Science, 2023;10:230053.
12. Van Houten R, Malenfant JEL, Reagan I, Sifrit K, Compton R, Tenenbaum J. Increasing seat belt use in service vehicle drivers with a gearshift delay. Journal of Applied Behavior Analysis, 2010;43:369.
13. Kidd DG, Singer J. The effects of persistent audible seat belt reminders and a speed-limiting interlock on the seat belt use of drivers who do not always use a seat belt. Journal of Safety Research, 2019;71:13–24.
14. Park J, et al. The effectiveness of seat belt reminder (SBR)s by analyzing the result of the pilot project of an interurban bus with SBR. Korea Automobile Testing & Research Institute (KATRI), Paper Number 23-0190, Proceedings 27th International Technical Conference on the Enhanced Safety of Vehicles, 2023.
15. Hagenzieker MP, Bijleveld FD, Davidse RJ. Effects of incentive programs to stimulate safety belt use: a meta-analysis. Accident Analysis and Prevention, 1997;29(6):759–777.
16. Ebert JP, Yan R, Friday S, Bartolozzi K, McDonald CC, Small D, Delgado MK. Behavioral interventions for increasing seat belt use and decreasing distracted driving using telematics: a national randomized trial. American Journal of Public Health, 2025;115(5):758–768.
17. Krause TR, Seymour KJ, Sloat KCM. Long-term evaluation of a behavior-based method for improving safety performance: a meta-analysis of 73 interrupted time-series replications. Safety Science, 1999;32(1):1–18.
18. Nagin DS. Deterrence: a review of the evidence by a criminologist for economists. Arizona State University, Academy for Justice.
19. National Highway Traffic Safety Administration. Analyzing the first years of the Click It or Ticket mobilizations. DOT HS 811 232; and Strengthening seat belt use laws, DOT HS 810 890W.
20. Effects of gamification on behavioral change in education: a meta-analysis. 2021.
21. Goldstein NJ, Cialdini RB, Griskevicius V. A room with a viewpoint: using social norms to motivate environmental conservation in hotels. Journal of Consumer Research, 2008;35:472–482.
22. Bellicha A, et al. Getting more people on the stairs: the impact of point-of-decision prompts. Social Science & Medicine, 2017; and Community Preventive Services Task Force, point-of-decision prompts to encourage use of stairs.
23. Geller ES, Bruff CD, Nimmer JG. “Flash for life”: community-based prompting for safety belt promotion. Journal of Applied Behavior Analysis, 1985; and Cox CD, Cox BS, Cox DJ. Prompting safety belt use: comparative impact on the target behavior and relevant body language. Journal of Applied Behavior Analysis, 2010;43:321–325.
24. Mertens S, Herberz M, Hahnel UJJ, Brosch T. The effectiveness of nudging: a meta-analysis of choice architecture interventions across behavioral domains. PNAS, 2022;119(1):e2107346118.
25. Maier M, et al. No evidence for nudging after adjusting for publication bias. PNAS, 2022;119(31).
26. Szaszi B, et al. No reason to expect large and consistent effects of nudge interventions. PNAS, 2022;119(31).
27. Rotella A, Sparks AM, Mishra S, Barclay P. No effect of ‘watching eyes’: an attempted replication and extension investigating individual differences. PLOS ONE, 2021;16(10):e0255531.
28. Crowding-out effects of opt-out defaults: evidence from organ donation policies. PNAS Nexus, 2025;4(10); and Opt-out defaults do not increase organ donation rates. Preventive Medicine, 2024.
29. Lourens A, Sinclair M, Willems B, Young T. Education, incentive, and engineering-based interventions to promote the use of seat belts. Cochrane Database of Systematic Reviews, 2024, CD011218.pub2.
30. Daniels AC. PIC/NIC Analysis and the antecedent-behaviour-consequence model in industrial safety. Aubrey Daniels International.
31. Ederer DJ, et al. The Safe Systems Pyramid: a new framework for traffic safety. Transportation Research Interdisciplinary Perspectives, 2023;21:100905.
32. Shingo S. Zero Quality Control: Source Inspection and the Poka-Yoke System. Productivity Press, 1986; and Norman DA. The Design of Everyday Things. 1988.
33. Kidd DG, McCartt AT, Oesch NJ. Attitudes toward seat belt use and in-vehicle technologies for encouraging belt use. Traffic Injury Prevention, 2014.
34. Federal Aviation Administration. Advisory Circular 120-88A: Preventing injuries caused by turbulence.
35. Seneviratne D, Molesworth BRC. Employing humour and celebrities to manipulate passengers’ attention to pre-flight safety briefing videos in commercial aviation. Safety Science, 2015.
36. Muir HC, et al. Influence of cabin crew during emergency evacuations at floor level exits. UK Civil Aviation Authority, CAA Paper 95006, 1996.
37. Cvach M. Monitor alarm fatigue: an integrative review. Biomedical Instrumentation & Technology, 2012;46(4):268–277; and Computational approaches to alleviate alarm fatigue in intensive care medicine: a systematic literature review, 2022.
38. Electronic monitoring systems for hand hygiene: systematic review of technology. Journal of Medical Internet Research, 2021;23(11):e27880.
39. Analysis of the face mask use by public transport passengers and workers during the COVID-19 pandemic. International Journal of Environmental Research and Public Health, 2022.
40. United Nations Economic Commission for Europe. UN Regulation No. 16: Uniform provisions concerning the approval of safety-belts, restraint systems, child restraint systems and ISOFIX child restraint systems.
41. National Highway Traffic Safety Administration. Federal Motor Vehicle Safety Standards; Occupant Crash Protection, Seat Belt Reminder Systems, Controls and Displays. Final rule, December 2024.
42. Marshall DC, Lee JD, Austria PA. Alerts for in-vehicle information systems: annoyance, urgency, and appropriateness. Human Factors, 2007;49(1):145–157; and Graded auditory warnings during in-vehicle use, AutomotiveUI, 2012.
43. Chater N, Loewenstein G. The i-frame and the s-frame: how focusing on individual-level solutions has led behavioral public policy astray. Behavioral and Brain Sciences, 2023;46:e147.