There are few better ways to dismiss an argument, particularly in the criminal courts, than to brand it “theoretical”. Or worse, “hypothetical”. Or, almost worthy of a charge of professional misconduct, “metaphysical”. And it’s not hard to see why.
When a word is capable of a neat definition then the law can neatly define it. But when it’s not – when, as is often the case, a word’s meaning is as vague and inconsistent a mess as our everyday moral intuitions – then any attempt at a neat definition, any bold metaphysical adventure, is likely to end in tears. Safer to declare it a perfectly ordinary English word (God damn it!) while pounding the invisible gavel and reminding everyone of the inconceivably vast quantities of common sense possessed by juries.
So it’s always exciting when a particular combination of bad statutory drafting and gremloid facts leads the Court of Appeal to remove the safety latch from the floodgates, or at least peer through the fish-eye peephole at the metaphysical horrors beyond.
Just over a year ago Sir Brian Leveson gave judgment in R v Wilson  EWCA Crim 1184, a case of Causing Death by Driving Whilst Uninsured, and made some interesting observations on the connection between risk and causation. The case didn’t make headlines but will have a significant effect on how fatal driving cases are charged and summed up (as it did in two recent cases of mine). The effect will not usually operate in a defendant’s favour.
The Court of Appeal held that for a defendant to be criminally liable the bad driving must “significantly or materially increase the risk of death”. But beyond that, a “close comparative analysis of what would have happened with what did happen” is discouraged.
43. “In our judgment, section 3ZB is satisfied and the second limb of the test in Hughes is made out – “contributes in some more than minimal way to the death” – if the driving at the critical time was such as significantly or materially to increase the risk of death resulting from the Appellant’s culpable acts or omissions. Further than this, the purpose of the provision is not advanced by requiring a close comparative analysis of what would have happened with what did happen.”
The phrase “further than this” in that paragraph is, as they say, doing a lot of work… How can we determine the increase in risk (as the first sentence requires us to do) other than by comparing the risk of the bad driving to the risk without the bad driving – i.e. by performing a “close comparative analysis of what would have happened with what did happen”? Courts don’t like counterfactuals, even when recommending them.
(The trial judge went much further in his half-time ruling: “It does not matter whether or not [the deceased] would have lived or died at 30mph because the jury was not concerned with that theoretical scenario.” This is clearly wrong although the Court of Appeal does not criticise it directly. It’s perhaps worth noting that many a philosopher (e.g. David Lewis and, in places, David Hume) define causation in terms of counterfactuals.)
But what about this new “significantly increased risk of death” test? It works nicely on the facts in Wilson (an increase in risk of death from 7% to 31%) and in many other scenarios but I’m not sure it adequately captures our moral intuitions in all cases.
A young man rushes into the road without looking and gets hit by a speeding car. This is what happened in Wilson, and in my recent case, and I dare say in a fair proportion of other fatal driving trials (in 2018 pedestrians accounted for 25% of fatal road accident victims). Contributory negligence on the victim’s part may be relevant to sentencing, but not generally to the issue of causation. What matters is whether, given the actions of the victim, the bad driving caused the death – or more precisely whether it was “more than a minimal cause” of the death.
In our example, as in the majority of cases, the “but for” test is all we need to consider: questions about whether the cause was sufficiently proximate to warrant liability, and intervening acts breaking the chain of causation do not arise.
“But for the defendant’s bad driving, would the death have occurred?”
That’s the question. Note: it’s the bad driving, or the badness of the driving that needs to have caused the death, not the mere fact of the driving which may happen to have been bad.
Example: an overloaded vehicle spilling cargo onto the road behind it, travelling at the 60mph speed limit, hits a pedestrian who runs blindly into its path. But for the fact of the driving (which was in fact dangerous because of the unsecured cargo) death would not have occurred. But the defendant isn’t criminally liable because the “but for” test has to apply to the badness of the driving. The question would be, “But for the dangerously unsecured cargo, would death have occurred?” The answer is “Yes, death would still have occurred” – so the defendant would be Not Guilty.
In the example involving excess speed the question is whether but for the excess speed death would have occurred. And that can only be answered by a close analysis of what the effects of the collision would have been had the vehicle kept within the speed limit.
So, let’s suppose our pedestrian runs blindly onto a motorway thirty meters in front of a car travelling at 90mph. The driver immediately slams on the brakes but hits the pedestrian and kills him. An expert witness performs some calculations and tells the jury what would have happened if the car had been travelling at the 70mph speed limit. He then explains the counterfactual to the jury: if the pedestrian had run blindly onto the motorway thirty meters from the car but the car had been going at only 70mph, the car would have hit the pedestrian anyway; the speed at the point of impact would have been slower (say, 68mph instead of 88mph) but death would still have occurred. Therefore no causation, so Not Guilty.
If, by contrast, the car was travelling at 40mph in a 30mph zone, and the pedestrian ran into the road 30 meters in front of the car, the expert would (assuming necessary details) say that if the driver had been travelling at or below the speed limit then there would have been no collision at all. So the excess speed is what caused the death, so the defendant is Guilty.
And this fits with our intuitions doesn’t it? In case not – in case breaking distances and the path of a moving pedestrian obscure your intuitions – imagine an unusually bold protestor who tries to establish a one-man road-block by lying down in the road just after a bend, in an unrestricted speed limit zone on a foggy night.
He’s going to get killed whatever the speed of the next car round that corner. So if the next car is travelling too fast the driver should not be held criminally liable for the death, because his excess speed had nothing to do with the death.
But if the lone protester had been slightly less incautious and had blocked the road by sitting inside a stationary vehicle rather than lying down on the tarmac, and was killed by a speeding car in circumstances where death would not have occurred below the speed limit, our moral intuitions would want to find the driver guilty: it makes sense to say that it was the excess speed that caused the death.
But the examples above all compare the fatal collision that actually happened with one or other of the following:
- No collision
- A slower fatal collision
- A slower non-fatal collision
That might seem to cover all bases but of course the expert will often (or indeed in most cases) not be able to say for sure whether the hypothetical slower collision would have been fatal. Instead, he’ll only be able to assign death a probability.
In the Wilson case the defendant said he was travelling at about 40mph in a 30mph zone. The Agreed Facts were that at 40mph death was at least four times more likely than at 30mph (para 13). The probabilities given were 31% compared to 7%.
So, the Court of Appeal say that the test for causation (“contributes in a more than minimal way to the death”) is made out if the excess speed (or other bad driving) significantly increases the risk of death.
A first query might be this: should the significance of the increase in risk be assessed proportionally or in absolute terms? Common sense might suggest proportionally. Increasing the risk of death by five percentage points from 1% to 6% is a sixfold increase – 600%. Probably significant. But add the five percentage points on to 80% to get 85%, and that’s only a 6.25% increase. Less clearly significant.
But what if the contrasting probabilities are both very low indeed: 1% increased to 2% – like those food-scare stories about bacon doubling your risk of cancer, it might be a doubling but it doesn’t feel very significant when the risk remains tiny.
I suppose jurors can be left to answer the question as best they can and barristers can try to persuade them of the interpretation of “significant increase” most favourable to their client. But I can see certain lines of argument being unfairly proscribed by judges anxious to give an uncomplicated summing-up.
And, in any event, does a significant increase in risk of death necessarily mean that there has been a more than minimal contribution to death? If the risk increases from 75% to 85% then even if it was accepted that those ten percentage points (or a proportional increase of 13.3%) amounted to a “significant” increase, is it really accurate to say that the driver caused the death when the deceased was (very) likely to have died anyway? I feel that a defendant should be allowed to try to persuade a jury that it isn’t. Consider the following scenario.
A group of mafiosi hold a gun to their victim’s head and play a grim and elaborate form of Russian Roulette. On a table in front of him there are ten upturned opaque beakers, and a large jar of lottery balls numbered 1 to 100. The victim is told that while he is blindoflded ten balls will be randomly selected from the jar and one placed under each cup. The victim must then select a cup. If the number on the ball beneath the cup that he selects is over 50, he will be shot and killed, otherwise he will be let go. The balls are selected and placed under the cups. Let’s suppose that, as it turns out, six of them are above 50. While this is going on, I (secretly and without causing any delay) replace the number 9 ball that’s under one of the cups with the (as yet unselected) 99 ball. Now seven of the balls have numbers above 50. The victim chooses a cup. The gunman sees the number of the ball beneath it and – following the rules – shoots and kills the victim. Did I cause his death?
Surely only if he actually picked the 99 ball. I increased his risk of death, certainly, but was my act an operating cause? But for my adding of the 99 ball, would he have been shot? If he in fact picked the 99 ball, then no: he would not have been shot but for my ball swap. But if he in fact picked a ball over 50 but not the 99 ball then it can be said: he would have been shot anyway. In which case I didn’t cause his death. My swapping in of the 99 ball had nothing to do with his death.
On one view, the important difference between this and the equivalent hypothetical fatal driving scenario is that in the latter we don’t know ‘which number was picked’. That is, we don’t know whether the extra ten percent death risk is what actually made the difference.
Well, let’s suppose that after the balls are placed beneath the cups and the risk-increasing swap is made, the balls are (automatically) reshuffled. And, to be absolutely safe from worries about intervening acts let’s suppose the ball number recognition and consequent shooting is an entirely automated process, and that the table is then (automatically) upturned. That is, we never know, and there can never be any way of knowing, which ball the victim picked. We just know it was a ball over 50 because he was shot, but that’s all.
So: what’s the probability that I caused the death? Answer: 1 in 7, or c.14%. I would not like to send someone to prison on the basis of that percentage.
All the above considerations pre-suppose that an expert can fairly calculate probabilities of death for counterfactual, and actual, scenarios. Most starkly of course an expert needs to be able to first work out whether a collision would even have occurred without the excess speed. What set of alternative facts should the expert use? The Court said in Wilson:
42. We have not embarked on a reasoning process which entails envisaging exactly where the Appellant’s vehicle would have been on the highway had he been travelling at the speed limit. On one analysis, it could be said that he would have been further back in the road. However, such reasoning would be close to engaging in an exercise in metaphysics.
These problems are serious and difficult and are often not considered sufficiently carefully – or consistently – by experts and lawyers. In my mafia lottery murder example I had to go to some lengths to keep the ball-swapping causally insulated from the victim’s cup selection. It’s not easy to do in real-life situations.
When trying to work out whether a driver would have hit a pedestrian had he not been speeding, which circumstances (timings, speeds, reactions, etc.) should we keep constant and which should we change? If in the counterfactual scenario the pedestrian runs into the road at the exact same time of day that he did in reality, and the driver instead of speeding had been travelling at the speed limit for some appreciable time prior to arriving at the location of the collision, it will almost always be correct to say that there would not have been a collision at all. (This would of course operate to a defendant’s disadvantage). Instead, a defendant should surely be allowed, if it helps him, to have the counterfactual analysed on the basis that the pedestrian ran into the road when the driver was first able to see him as he rounded the corner. Or should that be ‘when the driver first did see him’? And should, for example, a defendant’s unusually swift braking reactions operate against him in the analysis of the counterfactual? I may explore these issues in a future post.