(dissenting) :
I respectfully dissent from the majority opinion as it relates to evidence of incidents concerning other children, and especially as it holds such evidence may be considered proof of the corpus delicti (which may include the criminal agency of the accused). I agree with part III of the opinion concerning the sentence.
While I do not agree with the majority that the evidence of prior occurrences was admissible under either the signature exception or to show lack of accident, since the majority gives no reason for its holding, the reason for my dissent will not be further mentioned here. I treat the same my disagreement with the conclusion of the majority that the effect of such evidence, even if properly admitted, should not be limited by a proper instruction. I would find it unnecessary to consider the application of Ashe v. Swenson, supra, for the evidence concerning Judy should have been stricken from consideration for the same reasons the evidence concerning the other children should not have been admitted.
I am convinced that the defendant here did not receive a fair trial because the evidence of prior occurrences, fitting no recognized exception to the general rule, was so highly inflammatory and prejudicial, as well as being neither plain, nor clear, nor convincing, that it should not have been admitted for any purpose, much less for all purposes.
Through the years, the courts of this circuit have adhered to the general rule that evidence of crimes not alleged against the defendant within the indictment may not be adduced at trial as proof of the defendant’s guilt. E. g., Benton v. United States, 233 F.2d 491 (4th Cir. 1956); Lovely v. United States, 169 F.2d 386 (4th Cir. 1948); Simpkins v. United States, 78 F.2d 594 (4th Cir. 1935).
The Supreme Court has adhered to the same rule which I consider binds us not only by way of precedent, but also as a matter of sound policy supported by reason and justice. Mr. Justice Jackson articulated the underlying considerations which support this accepted rule of criminal evidence in Michelson v. United States, 335 U.S. 469, 69 S.Ct. 213, 93 L.Ed. 168 (1948):
“Courts that follow the common-law tradition almost unanimously have come to disallow resort by the prosecution to any kind of evidence of a defendant’s evil character to establish a probability of his guilt. The State may not show defendant’s prior trouble with the law, specific *140criminal acts, or ill fame among his neighbors, even though such facts might logically be persuasive that he is by propensity a probable perpetrator of the crime. The inquiry is not rejected because character is irrelevant; on the contrary, it is said to weigh too much with the jury and to so overpersuade them as to prejudge one with a bad general record and deny him a fair opportunity to defend against a particular charge. The overriding policy of excluding such evidence, despite its admitted probative value, is the practical experience that its disallowance tends to prevent confusion of issues, unfair surprise and undue prejudice.” 335 U.S. 469, 475-476, 69 S.Ct. 213, 218, 93 L.Ed. 168.
In accord are Hall v. United States, 150 U.S. 76, 14 S.Ct. 22, 37 L.Ed. 1003 (1893) and Boyd v. United States, 142 U.S. 450, 12 S.Ct. 292, 35 L.Ed. 1077 (1892). As we previously have stated, this is not merely a technical rule of law; rather, it reflects the “fundamental demand for justice and fairness which lies at the basis of our jurisprudence.” Lovely v. United States, 169 F. 2d 386, 389 (4th Cir. 1948).
In applying the rule excluding evidence of a defendant’s prior crimes or acts, we have recognized a number of circumscribed exceptions which may admit evidence introduced for some relevant purpose other than a showing that the defendant probably committed the crime. Accordingly, evidence of prior acts or crimes has been permitted for the limited purposes of proving knowledge, intent and motive.1 Also, in appropriate circumstances, courts have admitted prior-crime evidence for several additional purposes. A case by case analysis of these decisions might yield the following exceptions: (1) to complete the story of the crime; (2) to show the existence of a larger continuing plan, scheme, or conspiracy; (3) to show that other crimes by the accused are so nearly identical as to earmark them as the handiwork of the accused; (4) to show a propensity toward illicit sexual relations with the particular person concerned; (5) to show that the act for which the accused is on trial was not inadvertent, accidental, unintentional or without knowledge; (6) to establish motive; (7) to show malice, deliberation or ill will; (8) to prove identity; (9) to show admissions by conduct tending to obstruct justice or to avoid punishment for the present crime; (10) to impeach the accused after he testifies on his own behalf. See McCormick on Evidence § 190, at 448-51 (2d ed. 1972).2
Even when the trial court determines that evidence of a prior act or crime is admissible as coming within one of the exceptions to the general rule, it still must determine whether the proof of the prior act is of a sufficient quality to permit its admissibility, and the courts have held, with considerable uniformity, that in cases falling under exceptions to the rule it is essential that proof of the prior act or offense be plain, clear and conclusive. Evidence concerning prior acts of a vague or uncertain character is not admissible. Kraft v. United States, 238 F.2d 794, 802 (8th Cir. 1956).3 Consequently, it should be apparent that *141evidence of prior merely suspicious occurrences is no substitute for clear and conclusive proof.
Yet, assuming that the proffered evidence of prior acts fits a given exception and is proven by clear and convincing evidence, considerable reputable authority maintains the court’s inquiry as to admissibility is not even then completed. Careful consideration still should be given to the prejudicial potential of the evidence of other acts or crimes on the minds of the jurors. As Judge Sobeloff cautioned in his dissent in United States v. Baldivid, 465 F.2d 1277, 1283 (4th Cir. 1972), “the danger of undue prejudice is not cured merely because the evidence falls within an exception to the rule; the prejudice'remains inherent in the testimony.” 4
In applying a balancing test, the considerations confronting the district court may be summarized succinctly: assuming an exception to the rule is applicable, before admitting evidence of prior acts or crimes, the court should determine, first, whether the evidence is relevant; secondly, whether the evidence is unduly prejudicial notwithstanding its relevancy; and, thirdly, having met both prior tests, whether the evidence is truly needed by the prosecution.5
*142Viewed in any light, the testimony offered by the government of prior incidents involving children other than Paul D. Woods was improperly admitted. Not only was it inadmissible to establish the corpus delicti, it also failed to meet the plain, clear and conclusive test as to the quality of proof required. The resulting prejudicial effect upon the jury so outweighed the probative value of the evidence that its admission was reversible error. I would grant a new trial.
The ancient doctrine of corpus delicti was developed to prevent conviction for a crime never committed. See 4 Blackstone, Commentaries, at 358-59 (Lewis ed. 1962). As one commentator has noted, “the execution of an innocent man is equally tragic no matter what caused justice to miscarry in his case, but it seems most shocking where the crime itself is found to have been nonexistent.6 Accordingly, the common law adopted the two elements of corpus delicti in homicide eases and insisted that proof be made not only of the fact of the death of the victim, but also that death was occasioned by criminal means.7
Federal courts have recognized and applied this definition of corpus delicti in murder prosecutions, requiring that both elements be proved beyond a reasonable doubt.8 In Evans v. United States, 122 F.2d 461, 465 (10th Cir. 1941), the court defined the scope of the doctrine in the following terms:
“The corpus delicti, as relates to homicide, is composed of two elements: (a) The death of the person alleged to have been killed; (b) that some criminal agency caused such death. . .
Of course, both of these elements must be established beyond a reasonable doubt. When the jury find these established, the next inquiry is as to the identity of the criminal agency; and this, too, must be established beyond a reasonable doubt, but is properly no part of the corpus delicti.”9
In the instant case, the government attempted to establish the corpus delicti and the criminal agency of the accused for the alleged murder of Paul Woods in two stages. The first stage involved the consideration of the testimony, hospital records and autopsy relating only to Paul Woods. The second stage involved the consideration of records and testimony, bolstered by the opinion of Dr. De-Maio, concerning past occurrences wherein other children had suffered illness under allegedly similar circumstances, some of whom died and some of whom did not.
Examination of the record forcibly demonstrates, as the majority opinion concedes, the failure of the government to establish either the corpus delicti or the criminal agency of the accused through evidence relating solely to Paul Woods. The government’s principal witness on the subject of corpus delicti was Dr. Vincent J. DeMaio, the forensic pathologist who performed the autopsy on Paul. Before rendering an opinion, Dr. DeMaio considered the hospital records of Paul Woods at Johns Hopkins Hospital, Kirk Army Hospital, and Walter Reed General Hospital; the autopsy which he performed; and the oral testimony of witnesses pertaining to Paul Wood’s history. Based upon those facts, Dr. DeMaio testified that he was only 75% certain that “the death was most probably a homicide.” As a logical corollary, he admitted there was a 25% *143chance that the death of Paul Woods was due to natural causes.10
Unable to prove the corpus delicti utilizing evidence relating solely to Paul Woods, the government turned to evidence of past occurrences involving children other than Paul Woods. In my opinion, such evidence was improper on any account. It was not within any exception to the rule; it was highly prejudicial in the context of this trial; and most or all of such evidence failed to meet the required evidentiary standard of plain, clear and conclusive. Of equal force is the total lack of authority or precedent permitting use of evidence of other crimes, much less evidence of prior occurrences, to prove the corpus delicti in a homicide.11
Being unable to justify admitting the proffered evidence by the authorities, the majority draws analogies from arson and confession cases. Upon inspection, however, neither such class of cases offers persuasive support for use of evidence of prior occurrences to prove the criminal agency of the accused in a homicide trial. Analysis of People v. Wolf, supra, relied upon by the majority, for example, discloses that testimony of prior acts was admitted for the primary purpose of showing a connected scheme or purpose, i.e., to defraud an insurance company, and then apparently was allowed to be considered for all purposes. But where the issue has been squarely presented, courts have refused to admit such evidence to prove the corpus delicti *144of arson.12 The confession cases, moreover, seem quite inapposite since the confession itself, if corroborated by substantial independent evidence, supplies the proof. The evidentiary problems, then, in confession cases, rather than relating to proof of the corpus delicti by proof of similar occurrences, should concern establishment of the trustworthiness of the confession and some independent evidence touching the corpus delicti so that the confession may be admitted13
I am unable, then, to find support, either by precedent or by force of reasoning, for the position of the majority that evidence of prior occurrences is admissible to prove the corpus delicti, whether or not it may be said to include the criminal agency of the accused. Indeed, it is clear to me that such use of prior occurrences is an impermissible erosion of the policy underpinnings of the common law requirement of proof of corpus delicti. In the context of this trial, it has become a bootstrap operation allowing the government to circumvent proof of a major element of its case, and its allowance ought to be reversible error.
Regardless of the purpose for which evidence of prior occurrences is sought to be introduced, such evidence is never admissible unless proven by plain, clear, and conclusive evidence. Here lies the basis for an equally fundamental objection to the government’s use of evidence of prior occurrences.14
In no instance cited by the government involving other children was the defendant convicted of the crimes alleged by the government; nor was she indicted or even charged with the commission of such crimes. To the contrary, in each instance involving the death of another child a formal death certificate or autopsy report was issued stating the child died of natural causes and naming the disorder.15
A brief review of the more compelling illustrations may be useful. Charles Lewis Stewart was born prematurely in 1946 and died less than two months later. The government’s testimony involved a review by Dr. DeMaio of the death certificate and the hospital records. From those he gave an opinion that the cause of death could not have been an enlarged thymus as certified on the certificate since recent medical thinking has been that it is normal for children to have an enlarged thymus. Nevertheless, Dr. DeMaio admitted that the symptoms leading to the first hospital admission were as consistent with the diagnosed condition of malnutrition as they were with the government’s theory of smothering. He admitted that, given malnutrition, he could not argue with a finding that the death was a natural crib death, which he stated is now generally the diagnosis for death which used to be attributed to an enlarged thymus.
The death in 1946 of John Wise, the defendant’s nephew, offers another ex*145ample where evidence of a prior occurrence fell far below the clear and convincing standard. According to the death certificate and the autopsy findings, John Wise died of diphtheria. The government’s own proof established that two other children were taken from the same household during the same week to the hospital isolation unit with similar cases of severe diphtheria. The only evidence adduced by the government to challenge this finding of natural death was the statement by Dr. DeMaio that, had he performed the autopsy, he would have preferred to examine the neck organs in order to be certain of the diagnosis of diphtheria. Indeed, he admitted that the diagnosis was consistent with diphtheria.
The government’s evidence also included five other occurrences where children in the defendant’s presence or under her care suffered death. The death certificates are summarized in note 10. In each instance, Dr. DeMaio, the government’s chief witness, stated his reasons for disagreeing with the specific findings in each death certificate or autopsy report. In every case, however, Dr. DeMaio was careful to state that the evidence was not inconsistent with a natural death. In my opinion, such evidence was inconclusive and failed dismally to meet the required standard of plain, clear, and conclusive.
In view of the highly prejudicial nature of the evidence of other deaths, it is clear that error in admitting evidence of any of the other deaths should entitle the defendant to a new trial. This is not a case of harmless error since “there is a reasonable possibility that [improper evidence] contributed to the conviction.” Fahy v. Connecticut, 375 U.S. 85, 84 S.Ct. 229,11 L.Ed.2d 171 (1963).
I feel that the admission of other-occurrence evidence severely impaired the fairness of this trial to the defendant. As the Supreme Court noted in Brinegar v. United States, 338 U.S. 160, 174, 69 S.Ct. 1302, 93 L.Ed. 1879 (1949) :
“Guilt in a criminal case must be proved beyond a reasonable doubt and by evidence confined to that which long experience in the common-law tradition, to some extent embodied in the Constitution, has crystallized into rules of evidence consistent with that standard. These rules are historically grounded rights of our system, developed to safeguard men from dubious and unjust convictions, with resulting forfeitures of life, liberty and property.”
In my opinion, the defendant was denied those safeguards here.
I would reverse and remand for a new trial.
. United States v. Baldivid, 465 F.2d 1277 (4th Cir. 1972) ; United States v. Mastrototaro, 455 F.2d 802 (4th Cir.), cert, den., 406 U.S. 967, 92 S.Ct. 2411, 32 L.Ed.2d 666 (1972) ; United States v. Smith, 446 F.2d 200 (4th Cir. 1971) ; United States v. Samuel, 431 F.2d 610 (4th Cir. 1970) ; United States v. Dutsch, 357 F.2d 331 (4th Cir. 1966).
We said in Samuel, 431 F.2d at p. 612: “Such evidence is admissible only where it is relevant for some purposes other than to show that the accused, because he is a man of criminal character, committed the crime charged.”
. I have found no authority, and none has been cited to the court, which permits proof of prior crimes or acts, much less proof of prior occurrences, as an acceptable means of proving the corpus delicti. The majority agrees with this so far as homicide goes but finds exception for arson cases and cases involving confessions.
. See also United States v. Machen, 430 F. 2d 523, 526 (7th Cir. 1970) (crisp, concise, and persuasive) ; Labiosa v. Canal, 198 F.2d 282, 284-285 (4th Cir. 1952) (clear).
. See United States v. Phillips, 401 F.2d 301, 305-306 (7th Cir. 1968) ; DeVore v. United States, 368 F.2d 396 (9th Cir. 1966).
Compare the Proposed Federal Buies of Evidence, Rule 403, which offers a balancing test excluding relevant evidence under certain circumstances:
“Although relevant, evidence may be excluded if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury, or by considerations of undue delay, waste of time, or needless presentation of cumulative evidence.”
“Unfair prejudice” is defined by the Advisory Committee as “an undue tendency to suggest decision on an improper basis, commonly, though not necessarily, an emotional one.”
In Baldivid, the admissibility of the evidence under the intent and knowledge exception was acknowledged. The difference in the court was on the extent of the prejudice.. Since the majority here does not depend on the signature or lack of accident exceptions, I need not apply the weighing test in precisely the same sense as applied by the dissent in Baldivid.
. Tlie majority alters the balance by weighing the relevance of the evidence plus the need for the evidence against the prejudice resulting to the defendant. It also applies the test without first finding an exception to gain admissibility. The infirmity in this formulation is that as the need for the evidence increases, the probative value may decrease and still the evidence will be admissible. This new test given effect by today’s ruling, elevates need, if accompanied by the slightest probative value, to the status of an exception in compelling cases. This, of course, coupled with the absence of a limiting instruction, gives no protection to the defendant in a case already clouded with emotional issues.
Indeed, I find that such emphasis on need comes perilously close to a rule in the law of evidence that the end may justify the means, and my inhibition at such a departure from the common law tradition is at the root of my dissent. While the executive branch of government must frequently, of necessity, make the decision as to need as against consequences in securing evidence, such a balance should not be given dignity by the courts in ascertaining admissibility absent an exception to the general rule. The use of need in any balancing test properly ought to be limited to applying, in a negative sense, an exclusionary rule, not to bolstering admissibility. McCormick on Evidence, 1954 Ed. § 157, clearly points this out, and advocates that the test be used as a further limit on admissibility, not as an extension of it.
Rule 403 of the proposed Federal Rules of Evidence (see n. 4) must be read together with Rule 404(b), quoted below, as adopting the existing rules of evidence, limited, however, as advocated by McCormick and Judge Sobeloff’s dissent in Baldivid. The comment to Rule 404(b) explains the rule is so deeply imbedded in our jurisprudence as to assume almost constitutional proportions. Thus, the decision of the majority takes a new departure.
Rule 404(b)—
“(b) Other crimes, wrongs, or acts. Evidence of other crimes, wrongs, or acts is not admissible to prove the character of a person in order to show that he acted in conformity therewith. This subdivision does not exclude the evidence when offered for other purposes, such as proof of motive, opportunity, intent, preparation, plan, knowledge, identity, or absence of mistake or accident.”
. Perkins, The Corpus Delicti of Murder, 48 Va.Law Rev. 173-74 (1962).
. See, e. g., 1 Wharton’s Criminal Evidence § 17 at 48 (12th ed. 1955) ; Lefave, Criminal Law § 4 at 16-17 (1972).
. Evans v. United States, 122 F.2d 461, 465 (10th Cir. 1941), cert. den. 314 U.S. 698, 62 S.Ct. 478, 86 L.Ed. 558 (1942) ; Murray v. United States, 288 F. 1008, 1016, 53 App. D.C. 119 (1923).
. The majority faces the problem squarely and decides that proof of prior occurrences is admissible to prove not only the death and criminal agency of someone, but also the criminal agency of the accused. Opinion p. 15-16. Cf. Bowie v. Commonwealth, 184 Va. 381, 389, 35 S.E.2d 345 (1945) (Proof of the charge compared). See also Evans, supra.
. I am of opinion that evidence of a 75-25 percentage has no place in a criminal prosecution. It is an invasion of the province of the jury.
I am further very doubtful as to the relevancy and competency of most of Dr. De-Maio’s testimony in view of at least two factors. First, the prior occurrences to which he testified were remote in time: Charles Stewart and John Wise died more than 25 years before trial; Mary E. Huston —21 years; Carol A. Huston — 20 years; Lillie Marie Stewart — 14 years; and Marian Rash — 8 years. See United States v. Mach-en, 430 F.2d 523, 526 (7th Cir. 1970). Second, Dr. DeMaio neither examined, attended, nor performed autopsies upon any of these children. The physicians who did were almost certainly more qualified to testify at trial as to the cause of death. Yet, their death certificates, presumed to be correct, have been allowed to be outweighed by opinion evidence arrived at in some cases twenty-five years later.
Death certificates were signed by physicians in each prior death, other than Marian Rash. For him, the record contains an autopsy report. The following information may be summarized:
a. Charles Lewis Stewart — Ohio Death Cert. Exhibit 12-C. Died 8-23-46 of “Enlarged Thymus” and “Status Lymphat-icus.” Signed by John B. Gravis, M.D.
b. Mary Elizabeth Huston — Ohio Death Cert. Exhibit 13-D. Died 8-25-50 of “Asphyxiation due to mucous plug” and “patent foramen ovale.” No autopsy. Signed by F. Richardson, M.D.
c. Carol Ann Huston — Ohio Death Cert. Exhibit 14-D. Died 5-12-52 of “Epiglot-tifus” — No autopsy. Signed by Sol Mag-gied, M.D.
d. John Wise — Ohio Death Cert. Exhibit 15-B. Died 12-26-46 of “diptheria.” Signed by John B. Gravis, M.D. Autopsy performed (Exh. 15-C) by R. M. Hart-well.
e. Lillie Marie Stewart — Ohio Death Cert. Exhibit 16-B. Died 5-18-58 of “Fulminating pneumonia” and “broncholitis.” Signed by Schmelzer, M.D. Whether autopsy performed is not indicated on certificate.
f. Marian L. Rash — Colorado autopsy report. Exhibit 17-E. Signed by Lt. Col. Clearkin, M.D. Child died 5-4-64 of bronchiolitis. Autopsy performed immediately.
. There is considerable authority supporting the proposition that evidence of other occurrences allegedly involving the accused is not admissible for the purpose of establishing the corpus delicti in a homicide case.
We have considered the very question presented here by way of dicta in Lovely, supra, 169 F.2d at p. 391, a rape case, in which we said by way of emphasis that evidence of a prior rape had no more tendency to establish a plan or scheme within the meaning of the rule “ . . . than similar incidents of a prior homicide or burglary would render evidence thereof admissible on a trial for those offenses.” [Emphasis added]
With Judge Parker in Lovely, I find it impossible to understand a rule which will admit evidence as proof of guilt which would be inadmissible even as proof of the bad character of the defendant if she put the matter in issue. Lovely, p. 388.
. Kahn v. State, 182 Ind. 1, 105 N.E. 385 (Ind.1914). Reversing a lower court conviction of arson, the Indiana Supreme Court in Kahn noted:
“Other crimes than the one for which the defendant is upon trial are never permitted to be shown to prove the corpus de-licti. It is only permissible in the trial of a criminal charge, in those cases where the act constituting the crime under investigation has been clearly established and the motive, intent, or guilty knowledge of the defendant is an issue.” 105 N.E. 385, 386.
. E. g., Opper v. United States, 348 U.S. 84, 75 S.Ct. 158, 99 L.Ed. 101 (1954) ; United States v. Waller, 326 F.2d 314 (4th Cir. 1963), cert. den. 377 U.S. 946, 84 S.Ct. 1355, 12 L.Ed.2d 309 (1964).
. See cases cited at note 3 and accompanying text. See also note 10.
. As a general rule, death certificates are, by statute, prima facie evidence of the facts stated therein. Each of the states here involved had such provisions. Md.Code Ann. Art. 43 § 26 (Repl. vol. 1971) ; Colo.Rev. Statutes § 66-8-24 (1963) ; Ohio Rev.Code Ann. § 3705.05 (1953).
See also note 10.