Report of collision Bedford (Elstow) - 19/06/26

[6Gman]

As a car driver, not a train driver, I know to stop at a red traffic light, even if the next one along the road is green...

And if I pass a red traffic light, I don't get a warning in the car - potentially I get a fixed-penalty notice in the post.

As a car driver I have driven through a red light because of "reading through" to the next set of lights. On a long straight road with multiple sets of lights in view.

 

[JammyJames08]

For me (as a driver) my thoughts at this stage are:-

What was the previous signal showing?

&

What causes a distraction long enough up to and beyond the red signal? (which the AWS was cancelled for) my opinion is that he wasn't even looking at it, to not even apply the emergency brakes on seeing the red and make an attempt to stop until the last few seconds, suggests to me a distraction in the cab.

 

[sharpener]

As a car driver I have driven through a red light because of "reading through" to the next set of lights. On a long straight road with multiple sets of lights in view.

Easily done. I have also stopped at the "wrong" stop line because there was a central pedestrian island with two sets of lights on it and the same on the far side of the junction. As at the Longdon Colney roundabout, more is not necessarily better.

 

[bramling]

For me (as a driver) my thoughts at this stage are:-

What was the previous signal showing?

&

What causes a distraction long enough up to and beyond the red signal? (which the AWS was cancelled for) my opinion is that he wasn't even looking at it, to not even apply the emergency brakes on seeing the red and make an attempt to stop until the last few seconds, suggests to me a distraction in the cab.

There’s a pretty much infinite list of possibilities, some work-related, others not.

The industry is quite robust about training people to stay focused when on yellow aspects. The difficulty here is that a yellow aspect was quite typical at this signal, so it’s very easy to see how the driver could have gone onto “auto pilot”, and driven as he would have done probably hundreds of other times, especially if he’d seen the 810 coming past at full speed. That is all quite easy to explain.

But my feeling, for what it’s worth, is something else then happened, even if only for a few seconds. A micro sleep would certainly explain it, although just half a minute later the driver seems to have been pretty alert in seeing the train standing ahead. It’s interesting to consider what might have gone on, but realistically who really knows at this point.

 

[norbitonflyer]

There's also here (also on the A308) where you can see the two sets of lights (controlling pedestrian crossings) but miss the Give Way line inbetween where you merge with the A3 coming in from the right

Street View · Google Maps

Explore a place in a more immersive way in Google Maps.

www.google.co.uk

 

[sharpener]

There's also here (also on the A308) where you can see the two sets of lights (controlling pedestrian crossings) but miss the Give Way line inbetween where you merge with the A3 coming in from the right

Street View · Google Maps

Explore a place in a more immersive way in Google Maps.

www.google.co.uk

Thanks for reminding me @norbitonflyer. Came off my C15 motorbike on the approach to Robin Hood roundabout ~50 yrs ago! Possible to turn L into the park back then but quite dangerous because traffic behind didn't expect it.

Sorry mods.

 

[InvernessTMD]

How did a thread on a rail accident turn into a debate over the finer points of traffic lights and intersections?!

 

[DelW]

How did a thread on a rail accident turn into a debate over the finer points of traffic lights and intersections?!

It seems certain that the collision was a result of human factors (driver actions and behaviours) rather a failure of signalling or train systems. As such, relating those behaviours to roughly analogous situations that far more of us have experienced, helps with understanding of what might have happened in that critical few seconds.

 

[Taunton]

Even discounting age, the threshold for greater energy absorption comes in at 190km/h (118mph), so on basic principles you'd expect the 810 to fare much better as it should be designed to the higher standard for higher speeds.

It's notable how, over time, innumerable accidents have been between a main line train and a more local service train, and very rarely a main line train travelling at full speed; one wonders why the two types should have different survival requirements.

 

[Harpo]

But my feeling, for what it’s worth, is something else then happened, even if only for a few seconds. A micro sleep would certainly explain it,

An AWS bell needs cancelling within around 3 seconds and the driver achieved that.

although just half a minute later the driver seems to have been pretty alert in seeing the train standing ahead.

The RAIB timeline shows 27 seconds between the AWS and (FFCCTV) visibility of the train ahead, but a further 9 seconds for power off/brakes. That 9 seconds is quite a long time.

 

[edwin_m]

It's very clear that this is the accident that has been waiting to happen ever since the decision was made to fit TPWS only at high risk signals. That decision was right in cost effectiveness terms, but there was never any doubt that one day the holes in the Swiss cheese would line up.

Finding out what the contributory causes were, to the extent that that is possible, is of course highly necessary and the Final Report will bring them out, but addressing them can only mitigate the risk to a minor extent. One death is a tragedy, but it's already clear that the fact that the toll was so small is down to the vastly improved crashworthiness of trains and to luck combined with prompt action by train crew. There will undoubtedly be recommendations to mitigate the consequential risk as well - and some ideas have been discussed on this thread already.

It will be very interesting to see what the recommendations say about the fundamental problem of providing train protection at every signal.

The "crumple zones" at the end of each coach are specifically aimed at mitigating the effect of end-on collisions, they have little benefit in any other scenario and I believe this is the first time they have made a difference. This is precisely the type of collision that will result from passing signals without TPWS, so it could be said that this is the mitigation for those collisions. Other crashworthiness measures such as avoiding aggressive corners in the vehicle interiors will have helped mitigate the consequences of multiple accidents.

Also, possibly about the inclusion of a "nosecone" to take the brunt of the impact in the event of an accident - the flat fronted Desiro is obviously and tragically much, much worse off in the collision with the 810. Would a collision between two 810s both with long bonnets to absorb some of the energy and crush less of the cab have changed the outcome?

It's notable how, over time, innumerable accidents have been between a main line train and a more local service train, and very rarely a main line train travelling at full speed; one wonders why the two types should have different survival requirements.

There is a limit on the amount of energy absorption it is physically possible to incorporate in a train front end, although I think the 360 pre-dates the latest regulations on this. This limit is far less than the energy that has to be dissipated in a high-speed collisions such as this, though it may make more difference in a lower-energy collision such as striking a vehicle at a level crossing.

 

[Snow1964]

The RAIB timeline shows 27 seconds between the AWS and (FFCCTV) visibility of the train ahead, but a further 9 seconds for power off/brakes. That 9 seconds is quite a long time.

The location of a camera and the drivers viewpoint are slightly different, so going around a long gentle corner changes when each first saw it.

In reality there is going to be some thinking and reaction time, to an unexpected event, before action to cut power and apply max braking. Agree 9 seconds is quite long, but unlikely to be less than about 3-5 seconds at best, as not a reflex action.

 

[Bill57p9]

The "crumple zones" at the end of each coach are specifically aimed at mitigating the effect of end-on collisions, they have little benefit in any other scenario and I believe this is the first time they have made a difference. This is precisely the type of collision that will result from passing signals without TPWS, so it could be said that this is the mitigation for those collisions. Other crashworthiness measures such as avoiding aggressive corners in the vehicle interiors will have helped mitigate the consequences of multiple accidents.



There is a limit on the amount of energy absorption it is physically possible to incorporate in a train front end, although I think the 360 pre-dates the latest regulations on this. This limit is far less than the energy that has to be dissipated in a high-speed collisions such as this, though it may make more difference in a lower-energy collision such as striking a vehicle at a level crossing.

Is it still the case that the front third of UK rail vehicles rated for speeds over 110mph cannot be passenger accommodation?
Or has this been overtaken by outcome based requirements.

 

[AlterEgo]

An AWS bell needs cancelling within around 3 seconds and the driver achieved that.

The RAIB timeline shows 27 seconds between the AWS and (FFCCTV) visibility of the train ahead, but a further 9 seconds for power off/brakes. That 9 seconds is quite a long time.

It's actually 11 seconds.

Train 1B67 first becomes visible in FFCCTV images from train 1H46 at 17:13:13.

Analysis of the data from train 1H46’s OTDR shows that the driver reduced the traction power demand from 17:13:24

 

[fishwomp]

It's notable how, over time, innumerable accidents have been between a main line train and a more local service train, and very rarely a main line train travelling at full speed; one wonders why the two types should have different survival requirements.

They are most certainly numerable. I listed the last few earlier in another thread - Bedford, 2026; Talerddig, Oct 2024 ; low speed at Aviemore (2023) ; Salisbury (2021). Meanwhile, level crossing accidents would be a few times per year - often user worked crossings but not always.

Most accidents don't involve other trains - eg. tractors, cars, buffers, landslips, cattle.

Those kinds of accident do depend on the max speed of the line and unit/loco.

 

[Flying_Turtle]

This is an interesting comment for the enquiry to consider. Stepping up signals just a few seconds before the train reaches them was the key factor in the 1995 Toronto Subway major accident, which if this is actually the case with WH356 it would be instructive for the Bedford enquiry to read about for parallels - for speed control purposes signals were clearing on a timer, and drivers got used to this over time and would approach red signals at speed, with them clearing just in time, sometimes just as they passed from view. I saw this myself from the train front windows before the accident and was surprised at the approach, and an article in a UK rail magazine made similar comments.

Protection for the section ahead was by trip train stops, like the London Underground, and indeed they had periodic trips. In the Toronto accident the train ahead was unusually delayed (sound familiar from Bedford?), but the trip mechanism was out of alignment. Here's a summary ...

1995 Russell Hill subway accident - Wikipedia

en.wikipedia.org

Portugal doesn't have the problem of step ups (you show the lowest relevant aspect and keep it like that) but a few decades ago a situation somewhat similar to this, prompted an important rule change. A signal indication is always valid until the next signal... so if one comes from yellow to red and see the signal clearing, to green It can only comply with the green once it has passed it.

 

[hwl]

Is it still the case that the front third of UK rail vehicles rated for speeds over 110mph cannot be passenger accommodation?
Or has this been overtaken by outcome based requirements.

The threshold is 190kmh (~118mph) so you could do 115mph but the gearing and extra installed power requirements make 110mph a sweet spot.

 

[bramling]

There may have been a deficiency with the signal, but I would expect the RAIB to examine the CCTV carefully to determine if the red aspect of the signal was sufficiency visable.
We know it was at least lit, because they would have reported it it wasn't. If the signalling failed to detect sufficient electrical current being used by the signal (indicating that it was lit) then the previous signal (here the junction signal) would not have shown a proceed aspect.

In any event, in daylight, when it's easier to see a signal that has a defect, if an signal aspect has a deficiency, drivers are supposed to treat it as a red and stop and report the problem.

What springs to mind is Cowden where I recall the signal was said to have been producing “reasonable light”, but there was some debate over the implications of this particularly in thick fog which was the conditions at the time.

No doubt things will be different nowadays due to the widespread ditching of filament lamps, but LEDs can have their own characteristics.

 

[bramling]

It's actually 11 seconds.

Train 1B67 first becomes visible in FFCCTV images from train 1H46 at 17:13:13.

Analysis of the data from train 1H46’s OTDR shows that the driver reduced the traction power demand from 17:13:24

I still don’t really see that as being a long time. So from the time the train first came into view, the driver has to have seen it, worked out what line it was on (bearing in mind a busy 4-track railway), probably had a double take and a moment to process if what he was seeing was actually real, followed by the realisation of the situation.

Seeing a train a few hundred metres away is an everyday sight for a train driver, whilst driving straight under a correctly aligned and illuminated red signal in perfect weather conditions - if that all turns out to be the case - most definitely isn’t routine.

 

[swt_passenger]

Is it still the case that the front third of UK rail vehicles rated for speeds over 110mph cannot be passenger accommodation?
Or has this been overtaken by outcome based requirements.

No longer a rule. That explicit requirement was removed from the group standards quite a while ago, ie well over 10 years. I’ll try and find the thread we had about it…

So in this thread about DVTs from 2013:

Why cant East Coast 225 DVTs be converted for passenger use?

Is there any reason the DVTs on East Coast can't be converted for use by passengers? I assume there are good reasons or it would have already been done.

www.railforums.co.uk

…I had found that the relevant group standards at issue 3 had the phrase removed by 2000, and no editions ever actually had a ’one third’ rule.

 

[yorkie]

For discussion regarding whether or not there should be AWS/TPWS or any other type of protection at all signals (or more signals), please use the following thread:

Should train protection be provided at every signal?

Moderator note: Split from https://www.railforums.co.uk/threads/report-of-collision-bedford-elstow-19-06-26.303537/ It's very clear that this is the accident that has been waiting to happen ever since the decision was made to fit TPWS only at high risk signals. That decision was right in cost...

www.railforums.co.uk

Also, comparisons either between modes of transport (e.g. driving cars vs driving trains), or between different incidents, are also best posted in separate threads (if anyone would like to create one, feel free to link to any such thread from here)

This thread should remain concise and focus on the specific incident in the thread title, please.

Thanks

 

[Archvile]

Not always, by any means. I've been caught out when an amber came on, only for the subsequent green to be an arrow for a direction I wasn't planning on going.

Needless to say, in almost all situations on the road, there is ample time and brake force available to solve most moderate (even severe) mistakes.

The railway punishes relatively minor errors readily.

There is also an overwhelmingly greater standardisation governing road infrastructure than railway infrastructure, hence it is far more difficult to make the mistakes in the first place when driving a car.

 

[Signal Head]

What springs to mind is Cowden where I recall the signal was said to have been producing “reasonable light”, but there was some debate over the implications of this particularly in thick fog which was the conditions at the time.

No doubt things will be different nowadays due to the widespread ditching of filament lamps, but LEDs can have their own characteristics.

The signal at Bedford platform is an LED head, the junction signal (WH356) is still a filament style head but appears to have been fitted with LED inserts ('light engines') in place of the SL35 filament lamps. It's reasonable to assume the other original heads in the area have been similarly fitted as is usual practice.

The light engines produce aspects of comparable brightness to the lamps, but colour rendition, particularly of the green, is noticeably stronger. A modern LED head appears much brighter than the older style heads.

 

[AHBD]

Yes, he cleared the junction and just kept on accelerating. Don't believe he read thru, or across, he was experienced, but why,
Don't think we'll ever know.

Do train drivers learn how avoid being caught out by eye saccades, where the brain 'makes up stuff based on previous view' as it moves your eyeball: probably a cause of looking but not seeing road collisions:

Original RAF members article is in Dropbox link

 

[MadMac]

I still don’t really see that as being a long time. So from the time the train first came into view, the driver has to have seen it, worked out what line it was on (bearing in mind a busy 4-track railway), probably had a double take and a moment to process if what he was seeing was actually real, followed by the realisation of the situation.

Seeing a train a few hundred metres away is an everyday sight for a train driver, whilst driving straight under a correctly aligned and illuminated red signal in perfect weather conditions - if that all turns out to be the case - most definitely isn’t routine.

Correct.There has to be a certain ”suspension of disbelief” when you see a train ahead after having it drummed into you repeatedly that “This can’t happen”. That eats into critical time.

 

[Hairy Bear]

Do train drivers learn how avoid being caught out by eye saccades, where the brain 'makes up stuff based on previous view' as it moves your eyeball: probably a cause of looking but not seeing road collisions:

Original RAF members article is in Dropbox link

A good driver teaching the route will point out potential read through sites, read across sites. There's only a couple on the whole line London to Sheffield, this is not one of them. As he made the move from up slow to up fast, at Bedford South junction, WH356 stepped to one yellow,

This should have told him that the 810 that just passed him up fast has not cleared the following section and stood out as unusual. Instead of keeping the acceleration and speed within limits preparing to stop next signal, it was full demand.

The next signal is on the crest of a hill climbing away and visible from the junction. But he just kept on accelerating, why ?, convinced himself that previous was green as normal received, does not explain why he didn't react to the oncoming Red.

As a driver of 38 years on the route, I am mystified .

 

[Belperpete]

Correct.There has to be a certain ”suspension of disbelief” when you see a train ahead after having it drummed into you repeatedly that “This can’t happen”. That eats into critical time.

It is also worth bearing in mind that on multiple track railways, drivers see trains ahead almost routinely. However, they become accustomed to ignoring them, as they are invariably on a different track.

In this case, it is likely that there will be a delay in the driver seeing the train ahead, then further delay before he realises and accepts the unthinkable that it is on his track.

It is also possible, as you say, that the driver then went into a state of shock, that temporarily paralysed him, if his brain was unable to accept the situation "this can't be happening". Like a rabbit in the headlights.

 

[MarkyT]

Correct.There has to be a certain ”suspension of disbelief” when you see a train ahead after having it drummed into you repeatedly that “This can’t happen”. That eats into critical time.

Also taking into account the cognitive dissonance of seeing a bright green signal aspect displayed prominently above it, very likely coming into conscious view a little before the short dark-hued train that shouldn't be there, in the driver's mind.

 

[Belperpete]

I think the intention of the delay to 'better than Yellow' was to force an AWS caution indication for a diverging movement, which at first thought seems a sensible idea - the train needs to be running at a reduced speed for the junction, so ensure the driver doesn't get an AWS 'bell'; the caution indication reinforces the junction indicator, not the aspect.

However - and here's where the law of unintended consequences steps in - doing that weakens the traditional meaning of the AWS.

Consider an arrangement where the signal releases straight up to Green (if aspects ahead allow) well before the train reaches the AWS magnet, which in most cases it probably will because there won't be anything immediately ahead, drivers are now used to getting a bell, so when on the odd occasion the signal cannot go all the way up to G, there's a audible difference as well as visual. Now the AWS caution reinforces the aspect, not the junction indicator.

When I first started in S&T design, junction release was designed to delay the clearing of the junction signal to slow the train down to the junction speed. It was therefore based on the line speed, junction speed, distance to the junction, and deceleration rate. As I recall, we had speed nomograms to help calculate it. This became impracticable with the introduction of modern high speed units with significantly different braking rates, such as HSTs. We therefore moved to the principle of holding the signal until the driver knew that he was taking the diverging route and relying on the driver to regulate the speed accordingly.

Even so, it was hard to shake off the concept that the signalling should be slowing the train down. So even when flashing aspects were introduced, it was still deemed wrong to give the driver a clear AWS indication at the junction signal. It was felt that giving the driver a bell might cause him to start accelerating, in excess of the junction speed. So the junction signal was held at yellow until the train had passed the AWS magnet. This often meant having to provide an additional track circuit specifically to detect the train passing the magnet, so doing this came at significant cost.

The drawbacks of course was that it meant that drivers had to assume that the following signal would be at red, which often meant slowing unnecessarily. Or drivers becoming accustomed to the following signal not actually being at red. This arrangement is therefore no longer used in modern installations.

However, there must be hundreds of existing installations that follow this practice. It would be nice if they could be amended, but the cost would be enormous. Each particular case would need to be risk assessed to see if it was justified in this case, and what other mitigations might be necessary instead. The change would need to be designed, checked, installed and tested. Staff would need to be got to site to make the changes. There would be project overhead costs to manage the work, etc etc. The resource and budget costs would be significant. With ETCS on the horizon, would it be worth it?

 

[AHBD]

... does not explain why he didn't react to the oncoming Red.

Bit if, due to eye movement / saccading the unconscious bits of the brain were feeding the conscious bit, made up data based on 'last input'. as described in the link, along with how to mitigate it (from RAF) , then he might not have seen a last minute change?

How human vision works seems pretty alarming use in for fast vehicles...