Reconsider Problem II-22. If you conclude that a directed verdict of acquittal is appropriate in the first instance in that problem but not the second, then consider the following.
C. NESSON, REASONABLE DOUBT AND PERMISSIVE INFERENCE: THE VALUE OF COMPLEXITY
92 Harv. L. Rev. 1187, 1194 (1979):
Why should it be that the high likelihood but starkly numerical case is thrown out of court while the cases based on self-serving testimony or additional circumstantial evidence will be put to the jury? The question becomes truly puzzling when one considers that even a case in which the quantifiable likelihood of guilt was much lower--for example, where originally only two prisoners were in the yard--might be allowed to go to the jury as long as the prosecutor's case was bolstered by additional circumstantial evidence or by other evidence distinguishing the defendant.
Why should evidence which generates a clearcut mathematical statement of the likelihood of guilt be considered insufficient, even when the probability of guilt is high, while other evidence of a testimonial or circumstantial nature is much more readily considered by the courts to be sufficient to sustain a prosecutor's case? Do we actually consider the jury to be accurate in assessing the credibility of witnesses and the strength of circumstantial evidence? Or are some risks of inaccurate verdicts more acceptable than others?
Is the problem of speculation limited to criminal cases? If the prison yard case were a civil action by the estate of the murdered guard against prisoner #1, would the analysis of issues be different?
In a civil case, is it useful to think in terms of odds or wagers when considering admissibility or sufficiency of evidence? Consider the following.
L. COHEN, THE PROBABLE AND THE PROVABLE §30 (1977)
The inapplicability of betting odds. Another commonly invoked criterion for the assignment of a mathematical probability is the acceptance or acceptability of appropriate betting odds within a coherent betting policy. So perhaps well-behaved jurors, who are fully self-conscious in their reasoning, should be supposed to measure the strength of their belief in the correctness of a particular verdict by reference to the odds they would accept if wagering on its correctness? This measure, the betting quotient, would be the ratio of the favourable figure to the sum of the two figures, and would have the structure of a mathematical probability. Odds of 4 to 1 on the plaintiff's case, say, would give a betting quotient or mathematical probability of .8. It would then apparently be possible for judges or legislators to stipulate a degree of mathematical probability that could be taken as putting the guilt of an accused person beyond reasonable doubt and a lower degree of mathematical probability that would suffice for the decision of civil cases.
But in fact such a procedure would be grossly fallacious. A reasonable man's betting practice is subject to two additional constraints, besides his knowledge of relevant data.
One constraint is that he only wagers on discoverable outcomes. Bets must be settlable. In each case the outcome must be knowable otherwise than from the data on which the odds themselves are based. When the horse-race is finally run, the winner is photographed as he passes the winning-post. When the football match is finally played, each goal is seen as it is kicked. Consequently wagers on past events or on the truth of scientific generalizations over an unbounded domain, or on any issue where the whole truth cannot be directly observed, are only intelligible in a context of total or partial ignorance about the relevant data. Knowing nothing, or only a little, about the local archaeological evidence I can wager you, on the basis of experience elsewhere, that there was no Roman settlement at Banbury; and to settle the bet we can excavate and see. But, if all the appropriate excavations have already been done, and we know their results, there is nothing to wager about. Similarly, to request a juryman to envisage a wager on a past event, when, ex hypothesi, he normally already knows all the relevant evidence that is likely to be readily obtainable, is to employ the concept of a wager in a context to which it is hardly appropriate. There is no time machine to take us back into the past. So there is no sure way of discovering whether the accused is guilty, or the plaintiff's statement of claim is true, except by looking at the relevant evidence. If one asks a juryman to envisage a wager in such a context, one is hardly entitled to expect a rational response.
Perhaps it will be objected that since people do sometimes wager sensibly about past events, as on the existence of a Roman settlement at Banbury, they can reasonably be assumed to have a general technique for assigning betting-quotient probabilities to past events on given evidence. Why then cannot a jury employ such a technique for the solution of its own special problems about probabilities? The answer to this objection is that it still does not succeed in attaching significance to the conception of judicial probabilities as betting quotients. The argument against a betting-quotient analysis of judicial probability is not that there is no sufficiently general technique for devising betting-quotients: the argument is rather that in certain situations the operational acceptance of a betting quotient is irrational. So talk about assigning probabilities, when probabilities are understood in the proposed way, involves the absurdity of talk about accepting reasonable bets on unsettlable issues.
Moreover, another constraint on rational betting practice has to be mentioned here. No one bothers about odds unless the amounts at stake are of some consequence. Only the very poor bet seriously in pennies. But when the amounts at stake begin to rise a prudent man tends to be more and more cautious about the odds he will accept for a given size of stake. Bookmakers shorten the odds on a horse not only when they hear of its latest wins elsewhere but also when they begin to be concerned about how much they would lose if it won the current race. So there is little sense in asking a man what odds he would accept on a certain outcome unless the value of the units wagered is also specified. Every juryman would have to be instructed separately by the judge, in accordance with the judge's estimate of what would be an appropriate sum of money for that juryman to envisage wagering. Consequently every accused would be at risk not only in relation to the evidence, but also in relation to the judge's estimates of how much importance each juryman attaches to the gain or loss of this or that sum of money. Such a system certainly does not yet exist anywhere, and its institution seems scarcely likely to promote the ends of justice.