Dear Colleagues:

Due to a technical issue, the following exchange between Jane Stapleton and Rob Stevens failed to post to the ODG. As it is very interesting, I repost it below: 

Jane Stapleton:  Rob says “she would have had an operation on a different day running
precisely the same risk anyway” but fails to make the correct deduction.
We need to keep separate two issues.

[1] That the failure to warn was a cause of the injury…
Remember that in Chester v Afshar the court accepted that the risk was
entirely random, like the chances at roulette (1 in 38).

Now, suppose one week a teacher brings his roulette wheel to school, and
says each lunchtime the students may play on it one by one: a student
would throw his marble on and see what number it ends up on. Unknown to
the students, the teacher has a mental quirk that means that if a
student’s marble ends on the number ‘30’ the teacher will impose a
detention on that student. During Monday lunchtime many students have
their turn and enjoy the game very much. But when Student R throws his
marble it lands on ‘30’ and the teacher imposes a detention on him. When
R is asked what he would have done had he known the detention risk
associated with playing he says he would have hesitated and not played
on Monday but that he was sure by Tuesday he would have overcome his
anxiety and played.

When we ask whether the teacher’s failure to warn of the detention risk
was a cause of R’s Monday detention we are comparing what happened in
the actual world (a failure to warn and an actual detention on Monday)
with what would have happened in the hypothetical world in which a
warning had been given. Since the latter is a hypothetical world the
best we have to go on are the inherent chances associated with the
transaction at stake. In this hypothetical world in which B had been
warned, he would not have played on Monday but would have played on
Tuesday.

It is true that on Tuesday the detention risk (1 in 38) is “precisely
the same” as it had been on Monday, but, ironically for Rob, this is
exactly what dictates the conclusion that on Tuesday it is likely R’s
marble would not have landed on ‘30’ and that R would not have suffered
a detention. [Since the chances are random, the fact that his marble
happened to have landed on ‘30’ on Monday is irrelevant to where it
WOULD, to the best of our estimation, HAVE LANDED on Tuesday.]

We express this contrast between the actual world (where R suffers a
detention) and the hypothetical world where the warning was given, by
saying the failure to warn was a “cause” of the actual injury (i.e. the
detention).

[2] That the injury was not a coincidental consequence of the failure to
warn…
Whereas the notion of factual cause is a hindsight judgment comparing
what happened to an individual in the actual world with what would have
happened to him in a hypothetical world, the notion of coincidence is a
relation between TYPES of conduct (or other factor) and TYPES of
outcome.

First, if we start with Rob’s strange definition …“A loss is
coincidental if, as things turn out, its risk was not increased FOR THIS
CLAIMANT by the wrong which occurred.” (Torts and Rights 165)…we must
note that it fails to identify the most widely acknowledged illustration
of a coincidental consequence: the speeding streetcar case of Berry v.
Sugar Notch Borough, 43 A. 240 (Pa. 1899) where, had the streetcar not
been speeding, it would not have passed a tree just as it collapsed on
the vehicle injuring someone aboard.
For the victim-claimant, the speeding DID increase his risk - from zero
to 100% - so Rob’s definition of a coincidence is not met.
But speeding in general DOES NOT increase the probability of the TYPE of
accident (trees falling on the speeding vehicle) so the orthodox
definition is satisfied.

May I venture to suggest the source of Rob’s confusion? I think it stems
from the abbreviated way the notion of coincidence is often expressed in
the US: e.g. “When tortious conduct does not generally increase the risk
of the TYPE of harm that occurred, the wrongful aspect of the actor’s
conduct is merely serendipitous or coincidental in causiRestatement Third of Torts (2010) s. 30, cmt a.

What is meant is the GENERAL TYPE of breach. I have tried to be more
explicit in my writings where I define: harm as a "coincidental"
consequence of the tortious aspect of the actor's conduct when the
GENERAL TYPE of tortious conduct engaged in by the actor (e.g.,
speeding) does not generally increase the risk of the TYPE OF HARM
suffered (e.g., a tree falling onto the speeding vehicle), see 44 Wake
Forest L. Rev. 1309 at 1327.

In other words, in determining whether a consequence of breach is a
coincidental consequence, the issue is not whether this particular
breach increased the risk for this particular claimant: this will nearly
always be the case because, as in the streetcar example, it will usually
be that, but-for the breach, the harm would not have occurred so that
the particular breach in issue will have increased the risk to the
individual claimant from 0 to 100%. Rather, whether a consequence is
coincidental turns on the general propensity (or not) of this type of
breach to produce this type of result.

In Chester and my roulette example, the type of breach (i.e. failure to
warn about the risk inherent in a transaction) DOES generally increase
the risk of the type of harm suffered relative to a world where the
warnings were given: the incidence of the injury is higher in a world
where patients are unwarned. When we say in an individual case that the
consequence of breach is not a coincidental consequence we are capturing
this general feature of the type of breach and the type of injury.

Rob Stevens: The complex hypothetical under (1) is a re-run of Chester and simply shows
that "but for" (or NESS if we really must) is satisfied. This is true in
Chester v Afshar and your example. Everyone accepts that. Indeed counsel
for the defence conceded it in Chester (I have the written submissions).
It is a distraction.

On (2) you say  "Whereas the notion of factual cause is a hindsight
judgment comparing what happened to an individual in the actual world with
what would have happened to him in a hypothetical world, the notion of
coincidence is a relation between TYPES of conduct (or other factor) and
TYPES of outcome."

This is false contrast because we need to know the outcome, in order to
know what question we are asking, and we can only know that with
hindsight. It is a hindsight issue to this extent: we need to know what
specifically happened in order to judge whether the risk of that event was
increased by this wrong. We must compare the world where the wrong
occurred with a hypothetical world where it did not and ask whether the
risk of this loss was increased for this claimant by this wrong.

On Berry v Sugar Notch Jane writes
"For the victim-claimant, the speeding DID increase his risk - from zero
to 100% - so Robert's definition of a coincidence is not met.
 But speeding in general DOES NOT increase the probability of the TYPE of
accident (trees falling on the speeding vehicle) so the orthodox
 definition is satisfied."

This is false. We have to take the specific event which with hindsight
happened, not just its general type. Did the speeding increase the risk of
this loss (with hindight) "for this wrong for this claimant"? The answer
is obviously no, in retrospect knowing in hindsight when and how it
happened. Was the probability of this loss increased for this claimant by
this wrong? No, because the probability of being by hit by trees is not
increased by speeding.

Focusing on “type” is too loose. The issue would then turn upon the degree
of generality of “type”. Was the risk of "serious injury" increased by the
speeding in Berry v Sugar Noch? The answer is yes. Why is the "type"
"serious injury from a tree fall" and not simply "serious injury"? We only
know that it is the former with hindsight.

Now, the Shavell definition doesn’t give the wrong answer in Berry if we
define the  loss as “serious injury from tree fall” and not simply
“serious injury”.

However, we can change the story to keep the loss coincidental but
unequivocally of a general “type” the risk of which is increased by the
defendant's wrong doing ex ante but, was a loss which with hindight did
not actually result in this actual loss. That is the House of Lords
decision in Carslogie. I am somewhat surprised that this, our leading
case, has not been cited or considered by Jane. It wasn’t adequately
considered by the House of Lords either, and I am critical of them for
that in my piece on the history of torts.

Was the risk of further storm damage increased by the initial wrong? Yes.
Is the 'but for' (or NESS) test satisfied? Yes. The 'Shavell' test give
the wrong answer in Carslogie and indicates that the further loss should
be recoverable. Ships which are smashed into *generally* run a higher risk
of further storm damage. But that is not enough to make this loss suffered
by this claimant anything other than coincidental.

Just looking at the simplified facts of Carslogie I give in my previous
reply, do you really think that the storm damage should be recoverable?
Storms are not unusual or unforseeable events at sea.

Chester v Afshar is I think the same as Carslogie. Persons *generally* who
are not warned will run a greater risk, just as ships with gaping holes
generally run a greater risk of storm damage. But comparing the world with
and without the wrong, there was no greater risk for anyone who would have
gone ahead anyway, including those such as the plaintiff who would have
delayed the operation. She would have run an identical risk on another
day. When we compare the world where the wrong occurred with a
hypothetical world where it did not she would have run the same risk of
this loss. She ran a higher risk of being injured on the particular day
that she was it is true, but the "type" of injury she suffered was not a
"Tuesday injury".

The Shavell definition only makes sense from a deterrence perspective, and
now that I have read it it is clear that that is the perspective he
adopts. We can forgive him this because he is an economist. He thinks the
law of torts is about deterrence. Which it isn’t. His view simply doesn’t
represent the actual law, something we commonly find with economists.
Carslogie is wrong on his (and Jane’s) definition.

Shavell’s definition commits Jane to the counter-intuitive view that in a
medical non-disclosure case, such as Chester, injuries suffered as a
result of all the other inherent risks associated with the
procedure in question – such as contracting MRSA, or an accident of
anaesthesia – are not coincidental consequences either.  It follows that
her approach to coincidental consequences is out of line with the majority
judges in both Chappel v Hart and Chester, since they thought that these
kinds of consequences were coincidental (this seems to be the import of
the analysis of both Gummow J and Kirby J in Chappel at [at [66]-[67] and
[96] respectively, and of Lord Walker in Chester at [94]).

All of the majority of the House of Lords thought they were departing from
orthodoxy for policy reasons. All members of the High Court of Australia
thought that it mattered that the plaintiff would have run a lower risk
operation if warned. Jane’s view is inconsistent with the reasoning of the
majorities and minorities in both cases.

That an American economics Professor adopts a definition of ‘coincidence’
doesn’t make it orthodox. This is especially so when it is inconsistent
with the result in our leading ultimate appellate court authority, and
with the reasoning of all of the judges in later cases.

Why do I think Jane is making this error? I think it is because she is
committed to the view that the only true sense of ‘causation’ is the NESS
test (which is in my view and Richard Wright’s a definition of
sufficiency). All other ‘scope of liability’ issues would then be
normative.

Coincidental losses falsify this view. Although NESS is satisfied, we
don’t think such losses are ‘caused’ by the wrong and no normative
question ever arises. Jane responds by saying that in England (but not the
US) coincidental loss is recoverable where the tort is deceit. However,
the English case she cites for this proposition doesn’t stand for it, nor
does anything said in it. I don’t think coincidental losses are ever
recoverable, and no normative issue ever arises, disproving the idea that
‘NESS’ (or indeed “but for”) is all we ever mean by “causation”. If event
X is coincidental to Y, Y did not cause X.

I think NESS is false as a test of ‘factual’ causation too, but that is a
separate matter.

Chester v Afshar is a good hypothetical to try out on dinner party guests.
I have never met anyone who thinks she should recover for her injuries.
That, however, may say more about my dinner party guests than the law.

RS: Another simpler way of making the point.

We are in a casino and you are choosing which number to bet on at
roulette. I advise you to place your money on number 22 on the next spin
of the wheel. You choose to follow my advice, choose this number which you
would otherwise not have done, and as luck would have it number 22 comes
up.

Was the winning coincidental to my advice?

Does this "general type" of conduct ("bet on 22") increase the general
chance of this outcome (someone about to place a bet winning?). No, says
the 'Shavell' test.

However, it is wrong to think that these winnings are coincidental as this
test would indicate. The chances of the specific win which actually
occurred was increased by the advice.

We could define the "type" of outcome much more specifically as "winning
on number 22" instead of "winning". But what if you are the only person in
the world who would have followed my advice? Would the win have been
coincidental because the chances of winning on 22 would not have generally
been increased for persons about to bet by my advice? Again, the answer is
no. Because the chances of this win by this person were increased by the
advice given it is not coincidental.

As I said intially, I think defining "type" suficiently precisely simply
collapses into my approach.