On the morning of 25 April 2005, at about 09:19, a West Japan Railway Company (JR West) rapid commuter train on the Fukuchiyama Line derailed on a tight left-hand curve at Amagasaki, in Hyōgo Prefecture, just before the morning peak had fully cleared. The seven-car train entered a curve of 304-metre radius — limited to 70 km/h — at about 116 km/h. The leading cars left the rails and slammed into a trackside apartment building; the first car was driven into the ground-floor parking area and the second was crushed against the structure. One hundred and seven people died, including the 23-year-old driver, and 562 were injured. It is one of the deadliest Japanese rail accidents of the postwar era.
The train was a 207-series electric multiple unit running a rapid service. Minutes before the curve, the driver had overrun his stop at Itami station, putting the train roughly 90 seconds behind schedule. Recovering lost time was not a trivial matter at JR West: the railway ran an intensely punctual operation, and a driver who fell behind or made an error faced disciplinary re-education. In the moments before the curve the driver appears to have been preoccupied — his attention drawn to the delay and to a radio exchange about how the overrun would be reported — and he braked late and inadequately, using the service brake rather than the emergency brake until it was too late to slow the train enough for the curve.
The accident was investigated by Japan’s Aircraft and Railway Accidents Investigation Commission (ARAIC), whose functions were later folded into the Japan Transport Safety Board, in report RA2007-3, published in 2007. The Commission did not reduce the disaster to the driver’s overspeed alone. It found the proximate cause was the train entering the curve far too fast because the driver did not brake in time, but it placed that error in a context that made it both more likely and more dangerous: JR West’s punitive retraining regime, known as nikkin kyōiku (“day-shift education”), which subjected erring drivers to humiliating non-driving duties and was found to have created fear that distracted drivers from driving; and the absence of an automatic train stop (ATS) system capable of enforcing the speed limit on that curve. The Finding is therefore Multi-factor.
The legal aftermath turned on whether senior JR West managers could be held criminally responsible for not having installed curve-protecting ATS. Successive prosecutions of former executives ended in acquittals, with courts finding the specific derailment had not been foreseeable in law even as they criticised the railway’s safety management.
On the morning of 5 October 1999, at about 08:09 BST, a Thames Trains Class 165 Turbo diesel unit leaving London Paddington passed signal SN109 while it was displaying a red stop aspect, ran on into the path of an oncoming First Great Western InterCity 125 high-speed train, and collided with it almost head-on roughly two miles out of the station, near Ladbroke Grove. The combined closing speed was around 130 mph. The diesel fuel from the Turbo ignited; the worst of the fire took hold in the leading coach of the HST. Thirty-one people died, including both drivers, and 417 were injured. It remains one of the worst rail accidents in late-twentieth-century Britain.
The public inquiry chaired by Lord Cullen did not reduce the disaster to a single mistake. The immediate event was a signal passed at danger (a SPAD): the 31-year-old Thames Trains driver, Michael Hodder, passed the single yellow caution at signal SN87 without slowing and then passed the red at SN109, some 563 metres before impact, apparently believing he had a proceed aspect. But Cullen found that the SPAD was made possible, and arguably made likely, by a system that had stacked the odds against the man in the cab. Signal SN109 was badly sighted — its aspect was obstructed and confused by the overhead-line gantry and the geometry of the approach, and on a bright autumn morning low sunlight washed out the signals. The signal had already been passed at danger eight times in the six years before the crash, yet no effective action had been taken and drivers were not warned. Hodder himself had qualified only about two weeks earlier, had no prior railway experience, and had been trained by instructors who did not know SN109 was a notorious multi-SPAD signal.
Because the cause was operational and distributed across the driver, the signal, the training regime and the management of a known hazard, the finding here is recorded as multi-factor. The infrastructure also lacked the one technology that would have stopped the train automatically: Automatic Train Protection (ATP) had been trialled but not adopted across the network, a decision taken largely on cost. Cullen’s two-part inquiry produced a sweeping set of recommendations and reshaped the regulation and management of British railway safety.
The legal reckoning fell on the corporate operators rather than on any individual. Thames Trains was fined two million pounds in 2004 for health-and-safety failures in driver training. Network Rail, as Railtrack’s successor in respect of the infrastructure, pleaded guilty and was fined four million pounds in 2007, both companies meeting substantial costs. No one was convicted of manslaughter.
At about 08:10 on the morning of 12 December 1988, a crowded commuter express from Basingstoke to London Waterloo ran into the back of a stationary train that had stopped, correctly, at a signal in the cutting just south of Clapham Junction station. Moments later a third, empty train travelling in the opposite direction struck the wreckage. Thirty-five people were killed and 484 injured, making it one of the worst British rail disasters of the post-war era. The trains, the drivers and the signalling rules were not at fault. The signal that should have protected the standing train had failed to the most dangerous state a signal can: it had shown a green proceed aspect on a section of line that was occupied.
The cause was a maintenance error of almost trivial mechanism and catastrophic consequence. During the Waterloo Area Resignalling Scheme (WARS), a signalling technician rewiring relays in the Clapham Junction “A” relay room had disconnected an old, now-redundant wire — but had not cut it back, insulated it, or tied it out of the way. Left loose at one end while still connected at the other, the redundant wire later came into contact with a relay terminal and created a false electrical feed, holding the signal at green even when the track circuit ahead was occupied by a stopped train. This is a “wrong-side” failure: instead of failing safe to red, the signal failed dangerously to clear. The Basingstoke driver behind it had no reason to expect a stationary train on a green road.
The public inquiry chaired by Anthony Hidden QC reported on 27 September 1989. Its central finding was that the disaster was caused by faulty wiring work, but its lasting force lay in why that faulty work had not been caught. There was no independent inspection or testing of the technician’s wiring; no wire count was performed; the man had never been told his working practices were wrong; and his judgement had been blunted by an extraordinary regime of overtime — he had worked seven days a week for thirteen consecutive weeks without a single day off. Hidden described a British Rail health-and-safety culture in which good intentions coexisted with sloppy practice. Because the root cause was defective installation and maintenance work, compounded by the absence of independent checks, the finding here is recorded as maintenance.
British Rail was fined 250,000 pounds for breaches of health-and-safety law; no individual was prosecuted for manslaughter. Hidden’s 93 recommendations reshaped British signalling discipline — mandatory independent testing of wiring, hard limits on safety-critical overtime, and reformed working practices — and the case is repeatedly cited as one of the spurs toward a corporate-manslaughter law in the United Kingdom.
At about 4:22 p.m. Pacific time on Friday, 12 September 2008, westbound Metrolink commuter train 111 ran past a red signal at Control Point Topanga, near the Chatsworth district of Los Angeles, and continued onto a single-track section straight into the path of eastbound Union Pacific freight train LOF65-12. The two trains met head-on. Twenty-five people died, including the Metrolink engineer; about 102 passengers were taken to hospitals, with dozens critically hurt. It was the deadliest accident in Metrolink’s history and, at the time, one of the worst US passenger-rail collisions in decades.
The investigation by the National Transportation Safety Board established the cause with unusual directness, because the engineer’s actions in the final minute were recorded in the cellular network. The Metrolink engineer, 46-year-old Robert Sanchez — employed by the contractor that supplied Metrolink’s train crews — had been exchanging text messages throughout his shift with a teenage railfan, and sent his last text at 4:22:01 p.m., roughly twenty-two seconds before impact. He did not observe and respond to the red stop signal at Control Point Topanga, did not slow, and never applied the brakes. The Union Pacific engineer applied emergency braking about two seconds before the collision, far too late to matter.
The NTSB’s probable cause is stated plainly: “the failure of the Metrolink engineer to observe and appropriately respond to the red signal aspect at Control Point Topanga because he was engaged in prohibited use of a wireless device, specifically text messaging, that distracted him from his duties.” Because the determining act was the operator’s distraction and failure at the controls, the finding here is recorded as driver. The Board further found, as a contributing factor, the absence of a positive train control system that would have automatically stopped train 111 short of the red signal and prevented the collision.
The case became the catalyst for a national change in US railway safety law. Within weeks Congress passed the Rail Safety Improvement Act of 2008, mandating the installation of positive train control (PTC) on main lines carrying passengers and on key freight routes. A federal liability cap limited the aggregate recovery for all passenger claims to 200 million dollars, which a court had to apportion among the many victims.
On the night of 12 May 2015, Amtrak Northeast Regional train 188 entered the Frankford Junction curve in the Port Richmond section of Philadelphia at 106 mph, more than twice the curve’s 50 mph speed restriction, and derailed. Eight passengers were killed and more than 185 people were taken to hospitals, several critically injured, from a train carrying about 250 occupants. It was the deadliest accident on the Northeast Corridor in decades and the event that forced the long-delayed completion of automatic speed enforcement on America’s busiest passenger railroad.
The train was a scheduled evening service from Washington, D.C., to New York City: one ACS-64 electric locomotive and seven Amfleet passenger cars, with a single locomotive engineer at the controls. There was no mechanical defect, no track failure, no obstruction. In the seconds before the derailment the engineer advanced the throttle and accelerated hard, as though the train had already cleared the sequence of speed-restricted curves that follow the departure from 30th Street Station. It had not. The locomotive and all seven cars left the rails; three cars overturned, and one was crushed and torn open.
The National Transportation Safety Board investigated and adopted its findings on 17 May 2016 in railroad accident report RAR-16/02. The Board’s probable cause was operational and human: the engineer accelerated to 106 mph entering a 50 mph curve because he had lost situational awareness, his attention most likely diverted by radio chatter about a nearby SEPTA commuter train that had made an emergency stop after being struck by a projectile. The Board found that the engineer believed he was at a point on the line where higher speed was authorized. It identified the absence of a positive train control (PTC) system on that stretch of track as a contributing factor — a safeguard that, the Board concluded, would have prevented the accident had it been active.
The legal arc ran for nearly seven years. The engineer, Brandon Bostian, faced charges that included involuntary manslaughter and reckless endangerment; after a series of dismissals, reinstatements, and appeals, a jury acquitted him on all counts on 4 March 2022. Amtrak, which accepted responsibility for the derailment, reached a $265 million settlement of victims’ claims in 2016.
At about 6:49 a.m. on 22 May 1915, beside the Quintinshill signal box near Gretna in southern Scotland, a southbound troop train ran at speed into a stationary local passenger train standing on the main line. Less than a minute later a northbound sleeping-car express ran into the wreckage. The gas-lit wooden carriages of the troop train caught fire and burned through the day. Most of the dead were soldiers of the 1/7th (Leith) Battalion of the Royal Scots, bound for Gallipoli. The death toll has never been exactly fixed, but it is generally given as about 226; it remains the deadliest rail disaster in British history.
The trains were not at fault, nor were the brakes, the rails, or the locomotives. The local passenger train had been shunted onto the main line to let the late-running express pass — a routine move. The two signalmen on duty, George Meakin and James Tinsley, then failed to protect it. Working through an irregular, unofficial shift-change that left the registers being copied up after the fact, and neglecting the block-working safeguards that should have made the lapse impossible, they forgot that the local train was standing on the main line in front of their box. With the line believed clear, the troop train was accepted and signalled forward into the occupied section.
The collision sequence was compounded by the period’s rolling stock and gas lighting. The troop train’s old wooden carriages, lit by compressed coal gas carried in cylinders beneath the floor, splintered on impact and then ignited; the fire spread through the wreckage and could not be extinguished until the following morning. Many soldiers who survived the impacts were trapped in the burning carriages.
The Board of Trade inquiry, conducted by its inspecting officer Lieutenant-Colonel Edward Druitt and reported in 1915, found the disaster due to the want of discipline on the part of the signalmen — a failure of operational signalling practice. Both men were prosecuted; tried in Scotland, they were convicted of culpable homicide. Meakin and Tinsley were imprisoned and released in late 1916. The finding is a signalling-operations verdict: human and procedural failure in the working of the block system, not a defect of any vehicle or structure.
On the morning of 31 January 2003, a CityRail Tangara interurban train left the rails on a tight curve in a rock cutting near Waterfall, about thirty-seven kilometres south of Sydney, killing seven people — including the driver — and injuring some forty more. The train was travelling at roughly twice the speed the curve was rated for. The reason it was travelling so fast was that the man controlling it was already dead or dying, and every system that should have stopped a train without a functioning driver had failed to do so.
The train was set G7, a four-car Tangara of the State Rail Authority’s CityRail network, running the early service from Sydney’s Central station south toward Port Kembla via the Illawarra line. At about 7:15 a.m., as the train ran through the descending, curving territory near Waterfall, the 53-year-old driver, Herman Zeides, suffered a sudden and fatal cardiac event. He did not brake, did not ease the power, and did not react to the approaching curve. The train accelerated to around 117 kilometres per hour and entered a curve rated for no more than 60. It derailed, the leading cars striking the rock walls of the cutting.
The reason the driver’s collapse did not stop the train was a defeated safety device. The Tangara’s “deadman” control was a foot pedal that the driver must hold within a defined range; releasing it or pressing it fully was meant to cut power and apply the brakes, on the assumption that an incapacitated driver’s foot would slip off. Investigators found that the unconscious driver’s body weight held the pedal within its active range, so the system read a live, attentive driver where there was none. The guard, the train’s second safety-critical crew member, was not monitoring speed and did not intervene with the emergency brake in time; the inquiry found the vigilance and guard-alerting arrangements inadequate to catch a silently incapacitated driver.
A Special Commission of Inquiry was established under the Honourable Peter Aloysius McInerney, who had also led the inquiry into the 1999 Glenbrook accident. McInerney’s final report, delivered in January 2005, located the immediate cause in the driver’s incapacitation and the resulting overspeed, while finding that an underdeveloped safety culture and a reactive approach to risk had left the railway without effective defences against exactly this scenario. The finding for this file is Driver — the proximate cause was the incapacitation of the man at the controls — but the report’s force lay in showing how predictable that single point of failure was, and how poorly the system had guarded against it.
On the afternoon of 19 September 1997, a Great Western Trains InterCity 125 high-speed train ran past two cautionary signals and a red stop signal at Southall, in west London, and collided with a heavy freight train that was crossing its path between the main lines and a goods yard. Seven people were killed and around 139 injured. The passenger train’s two protective systems — the Automatic Warning System fitted to all such trains, and the Automatic Train Protection installed on this particular fleet — were both inactive at the moment they were needed. The result was that a momentary lapse by a distracted driver was converted into a fatal collision that the technology aboard the train was specifically designed to prevent.
The HST, running the late-morning service from Swansea to London Paddington, was approaching Southall at line speed, around 125 mph, behind a single driver, Larry Harrison. The signals protecting the junction ahead were against him: a double-yellow warning, then a single-yellow warning, then red. The driver did not respond to the cautionary aspects — by his own account he was stowing belongings in his bag and did not register the warnings — and only applied the emergency brake when the red signal and the crossing freight came into view. By then the train could not be stopped. It struck the freight train’s wagons at well over 80 mph; the leading carriages were destroyed.
Why the train ran the signals without any automatic intervention was the heart of the inquiry. The HST’s AWS, which sounds a warning in the cab at every cautionary or red signal and applies the brakes if the driver does not acknowledge it, had been isolated earlier that day because of a fault, and crucially neither Railtrack nor the signaller had been told the train was running with it off. The more capable Automatic Train Protection, fitted to this Great Western fleet, was present but not in use because the driver was not trained to operate it and no qualified second person was in the cab. With both safeguards out of action, nothing stood between the driver’s distraction and the red signal but the driver himself.
The accident was examined in a public inquiry chaired by Professor John Uff QC. Its conclusion placed the immediate cause with the driver — a signal passed at danger after he failed to react to clear cautionary warnings — while condemning the operating regime that had allowed a high-speed train to run with its AWS isolated, its ATP unused, and only one person in the cab to catch a mistake. The criminal reckoning was uneven: Great Western Trains was fined £1.5 million under health-and-safety law, but manslaughter and corporate-manslaughter proceedings collapsed. The lasting legacy was the accelerated rollout of the Train Protection and Warning System across Britain’s network.
At about 2:35 a.m. on 4 February 2018, southbound Amtrak train 91 — the Silver Star, running overnight from New York to Miami — was diverted off the main track through a reversed hand-throw switch at Cayce, South Carolina, and collided head-on with a stationary CSX Transportation freight train parked on a siding. The two crew members in the lead locomotive, engineer Michael Kempf and conductor Michael Cella, were killed; about 116 other people aboard were injured out of the roughly 147 passengers and crew. The National Transportation Safety Board (NTSB) investigated and, in report RAR-19/02, found the collision was fundamentally an operational failure of risk management around a switch and a suspended signal system — not a track, equipment, or material defect.
The accident happened inside a stretch of CSX’s Columbia Subdivision where the automatic signal system had been deliberately taken out of service. CSX was installing components for positive train control (PTC), the federally mandated overlay designed, among other things, to stop a train approaching a misaligned switch. With the signals suspended, trains in the area were being run by track warrant and verbal authority, and the normal electronic backstop that would have flagged a wrongly lined switch was gone. A CSX local freight crew that had used the Charleston Highway siding earlier that night failed to return the hand-throw switch to its normal position lining the main track. Nothing else in the system caught the error.
The NTSB’s probable cause was unambiguous about where accountability lay. The board found that the collision was caused by CSX’s failure to assess and mitigate the risk of operating through a signal suspension — a condition that eliminated the system redundancy for detecting a switch in the wrong position — compounded by the CSX conductor’s failure to properly reposition the switch for the main track. The board further found that the Federal Railroad Administration’s (FRA) failure to implement effective regulation against misaligned-switch accidents, and Amtrak’s failure to conduct its own risk assessment before operating during the signal suspension, contributed to the accident. The verdict is therefore organizational and operational: a single human act of leaving a switch reversed was allowed to become lethal because the operators had stripped away every layer that should have stood behind it.
The Cayce collision came within a few weeks of two other high-profile Amtrak incidents and intensified scrutiny of the railroad’s safety oversight and of the slow national rollout of PTC. Litigation followed, with claims against CSX and Amtrak, and the NTSB used the case to renew long-standing recommendations that railroads operate at restricted speed near switches whenever a signal suspension is in effect.
At 8:22 a.m. on 2 December 1999, a crowded Sydney-bound CityRail interurban electric train collided with the rear of the stationary Indian Pacific transcontinental train on a curve east of Glenbrook station, in the lower Blue Mountains of New South Wales, Australia. Seven people on the interurban were killed and 51 were taken to hospital. The collision was not the product of a broken signal or a failed brake but of the rules and communications that governed how trains were allowed to move past signals that were already showing red — an operational failure that a judicial Special Commission of Inquiry, headed by Acting Justice Peter McInerney, traced to unsafe “pass-the-signal” procedures, poor communication between drivers and signallers, and an unreliable lineside telephone.
The sequence began with the Indian Pacific, the long Perth-to-Sydney passenger train, being authorised to pass a signal at danger at Glenbrook under the rules then in force. It proceeded and stopped at the next signal, which was also red. Its driver climbed down to use the trackside signal telephone to obtain authority to pass that second signal, but believed the phone was defective — a component was missing — and a delay of several minutes followed while the heavy train sat on the main line. Behind it, the CityRail V set interurban was itself authorised past the same red signal at Glenbrook and restarted up the grade. Required by the rule to proceed with extreme caution and to be able to stop short of any obstruction, the interurban instead caught up to the stationary Indian Pacific on the curve and struck its rear wagon.
The McInerney Special Commission of Inquiry — a judicial inquiry established under the New South Wales Special Commissions of Inquiry Act, not a transport-safety board issuing a “probable cause” — examined the disaster in an interim report (June 2000) and a final report presented on 11 April 2001. It declined to reduce the catastrophe to the interurban driver’s single error. Instead it found a web of contributing failures: a permissive signalling rule that routinely sent trains past red signals into territory that might be occupied; inadequate and ambiguous communication between drivers and signallers; the absence of a working, reliable means for a stopped driver to reach the signaller quickly; and shortfalls in training and procedure across the railway. The Commission made dozens of recommendations spanning rules, communications, training, and drug and alcohol testing, the great majority of which the State Government accepted.
The Glenbrook collision, coming after the 1977 Granville disaster and feeding directly into the same Commissioner’s later inquiry into the 2003 Waterfall crash, became a turning point in New South Wales rail safety, helping drive the overhaul of signalling rules and the creation of an independent rail safety regulator and investigator for the state.
At 7:34 a.m. on 18 December 2017, on the very first day of revenue service over a newly opened bypass route, southbound Amtrak Cascades train 501 entered a 30-mph curve near DuPont, Washington, at about 78 mph, derailed off a railroad overpass, and fell onto Interstate 5 below, striking highway vehicles. Three passengers were killed and 57 passengers and crew were injured; eight people in vehicles on the interstate were also injured. The National Transportation Safety Board (NTSB) investigated and, in report RAR-19/01, attributed the derailment not to a mechanical or track failure but to an organizational failure: the transit authority that owned and prepared the route had not effectively mitigated a known hazardous curve, had inadequately trained the engineer, and had begun service before positive train control (PTC) was operating to enforce the speed.
The accident occurred on the Point Defiance Bypass, a re-routed inland alignment of the Lakewood Subdivision that Amtrak Cascades was using for the first time that morning. The line ran from track posted for far higher speeds straight into a sharp 30-mph curve where it crossed over Interstate 5. The engineer, who had limited experience over the new territory, did not begin braking in time and took the curve at roughly 78 mph — more than twice the posted limit. The train’s locomotive and cars left the rails on the overpass; several cars plunged onto the freeway.
The NTSB’s probable cause centered on the Central Puget Sound Regional Transit Authority — Sound Transit — which owned the line and was responsible for preparing it for service. The board found the probable cause to be Sound Transit’s failure to provide an effective mitigation for the hazardous curve in the absence of operative positive train control, which allowed the engineer to enter the 30-mph curve far too fast because of his inadequate training on the territory and on the newer equipment he was operating. PTC, the federally mandated overlay that automatically enforces speed restrictions, had been installed in parts of the corridor but was not yet active in the curve where the train derailed. Had it been operating, it would have intervened to slow the train.
The board concluded the derailment was preventable and faulted the rush to inaugurate the bypass before its safeguards were complete, identifying failures by multiple agencies in planning, training, and oversight. The case sharpened national pressure to finish the long-delayed PTC rollout, produced civil litigation and settlements, and kept the bypass closed for years before service resumed under positive train control.