The failure of the DVSA’s hazard perception test to create perceptive drivers

As reported some considerable time ago, the DVSA have been working on replacing the filmed video clips for the hazard perception element of the theory test with computer generated imagery (CGI). It’s taken a while, but the latest news from the agency is that the hazard perception part of the driving theory test is set to be updated with from early next year.

The current filmed video clips are used to test candidates’ reactions to developing hazards on the road, but the DVSA acknowledge that whilst “the scenarios in these clips are still relevant… the image quality isn’t as clear or defined as modern digital technology allows.

“The first new clips show the same situations as the filmed clips, but are clearer on the screen and include updated vehicles, roads and surroundings to reflect modern day driving”.

DVSA Chief Executive, Alastair Peoples, said:

“Using CGI clips in the hazard perception test will allow us to present clearer, more up to date situations, ensuring the test fully reflects the realities of modern day driving.”

The DVSA worked with Nottingham University’s Accident Research Unit to trial the CGI clips. Research Fellow, Dr Peter Chapman, said:

“In our research we found that CGI clips retained all the benefits of traditional videos in discriminating between good and bad candidates, whilst allowing a more attractive, flexible, and up-to-date test.”

CGI will also allow for a wider range of hazards, for example including situations with vulnerable road users like children or cyclists and a range of driving conditions, such as night time driving or bad weather.

We agree the sample clip (link below) is much, much higher quality than the earlier examples of how the CGI clips would look.

That’s the good. Now for the bad.

Watch the clip. It runs through twice, with the second run having an explanatory voice-over.

The ‘developing hazard’ is pretty obvious when you’ve watched the clip to the end, but I found that the way the hazard develops makes it difficult to score the maximum score of 5. The demo clip has the candidate scoring the maximum 5 marks only if the mouse is clicked at the moment the vehicle comes into view, right on the very edge of the screen.

Haz Per Clip 04

Frankly, you’re out of that maximum score window almost before you can move your eyes to it, let alone click the mouse.

Even knowing it was coming, I found it difficult to click fast enough to get maximum points. Hopefully, as this is a demo clip, it can and will be sorted out.

But if this represents the final design, then there’s a problem. The vehicle appears on the far left edge of the screen but we need to shift our focus onto it, then monitor it for a short period in order to determine its speed and the potential for a collision with your own vehicle.

On the first pass, by the time I’d assessed the vehicle as being on collision course and being driven at a speed likely to be an issue – and clicked – I managed a fairly miserable 2 points – not enough for a pass score if that was my average result.

But even the definition of ‘developing hazard’ leads to confusion.

Look at the next still – now you have a white car approaching the main road from the left, and clearly in a position where it COULD pull out in front of us – the time and space we have to react is almost exactly the same as we have with the silver car.

Haz Per Clip 06

Look again at the silver car we passed a moment ago. Even though the silver car is approaching at the end of the road at speed, until it actually crosses the ‘Give Way’ marking, it’s still only a ‘potential hazard’ – it could still stop at the ‘Give Way’ line just like the white car does. We don’t actually NEED to take action until the silver car noses over the cycle lane, by which time we’ll only score two points.

Haz Per Clip 05

It gets worse. As the scoring starts BEFORE the silver car actually crosses the ‘Give Way’ line, then the only referent that differs between the two scenarios (bad, dangerous silver car driver and good, safe white car driver) is that the silver car is dangerous because it’s approaching at speed. Whilst that’s not a bad clue, the fact that the white car is NOT scored implies that the new driver has nothing to worry about; the white car can be safely dismissed as a hazard just because the driver has approached the junction slowly.

But as just mentioned, our speed and distance to the potential point of collision is about the same IF neither driver stops. What guarantee do we have that the driver of the white car WILL stop? Just because the driver’s approached the junction more cautiously doesn’t mean that he will not make a mistake and pull out – as a motorcyclist I know that only too well.

Either we treat both drivers and their vehicles as potential hazards UNTIL they stop, or we react ONLY to the silver car when it crosses the ‘Give Way’ line. Forget where the scoring happens for a moment, this is a huge logical inconsistency with the assessment of what constitutes the sort of problem a new driver needs to respond to, one that can only lead to more confusion between spotting a potential hazard (one that MAY force us to take action) and a developing hazard (one that we MUST take action to deal with).

So what’s the solution? By clicking on each and EVERY vehicle approaching down side streets, I could of course remove the need to assess speed and ‘time to collision’. But then I run the risk of clicking to a ‘pattern’, something that the briefing says will result in a zero score. So now I’m trying to ‘outguess the system’ by guessing which hazard will turn out to be the developing one without seeming to click to a pattern. Is that the intention? I very much doubt it.

But it gets worse. Far worse.

As a motorcyclist, there is a real issue with the way the DVSA clips ignore potential hazards.

Watch the video as it re-runs with the spoken explanation. At 1:18, a car appears in the side turning on the right. As there’s also a van approaching in the opposite lane, it’s clear that a moment later the car will be completely hidden by the van. Or to put it another way, the driver in that car won’t see us if he happens to look in our direction at that moment.

Haz Per Clip 01

This ‘eclipse’ duly happens at 1:20 and the car duly (and rather predictably) pulls out from behind the van, and straight into our path.

Haz Per Clip 02


…because the car pulls out just early enough and accelerates just quickly enough that the CGI driver doesn’t have to brake (there’s a carefully managed two seconds between the vehicles), this “doesn’t count as a developing hazard and you would not be scored on it”.

To my mind this logic is bizarre.

What if the driver emerging behind the van hadn’t accelerated so quickly? What if the driver had committed himself a second later, looking to his right rather than at us? What if he had glanced back after starting to pull out, spotted us then hit the brakes and blocked the road in front of us?

These are all very real possibilities and will instantly put us into the position of having to react. Sitting on my bike in real life, this would have grabbed at least as much of my attention as the silver car approaching the junction at speed or the white car waiting to turn at the junction just as I approached the potential point of conflict.

Ironically, there is a set of skid marks in the road suggesting precisely this scenario developed and someone had to brake hard to avoid a vehicle emerging from the right!

Haz Per Clip 03

Yet via the logic used in this clip the novice driver is encouraged to discount these very real POTENTIAL hazards because the situation didn’t actually go wrong.

It’s crazy. Ignoring this kind of potential hazard really creates a huge threat to a driver and more particularly to a biker.

There’s no doubt the quality is far better than the original filmed clips and in fact they are so realistic, they could easily be used for powering a driving or riding simulator, something we’ve said many times could make a huge difference to delivering driver training. The reason for that is that using a simulator could help novices make the mental link between the emergency responses like braking hard and swerving that have been practiced in safe areas, with the very real on-road hazard that needs that response.

But as it stands, the fact that this potential hazard can be included in the demo hazard perception clip AND be so lightly dismissed with a “don’t worry about it” messages demonstrates the failure of the hazard perception test to turn out new road users who can actually ‘perceive’. All it generates are reactive drivers.

The full clip can be found at:


One thought on “The failure of the DVSA’s hazard perception test to create perceptive drivers

  1. Pingback: The failure of the DVSA’s hazard perception test to create perceptive drivers | MIND over MOTORCYCLE

Comments are closed.