The most important assignments of error, both in the admission of evidence and in charging the jury as to its effect, relate to the admission, in the State's case in chief, of evidence to prove that the accused committed three other murders by arsenic poisoning, two before and one after the date of the crime for which she was on trial.
The argument against admitting evidence of other similar but unconnected crimes is not that it has no probative value. As Wigmore said: "It is objectionable, not because it has no appreciable probative value, but because it has too much" (§ 194); meaning, of course, that its appeal is not confined to the intellect or to the precise issue. Such evidence, when offered in chief, violates the rule of policy which forbids the State initially to attack the character of the accused, and also the rule of policy that bad character may not be proved by particular acts. Wigmore, § 57. These two rules of policy are firmly established, and they mark one important difference between the Anglo-American criminal procedure and the French. On the other hand, evidence of crimes so connected with the principal crime by circumstance, motive, design, or *Page 531 innate peculiarity, that the commission of the collateral crime tends directly to prove the commission of the principal crime, or the existence of any essential element of the principal crime, is admissible. The rules of policy have no application whatever to evidence of any crime which directly tends to prove that the accused is guilty of the specific offense for which he is on trial. The objection on the ground of policy applies exclusively to evidence of crimes which are logically unconnected with the principal crime. That is to say, to evidence the probative effect of which is indirect, in the sense that its direct application is exhausted in showing that the accused was possessed of a bad character or of a disposition to commit the particular crime of which he is accused, and thereby it furnishes a justification for a conviction rather than proof of guilt of the specific offense. The general rule upon the subject is well stated by Cushing, C. J., in State v. Lapage,57 N. H. 245, 289; "I think we may state the law in the following propositions: (1) It is not permitted to the prosecution to attack the character of the prisoner, unless he first puts that in issue by offering evidence of his good character. (2) It is not permitted to show the defendant's bad character by showing particular acts. (3) It is not permitted to show in the prisoner a tendency or disposition to commit the crime with which he is charged. (4) It is not permitted to give in evidence other crimes of the prisoner, unless they are so connected by circumstances with the particular crime in issue as that the proof of one fact with its circumstances has some bearing on the issue on trial other than such as is expressed in the foregoing three propositions."
This record does not present the case of relevant or connected crimes. No claim was made that the evidence as to the Gilligan, Smith and Gowdy murders, was properly admissible for the purpose of directly *Page 532 proving the commission of the principal murder, or the existence of any essential element thereof. It was claimed and admitted upon an entirely different theory, under the rule which in certain cases permits evidence of otherwise irrelevant similar acts by the accused for the limited purpose of eliminating innocent intent in the commission of the principal act, and to exclude the probability of accident and mistake.
The theory upon which evidence of other similar but unconnected acts of the accused is admitted for these purposes, involves tow antecedent hypotheses: that the particular act in question is equivocal, in the sense that its commission by the accused is consistent with an innocent as well as with a criminal intent, and that the accused did commit the act. These hypotheses being assumed, the argument, as Wigmore (§ 302) states it, "is purely from the point of view of the doctrine of chances, — the instinctive recognition of that logical process which eliminates the element of innocent intent by multiplying instances of the same result until it is perceived that this element cannot explain them all." As applied to cases of murder by poisoning, the argument is this: it is possible that one may administer a fatal dose of arsenic to one person by mistake or accident, but that he should administer arsenic to four persons with fatal result in one year by mistake is in the highest degree improbable. The administration of arsenic in fatal doses with innocent intent is abnormal; and the multiplication of similar acts tends to exclude this abnormal element in each case. It will thus be seen that the criminality of the other or collateral administrations of arsenic has nothing to do with the argument. In this case, for example, the evidence of the Gilligan, Smith and Gowdy poisonings, if admissible at all, was admissible solely for the purpose of reducing the probability that the accused could *Page 533 have administered the arsenic to Andrews accidentally — assuming that she did administer it, — by showing that she had done the same thing with the same fatal result four times in the same year. It is plain enough that evidence of the repetition of equivocal acts, with their results, does not necessarily violate any rule of policy. It is also plain enough that when the guilty intent necessarily accompanies the act, evidence of other unconnected crimes is as unnecessary as it is objectionable. But when, as in this case, the principal act is claimed to be equivocal and the commission of similar unconnected acts cannot be shown without at the same time showing that they were done with criminal intent, a conflict may arise between the general rule of policy and the special rule, for in such cases the legitimate argument "purely from the . . . doctrine of chances" is likely to be overshadowed by the inadmissible argument that because the accused is possessed of a disposition to murder by arsenic, she probably administered the arsenic to Andrews — assuming that she did administer it — with murderous intent.
The authorities on the subject are so numerous, and the relation between the commission of one offense and of another similar offense depends so much upon the nature of the offense and on the circumstances of each case, that we confine our discussion to the crime of murder by poisoning, except for a brief reference to two cases already decided by this court.
In State v. Ward, 49 Conn. 429, 440, which was a prosecution for receiving stolen goods knowing them to be stolen, evidence of the prior receipt by the accused of other goods from the same thief, knowing them to be stolen, was admitted to show guilty knowledge; and in State v. Raymond, 24 Conn. 204, we held that, on a prosecution for keeping intoxicating liquors with intent to sell, evidence of prior sales at the same place was *Page 534 admissible to show the intent with which the liquor was kept at that place. These cases evidently belong to the class of connected crimes. In the Ward case the evidence was not admitted for the purpose of multiplying unconnected instances of the receipt of stolen goods until the probability of their receipt with innocent intent was overwhelmed, but for the direct purpose of showing that the accused knew that he was dealing with a thief in the specific case. And in the Raymond case the continuing nature of the offense charged, coupled with the presumption of a continuance of the intent to sell manifested by the prior offense, afforded equally direct evidence of guilty intent in the specific case.
Returning to the subject of unconnected murders by poisoning, there is a conflict of authority upon the point which we regard as decisive of the present appeal, viz: whether in such case evidence in chief of other similar poisonings by the accused is always admissible, or whether it is admissible only when the administration of the poison is under the circumstances of the particular case reasonably consistent with a possible defense of accident or mistake.
In People v. Molineux, 168 N.Y. 264, 61 N.E. 286, which is the leading case in support of the latter doctrine, the New York Court of Appeals reversed a conviction of murder by poisoning, because the trial court admitted evidence of another similar, unconnected murder by poisoning by the accused, for the purpose of disproving innocent intent; saying that on the facts, as the State claimed to have proved them, no reasonable doubt could exist that the accused — if he committed the act in question — must have done it with criminal intent. The reasoning of the court is that the proof of unconnected crimes is contrary to the general policy of the law, and therefore inadmissible unless the hypothesis is first established on which the special *Page 535 exception is founded, viz, that the particular act in question is equivocal. Three justices dissented from the result in that case, on the ground that the two crimes were logically connected because the poison used was so rare and the mode of its administration so peculiar, that the commission of the collateral crime by the accused tended directly to show that he committed the principal crime also.
The leading case in this country in support of the other doctrine, that in poisoning cases the commission of similar unconnected poisonings may always be shown to eliminate accident or mistake, is State v.Hyde, 234 Mo. 200, 232, 136 S.W. 316. In that case a new trial was ordered, but the court proceeded to discuss the theory upon which evidence of other unconnected poisonings by the accused was admissible, saying: "It is true that the evidence in the record raises no question of intent. It may also be conceded that if the defendant knowingly administered deadly poison, the intent may be inferred from the act, and no further proof is required. But it must be remembered that in a criminal case the defendant files no written pleading. . . . It is a part of the State's case to show criminal intent, and in so doing to negative accident or mistake, by any proof competent for that purpose. The State was not required to rest upon the assumption that the proof would show that the intent accompanied the act, nor to hold back, for rebuttal, evidence showing intent." And in the course of its opinion the court quoted from Trogdon v. Commonwealth, 31 Gratt. (72 Va.) 862, 873 (a case of obtaining money under false pretenses), where the court observed that it could not tell what evidence of intent would satisfy the jury beyond a reasonable doubt, and put the question, "What would be thought of a judge who would thus prejudge the case and invade the province of the jury?" *Page 536
We think this argument ignores the essential fact that the underlying question which confronts the court is not simply whether the State has sufficiently proved criminal intent, but whether the particular evidence offered upon that issue is admissible. Courts are not infrequently required in criminal cases to pass upon preliminary questions of fact in order to determine the admissibility of evidence, and no doubt courts are vested with considerable discretionary powers in passing upon such preliminary questions. In this case the question is whether evidence, generally objectionable, shall be admitted for the limited purpose of eliminating accident or mistake, and we think it would be an abuse of discretion to permit proof of similar but unconnected poisonings in a case where the State's evidence had already gone so far toward eliminating accident or mistake as to leave no reasonable doubt, in the absence of rebutting evidence, that the poison, if administered by the accused, must have been knowingly administered. Otherwise evidence inadmissible on the general issue would be admitted for the special purpose of characterizing an equivocal act, when the act in question was not equivocal, and so the only practical effect which the evidence could have would be to prejudice the accused and violate the policy of the criminal law. In such cases the State should exhaust its other evidence of criminal intent before resorting to proof of other unconnected acts for that purpose, and it should not resort to such proof until it has also exhausted its evidence connecting the accused with the death. Then if evidence of other similar but otherwise irrelevant poisonings is offered for the limited purpose of reducing the probability of accident or mistake, it is the duty of the court to determine whether the administration of the poison, if done by the accused under the circumstances claimed by the State, is, in the absence *Page 537 of rebutting evidence, reasonably consistent with innocence, accident or mistake. If not, the evidence must be excluded without prejudice to its possible admission in rebuttal.
Applying these conclusions to this record, it is apparent that a large part of the circumstantial evidence which the State offered for the purpose of connecting the accused with the death of Andrews, necessarily tended also to negative the probability of accident or mistake. As stated upon the defendant's brief, this evidence included (1) secrecy in regard to the $500 loan matter; (2) plans for the reception of the Gowdys, with promises and lying statements made to them; (3) the purchase of arsenic four days before Andrews' death, with a false reason assigned for its purchase; (4) the fact that Andrews received at least two doses; (5) delay in calling the doctor the first time and again in calling him the second time; (6) false statements as to the nature of Andrews' illness to the doctor and others; (7) unreasonable failure to notify Andrews' relatives; and (8) the unseemly haste in getting rid of the body. We think it is evidently impossible that the jury should have accepted the State's claims in these particulars as true, without necessarily eliminating from their own minds any possibility that the accused administered the poison by accident or mistake. The single fact that the principal expert for the State gave it as his opinion that the deceased had received two doses of arsenic, one of them a very large dose administered shortly before death, is extremely significant upon this issue. Without going into any of the details, it would seem that the distribution of diffused arsenic in the body of Andrews and the quantity of undissolved crystals of arsenious acid found in his stomach, could hardly be explained upon any other hypothesis than that two doses, at least, were administered, *Page 538 and at or about the times suggested by the State's expert. If this is true, and if the poison was administered by the accused, it puts an end at once to any reasonable possibility of accident or mistake.
In this case the evidence in question was also admitted for the purpose of excluding suicide. For this purpose we think it was irrelevant and inadmissible. Suicide is a voluntary act, and it is impossible to see how the fact — if it were so — that the accused murdered three other persons by arsenic, one of whom survived Andrews, could increase or diminish any inclination toward suicide which Andrews might be supposed to have; especially as there is no evidence that Andrews suspected that they were murdered. The evidence indirectly tends to negative suicide, by showing that the accused was possessed of a disposition to murder by poison; but for that purpose it was inadmissible. Wigmore, § 363 (2A), appears to justify the admissibility of other instances of death by poison under similar circumstances to negative suicide, but the principle of anonymous intent to which he refers (§ 303) is more appropriately applied to other attempts on the life of the deceased.
The conclusions reached cover the evidence as to financial transactions between the accused and Smith and Gilligan, which was admitted for the purpose of showing that the accused had a motive for poisoning them, and also so much of the evidence of Mason as relates to purchases of arsenic connected with other poisonings and not connected with the poisoning of Andrews.
The charge of the court in so far as it is based upon the evidence as to other poisonings is erroneous, because inapplicable to the present case. That part of the charge which instructs the jury that evidence of other poisonings might be considered in determining *Page 539 whether the deceased committed suicide was also erroneous. In the other portions of the charge excepted to we find no error.
In view of the result it is unnecessary to pass upon the motion to correct the finding.
There is error and a new trial is ordered.
In this opinion PRENTICE, C. J., and SHUMWAY, J., concurred; RORABACK and WHEELER, Js., concurred in the result, but not in the opinion.