Defendant Vicki Jo Aten was convicted of second degree manslaughter after four-month-old Sandra Bibber died in the night while under Aten’s care. Aten confessed several times that she had placed her hand over Sandra’s mouth and nose to stop her from crying, causing Sandra’s death. Because one may not be convicted solely on the basis of a confession or admission, we must determine whether there is sufficient independent evidence of the corpus delicti, i.e., the body of the crime. We must also decide whether any of Aten’s statements may be used as independent evidence to corroborate Aten’s confessions and admissions. We hold that none of Aten’s statements about the event can be used, and that insufficient independent evidence corroborates her confessions and admissions. Accordingly, we reverse Aten’s conviction of second degree manslaughter.
Facts
To clarify our analysis of the corpus delicti doctrine, we divide the facts into four categories: the independent evidence; Aten’s description of the death; Aten’s conduct after Sandra’s death; and Aten’s confessions and admissions that she killed Sandra Bibber by placing her hand over Sandra’s mouth.
Independent Evidence of Corpus Delicti
Aten babysat for Sandra Bibber and Sandra’s three siblings while Sandra’s mother worked a night shift. Aten had cared for the children since Sandra’s birth four months earlier. Two days before Sandra’s death, her mother had taken Sandra to her doctor with a cold. Sandra had nasal congestion, but no inflammation, and no *82medicines were prescribed. On the night in question, Sandra was in good health, and her mother had had no problems with Sandra that day. Aten arrived at 10:00 p.m., went into Sandra’s room and took the sleeping infant from her crib, held her briefly, and then put her back down. The child was quiet when the mother left for work.
The following morning paramedics were summoned to Sandra’s home at 7:22 a.m. They found Aten holding Sandra in a blanket. Sandra was already dead and they could do nothing for her. An autopsy disclosed no abnormalities or signs of manual interference with Sandra’s breathing, and the pathologist concluded that Sandra had died from Sudden Infant Death Syndrome (SIDS). The pathologist testified that an autopsy cannot distinguish between suffocation of an infant and SIDS. The pathologist testified that "given a different history,” i.e., facts outside of the autopsy findings, he could make the inference that someone had manually interfered with Sandra’s breathing. But the pathologist was never asked if he had such an opinion, and never testified to that opinion.
Aten’s Description of the Death.
On the morning of Sandra’s death, Aten told the paramedics and the police that Sandra had had trouble breathing during the night, and that she had been up with her from 10:00 p.m. until 4:06 a.m. Sandra finally fell asleep on the couch, and Aten left the baby there because she did not want to wake her. Aten went to sleep in the bedroom. Upon awakening shortly after 7:00 a.m., she found Sandra nonresponsive on the couch and immediately called 911.
Aten’s Conduct After the Death.
In the two weeks after Sandra’s death, Aten gave away some of her household possessions to her children, who range in age from eight to twenty-one. She made arrangements for relatives to care for her two youngest children and asked people to care for her pets. She arranged to *83store the rest of her possessions, and told her adult daughter that she feared the sheriff might lock up her house.
One week after Sandra’s death, Aten checked herself into Olympic Memorial Hospital suffering from acute grief and depression.
Aten’s Confessions and Admissions.
Several days after Sandra’s death, Aten asked Sandra’s mother about the autopsy results. When told that the pathologist had diagnosed SIDS, Aten responded that, "it isn’t true.” Aten said no more at the time.
On the evening of February 10, ten days after Sandra’s death, Dr. Peterson spoke with Aten during his hospital rounds. Aten had appeared anxious in the preceding days, but this evening was quiet and calm. Aten spontaneously told Dr. Peterson that she had killed Sandra. Dr. Peterson recorded her exact words:
She wore me out. I would have never done that if I had had more sleep. I suffocated her. It’s not an excuse. It’s just what happened.
Aten asked to talk to a sheriff’s deputy that night.
Deputies arrived at the hospital and began interviewing Aten at 10:35 p.m. The interview was recorded and the tape and a transcription were admitted at trial. The deputies advised Aten of her rights, and asked if she wished to talk with them. Aten replied, "I really do, but I think I better have an attorney present just to see if maybe, ah, I might be messing up somewhere along the line.” Sergeant Martin told Aten he could not continue speaking with her. Aten asked the deputies to turn the tape off, and spoke with them without a taped record for forty minutes. They then resumed the recording and Aten said she wished to continue the interview.1
Aten told the deputies that Sandra was "real sick,” and *84that she was "fussing and fussing and fussing.” Aten was very tired. She carried Sandra into the front room where it was warmer and laid her on the couch. Aten held Sandra for a long time but Sandra wouldn’t calm down:
Something went wrong that night with me. And I just know this baby won’t shut up and I’m doing the best I can . . . . She’s being real persistent about this crying. I’m doing something I don’t want to do. But it takes care of the problem. I have her in my arms, she’s on my lap. I take my hand and I place it over her mouth and her nose and I just hold that for a little while and she calms down. And I lay her down. She’s still fussing. And then I leave. And I go to bed. And ah the clock says 4:06.
Aten stated that she just wanted Sandra to calm down, and did not hold onto her mouth until she went limp: "I did not do that. I lay a live baby down and wake up to a dead one.” Nor did she put a pillow over Sandra’s mouth. According to Aten, Sandra was still whimpering and breathing when Aten left her on the couch. Sandra was rubbing her fingers on the couch and her feet were moving as well. Aten felt responsible for Sandra’s death, " 'Cuz she’s dead and look what I did.”
Sandra’s mother visited Aten at the hospital.2 Sandra’s mother testified that Aten told her that Aten caused Sandra’s death. Aten said she had placed her hand over Sandra’s mouth and nose because Sandra was sobbing and then had suffocated her with a pillow. Aten said she felt very guilty, but she just had to let it out.
The day after Aten’s interview with the deputies, she spoke with a Child Protective Services (CPS) caseworker. Aten wanted to place her children into foster care before she was taken into custody following her confession. Aten told the caseworker that she had killed the baby when she *85put her hand over the baby’s mouth and that she didn’t want her children to know what had happened.
Aten was charged with second degree manslaughter. She waived her right to a jury trial and was tried by a superior court judge. Aten did not testify. The trial court found Aten guilty:
The defendant was exhausted and stressed at the time she committed that act due to the child’s crying. In so acting she failed to be aware of the substantial risk of suffocation and that failure was a gross deviation from the standard of care a reasonable person would exercise in that situation.
The standard range sentence for Aten was twelve to fourteen months. The court imposed an exceptional sentence of thirty months due to Sandra’s extreme vulnerability and Aten’s violation of trust.
Analysis
The corpus delicti doctrine provides that a court will not consider a confession as evidence of guilt unless it is corroborated by independent proof of the corpus delicti, or "body of the crime.”3 In its 1986 decision in Bremerton v. Corbett,4 our Supreme Court quoted a traditional description of the doctrine:
The confession of a person charged with the commission of a crime is not sufficient to establish the corpus delicti, but if there is independent proof thereof, such confession may then be considered in connection therewith and the corpus delicti established by a combination of the independent proof and the confession.
The independent evidence need not be of such a character as would establish the corpus delicti beyond a reasonable doubt, or even by a preponderance of the proof. It is sufficient if it prima facie establishes the corpus delicti.
The elements of corpus delicti are: (1) proof of injury or *86loss, here, the death of Sandra; and, (2) proof of someone’s criminal act as the cause of the death.5 In this case of second degree manslaughter, the criminal act is causing death through criminal negligence.6 A person acts negligently when she "fails to be aware of a substantial risk that a wrongful act may occur” and her lack of awareness "constitutes a gross deviation from the standard of care that a reasonable man would exercise in the same situation.”7 Thus, to corroborate Aten’s admissions and confessions, we must find independent evidence that Aten acted in a way which showed a lack of awareness of a substantial risk that a wrongful act might occur, and constituted a gross deviation from reasonable care.
Washington follows the majority rule that the confession or admission must be corroborated by independent evidence of both the injury and the criminal act.8 Dissatisfaction with the corpus delicti rule has led some jurisdictions to modify the doctrine and to require only corroboration of the trustworthiness of the confessions or admissions instead of corroboration of each element of the corpus delicti.9 Our Supreme Court has declined to adopt the trustworthiness doctrine.10 Thus, it is not enough to find evidence which shows that Aten’s confessions and admis*87sions are trustworthy. We must find independent evidence that criminal negligence caused Sandra’s death.
The confession can be admitted after independent proof has been presented. The State must then establish the corpus delicti beyond a reasonable doubt, but in evaluating the sufficiency of the evidence, the court can consider both the independent proof and the confession.11
Before we can evaluate the sufficiency of the proof of the corpus delicti, we must determine whether Aten’s statements can be viewed as independent evidence of the corpus delicti or whether such statements, like confessions, require corroboration. The corpus delicti rule clearly applies to Aten’s recorded statement to the sheriff’s deputies, which appears to be a classic confession. Does it also apply to: the statements to her doctor, the CPS caseworker, and Sandra’s mother; the statements to the paramedics and deputies on the morning of Sandra’s death; giving away her property and checking herself into the hospital?
Descriptions of the corpus delicti rule frequently use the term "confessions,” as in the traditional definition quoted above from Bremerton v. Corbett, 106 Wn.2d 569. In Corbett, our Supreme Court cited a Utah case for the distinction between "confessions” and "admissions.”12 The Utah court defined a confession as the defendant’s admission of all the crime’s necessary elements whereas an admission "merely admits some fact which connects or tends to connect the defendant with the offense but not with all the elements of the crime.”13 The Supreme Court did not find it necessary to address the distinction between confessions and admissions in Corbett, but seems to have assumed that the corpus delicti rule applies to both, stating that the "requirement of independent proof of the corpus delicti is . . . applicable to a defendant’s extrajudicial confessions *88and admissions.”14 A number of Washington cases since Corbett have similarly described the corpus delicti rule as applicable to admissions as well as confessions.15
Professor McCormick states that most jurisdictions recognize that independent proof must corroborate not only confessions, but also admissions and exculpatory statements, "because all involve the risks which the requirement is designed to reduce.”16 The decision to apply the corpus delicti rule to admissions turns on identifying the risks to be reduced by the corpus delicti rule. In Corbett, the Washington Supreme Court identified several reasons to distrust confessions — they may be involuntary, or they may be false for other psychological reasons:
This distrust stems from the possibility that the confession may have been misreported or misconstrued, elicited by force or coercion, based upon mistaken perception of the facts or law, or falsely given by a mentally disturbed individual. Note, [Proof of the Corpus Delicti Aliunde the Defendant’s Confession] 103 U. Pa. L. Rev. [638] at 642-46 [(1955)]; Note, Confession Corroboration in New York: A Replacement for the Corpus Delicti Rule, 46 Fordham L. Rev. 1205 (1978). Thus, it is clear that the corpus delicti rule was established to prevent not only the possibility that a false confession was secured by means of police coercion or abuse but also the possibility that a confession, though voluntarily given, is false.[17]
In addition, the Pennsylvania Law Review Note cited in Corbett quotes an observation made 150 years ago that the corpus delicti rule "certainly best accords with the human*89ity of the criminal code . . . .”18 This legal heritage may have been influenced by the distrust found in religious tradition of using confessions alone to establish the truth.19
In light of the weight of authority and the purposes of the doctrine, we hold that the corpus delicti rule requires corroboration of any statement made by the defendant, whether confession, admission, or even neutral description. The doctrine guards not only against coerced confessions, but against uncorroborated admissions springing from a false subjective sense of guilt. A defendant who falsely believes herself guilty may "admit” that guilt through any description of the events in question, whether that description is given to police or a close friend, whether inculpatory, exculpatory, or facially neutral. The purpose of the corpus delicti doctrine would be frustrated if the court allowed a false confession to be "corroborated” by a false admission, or even by seemingly innocent statements. The corpus delicti doctrine incorporates a policy that we will not find a defendant guilty beyond a reasonable doubt based solely on the defendant’s subjective belief; we require prima facie corroboration.20
Accordingly, none of Aten’s statements may be considered by the court as evidence of guilt absent independent corroborating evidence. The court cannot use Aten’s statements to paramedics and sheriffs deputies the morning of Sandra’s death. Without those statements, the circumstances of Sandra’s death are insufficient of themselves to *90provide even prima facie inference that criminal negligence led to Sandra’s death.
The State argues that Aten’s statements are corroborated by her actions in the two weeks after Sandra’s death— making arrangements for the care of her two youngest children, giving away her possessions and checking herself into the hospital for severe grief and depression. These actions cannot serve as independent corroboration of Aten’s statements for at least three reasons. First, Aten’s sense of guilt does not tend to show that Aten caused Sandra’s death through criminal negligence, i.e., that Aten disregarded a substantial risk that she would injure Sandra and that her actions were a gross deviation from reasonable care. Second, Aten’s actions are not "independent evidence” of the corpus delicti — that the baby’s death was the product of criminal agency. Rather, both Aten’s statements and actions are only really indicative of Aten’s subjective sense of guilt or responsibility. Because the corpus delicti doctrine is specifically designed to prevent convictions based solely on the defendant’s sense of guilt, Aten’s actions or what they indicate cannot be relied upon as independent evidence of the corpus delicti. Third, depression and a sense of guilt are common, if not inevitable, reactions to a SIDS death.21 One writer has observed that many SIDS parents (and babysitters) review the events of the death and think: "Healthy infants do not die without a cause. What must have happened is that there was something I did or did not do that caused the death. Therefore, I am the cause of the death of my child.”22 Aten’s own depression and subsequent actions simply fail to establish the corpus delicti even prima facie.
*91 The State also contends that the testimony of the pathologist corroborates Aten’s confessions and admissions. The pathologist testified that the autopsy showed that Sandra’s death could have been caused by SIDS or by manual interference with her breathing. " 'Prima facie’, in this context, means only that there be evidence of sufficient circumstances which would support a logical and reasonable inference [of criminality].”23 Evidence may lead to a reasonable inference of criminality, or it may lead to a reasonable inference of innocence. But evidence that simply fails to rule out criminality or innocence does not reasonably or logically support an inference of either. It would be speculative to conclude from the autopsy report that Aten was criminally negligent.
We do not reach this result lightly. This is an extremely close case and Aten has repeatedly said that she killed Sandra. Multiple confessions cannot corroborate one another.24 Sandra’s death was tragic, but we cannot compound that tragedy by affirming a conviction upon confessions and admissions that are not corroborated by independent evidence. Without the admission of Aten’s statements, the State failed to provide sufficient proof of Aten’s guilt. We therefore reverse Aten’s conviction and remand for dismissal.
Houghton, J., concurs.
At a pretrial hearing, the trial court found Aten’s statements to the deputies to be voluntary and admissible. Aten appeals this determination, but our resolution of the corpus delicti issue makes it unnecessary to reach this issue.
It is unclear whether this conversation preceded or followed Aten’s statement to the deputies. The mother testified she visited Aten on February 10 in the late afternoon, but also that she was "not good with dates.” She also testified that she reported this conversation to the sheriffs department on the same day it occurred, but none of the three deputies mentioned the mother’s report during their testimony.
See, e.g., State v. Riley, 121 Wn.2d 22, 32, 846 P.2d 1365 (1993).
Bremerton v. Corbett, 106 Wn.2d 569, 574-75, 723 P.2d 1135 (1986) (citations omitted) (quoting State v. Meyer, 37 Wn.2d 759, 763-64, 226 P.2d 204 (1951)).
Corbett, 106 Wn.2d at 573-74.
RCW 9A.32.070.
RCW 9A.08.010(l)(d).
See e.g., State v. Vangerpen, 125 Wn.2d 782, 888 P.2d 1177 (1995); State v. Smith, 115 Wn.2d 775, 781-82, 801 P.2d 975 (1990); Corbett, 106 Wn.2d at 573, 575, 577; 1 McCormick on Evidence § 145, at 557 (John W. Strong ed., 4th ed. 1992); 7 John Henry Wigmore, Evidence § 2071 (Chadbourne rev. 1978); E. H. Schopler, Annotation, Corroboration of Extrajudicial Confession or Admission, 45 A.L.R.2d 1316, 1327-29 (1956).
See, e.g., Opper v. United States, 348 U.S. 84, 75 S. Ct. 158, 99 L. Ed. 101 (1954); United States v. Lopez-Alvarez, 970 F.2d 583 (9th Cir.), cert. denied, 113 S. Ct. 504 (1992); Jacinth v. State, 593 P.2d 263, 266 (Alaska 1979); State v. Yoshida, 44 Haw. 352, 354 P.2d 986 (1960); State v. George, 109 N.H. 531, 257 A.2d 19 (1969); State v. Lucas, 30 N.J. 37, 52, 152 A.2d 50 (1959); State v. Paris, 76 N.M. 291, 294, 414 P.2d 512 (1966); State v. Parker, 315 N.C. 222, 337 S.E.2d 487 (1985); see generally 29A Am. Jur. 2d § 753 (1994).
Corbett, 106 Wn.2d at 577-78.
State v. Lung, 70 Wn.2d 365, 371-72, 423 P.2d 72 (1967); State v. Zuercher, 11 Wn. App. 91, 92-93, 521 P.2d 1184, review denied, 84 Wn.2d 1004 (1974).
Corbett, 106 Wn.2d at 575 n.1.
State v. Karumai, 101 Utah 592, 601, 126 P.2d 1047 (1942).
Corbett, 106 Wn.2d at 575.
See, e.g., Vangerpen, 125 Wn. 2d at 795-96; Riley, 121 Wn.2d at 32; State v. Nelson, 74 Wn. App. 380, 874 P.2d 170, review denied, 125 Wn.2d 1002 (1994); State v. Biles, 73 Wn. App. 281, 871 P.2d 159, review denied, 124 Wn.2d 1011 (1994); State v. Solomon, 73 Wn. App. 724, 727, 870 P.2d 1019, review denied, 124 Wn.2d 1028 (1994); State v. Cobelli, 56 Wn. App. 921, 924, 788 P.2d 1081 (1989).
1 McCormick on Evidence § 145, at 560; see also Opper, 348 U.S. at 84; Smith v. United States, 348 U.S. 147, 75 S. Ct. 194, 99 L. Ed. 192 (1954).
Corbett, 106 Wn.2d at 576-77.
103 U. Pa. L. Rev. 647-48 (quoting 1 Greenleaf Evidence 260 (2d ed. 1844)).
Cf. Deuteronomy 19:15 ("A matter must be established by the testimony of two or three witnesses.”), John 5:31 ("If I testify about myself, my testimony is not valid.”) (New International version); Ketuba 2:9 ("none may be believed when he testifies of himself.... None may testify of himself.”).
Our holding is contrary to the result in State v. Wright, 76 Wn. App. 811, 888 P.2d 1214 (1995), in which Division One held that the defendant’s confession was sufficiently corroborated by the defendant’s earlier false exculpatory statement to the police officer investigating the offense. It does not appear, however, that the defendant in Wright argued that his exculpatory statement could not be used as corroboration, and Division One never addressed the particular point we decide here.
E.g., John D. DeFrain, Deanne K. Jakub and Betty Lou Mendoza, The Psychological Effects of Sudden Infant Death on Grandmothers and Grandfathers, 24 Omega 165 (1992); John D. DeFrain, Learning About Grief From Normal Families: SIDS, Stillbirth, and Miscarriage, 17 J. Marital and Family Therapy 215 (1991); Julius Goldberg, The Counseling of SIDS Parents, 19 Clinics in Perinatology 927 (1992); Luisella Z. Schwartz, The Origin of Maternal Feelings of Guilt in SIDS, 533 Annals of N.Y. Academy of Sciences: The Sudden Infant Death Syndrome 132 (1988).
Golberg, supra, at 932.
Corbett, 106 Wn.2d at 578-79.
See State v. Neslund, 50 Wn. App. 531, 749 P.2d 725, review denied, 110 Wn.2d 1025 (1988); see also United States v. Calderon, 348 U.S. 160, 275 S. Ct. 186, 99 L. Ed. 20 (1954) (holding that additional admissions are insufficient to corroborate confession); United States v. Northrup, 482 F. Supp. 1032, 1037 (D. Nev. 1980) (stating that "If two admissions, in and of themselves, are untrustworthy, obviously, they cannot be bootstrapped together to raise each other to the level of trustworthiness.”); Bines v. State, 118 Ga. 320, 45 S.E. 376, 379 (1903) (stating that “even two positive confessions of guilt, without independent proof of the corpus delicti, would not be sufficient to authorize a conviction.”); Duncan v. State, 64 Md. App. 45, 494 A.2d 235, 239 (1985) ("multiple confessions by an accused cannot corroborate each other”); Lemons v. State, 49 Md. App. 467, 433 A.2d 1179 (1981); People v. Cuozzo, 292 N.Y. 85, 54 N.E.2d 20 (1944) (holding that there was insufficient independent evidence of the corpus delicti even though defendant confessed eleven times); E.H. Scholpler, Annotation, Corroboration of Extrajudicial Confession or Admission, 45 A.L.R.2d 1316, 1327.