The petitioner, the Estate of J. Douglas King, appeals an order of the Hillsborough County Probate Court (Cassavechia, J.) declaring the decedent to be intestate because the petitioner could not produce an original will. The petitioner argues that the probate court erred by relying upon the presumption of revocation even though the estate produced an original codicil to the original will. Alternately, the petitioner argues that it presented sufficient evidence in the probate court to overcome the presumption of revocation, and that the court relied upon inadmissible hearsay in finding otherwise. We reverse and remand.
The decedent, J. Douglas King, executed a will in 1994 that left his estate to his second wife of fifteen years, Laurel King, with whom he had two minor children. Douglas also had three adult children (Rebecca King Felmet, Rachel King and Jason King) from a previous marriage. The will conditioned bequests to these three children upon Laurel predeceasing Douglas. On July 24, 1997, after presenting a copy of the will to Ellen Peterson, a Massachusetts attorney, Douglas executed a codicil to the will. The codicil made no substantive changes to the will, but merely altered provisions for appointing the executor and trustees, and otherwise reconfirmed the 1994 will. Peterson retained the original codicil and the copy of the will.
Douglas died suddenly and unexpectedly in a motorcycle accident on September 10, 2000. The original 1994 will, last known to be in Douglas’ possession, was never found. Instead, Laurel obtained the 1997 original codicil and copy of the will from Peterson, and filed them with the probate court. Rebecca, Rachel, and Jason contested these documents, arguing that a presumption of revocation applied to the missing will. The probate court agreed that the presumption must be invoked, and conducted an evidentiary hearing to determine the likely fate of the will. See In re Estate of Fuller, 119 N.H. 132, 135 (1979).
Much of the testimony at the hearing concerned Douglas’ relationship with his family. Evidence established that Douglas’ marriage with Laurel, although generally loving, featured a series of arguments, separations and reunions. Many of these incidents concerned Douglas’ problems with alcohol and prescription drugs. On one occasion in 1989, Laurel filed for divorce, but withdrew the filing after Douglas promised to change. In 1993 or 1994, she consulted an attorney about a separation. Douglas and Laurel had another argument on September 2, 2000, shortly before Douglas’ death, after which Douglas wrote a missive to himself lamenting that his *229“life has come to an end.” Laurel, however, testified that they reconciled, and had planned to renew their wedding vows in October of 2000.
The witnesses agreed that throughout his second marriage, Douglas was close to his children from the first marriage with the exception of his oldest son, Jason, who had a substance abuse problem and was estranged from his father until 2000, when they reconciled. Douglas lavished gifts on these children, including direct financial assistance, a house, and several cars. Douglas assisted his children in this manner both before and after the execution of the 1994 will. Laurel testified that Douglas was generous to his adult children because he only intended to provide for them while he was alive.
Other testimony concerned Douglas’ financial plans. Douglas, who was not always a wealthy man, realized a major accession to his wealth between 1994 and 1997, at which time the joint estates of Douglas and Laurel were valued at approximately eight million dollars. Douglas purchased a home in Hillsborough, which he intended to be a communal family estate for all of his children. He consulted with three different attorneys, from 1996 to 1999, about the possibility of revising the testamentary plan in his will to minimize tax liability, although no evidence of any new will exists save the 1997 codicil. In 1999 Douglas and Laurel undertook to divide their marital assets, although they subsequently placed their assets back into a joint account.
The petitioner produced three witnesses who testified about statements Douglas made regarding the status of the 1994 will. First, Douglas Hatfield, Douglas’ attorney and neighbor in Hillsborough, testified that Douglas King had told him in September of 1999 that he “had an old will that he had amended that left everything to Laurel.” Hatfield understood Douglas to mean that this will was still effective at that time. Second, Donna Stafford, a friend of Douglas and Laurel, testified that Douglas told her in June of 2000 that “he had prepared his will, that he had left everything to Laurel, everything was in her name, and he knew that [Laurel] would take care of all of his children.” Stafford testified that Douglas told her one week before his death that “your husband will take care of [your stepson], just like Laurel will take care of my five children.” Finally, John Zelonis, Douglas’ friend and employee, testified that on August 31, 2000, Zelonis asked Douglas, “What would happen if you died tomorrow?” Douglas replied that “he had his affairs in order for years.” Zelonis also testified that Douglas and Laurel, prior to airplane trips, would leave him a list of names to call in the event “anything happens to both of us.” This list included Ellen Peterson, the attorney who kept the original codicil and copy of the 1994 will.
*230Finally, over the petitioner’s hearsay objection, the court admitted the testimony of Rebecca, Rachel and Jason regarding a statement made to them by Andrew Reyes, Douglas’ longtime accountant, several days after Douglas’, death. According to all three, Reyes told them that Douglas had said he had made specific provisions for his children in the event of his death. Reyes did not subsequently contact the children about these “specific provisions.” At the time of the hearing, Reyes was missing, and being sought by law enforcement agencies in connection with his suspected embezzlement of over a million dollars from Douglas’ estate.
On December .27, 2001, the probate court issued a written order concluding:
[I]t is as probable as not that Doug[las] destroyed the 1994 will (if he, in fact, did not prepare another which has not been located) knowing that intestacy would promote a degree of equality among his children, and between his children and his wife, that the 1994 will did not____Laurel’s proof is devoid of credible evidence establishing, directly or through reasoned inference, what became of the 1994 original will execution, or to otherwise account for its non-production.
On appeal, the petitioner first argues that the court erred by holding an evidentiary hearing to determine the fate of the missing 1994 original will because the' codicil was per se súfficient to prove the will. Second, it argues that it overcame the presumption of revocation at the evidentiary hearing. A third issue regarding the appointment of the administratrix of the estate is raised in the notice of appeal but not briefed, and therefore we deem it waived. See Herman v. Monadnock PR-24 Training Council, 147 N.H. 754, 758 (2002). We consider the preserved arguments in turn.
The petitioner first argues that the 1997 codicil is sufficient to prove the will and to obviate any inquiry into the will’s revocation. The petitioner relies upon the principle that a codicil, when executed with the formalities of a will, republishes the original will as of the date of the codicil’s execution. See Foster v. Farrand, 81 N.H. 448, 451 (1925). Thus, under New Hampshire law, a will is proved by proving a subsequent codicil. See Prob. Ct. R. 97. For example, if a flaw exists in the execution of a will, proof of a subsequent and properly executed codicil to that will validates the will’s provisions. Cf. Tyson v. Henry, 514 S.E.2d 564, 566 (N.C. 1999).
The petitioner interprets these principles to mean that the codicil becomes a separate testamentary instrument. Under this view, the original will, subsequent to the execution of a codicil republishing that will, is no longer relevant, as the codicil replaces the will. Therefore, the destruction of the original will has no legal effect unless accompanied by *231the destruction of the codicil, because otherwise the codicil instantly revives the revoked will’s terms by its reaffirmation and republication of the will. Under this logic, the petitioner submits that the introduction of the original codicil is sufficient to prove the terms of the 1994 will, regardless of whether Douglas destroyed the original document.
The petitioner cites only one case, In re Estate of Smith, 378 N.W.2d 555, 557 (Mich. Ct. App. 1985), that adopts its reasoning. Smith based its holding upon Michigan law defining a codicil as a “separate and independent testamentary instrument.” Id. at 557. The petitioner argues that RSA 21:22 (2000), which states, “[t]he word ‘will’ shall include codicils,” compels us to reach the same result.
RSA 21:22, however, is merely a rule of statutory construction that does not apply to this case. See RSA 21:1 (2000). New Hampshire, along with the majority of other jurisdictions, has adopted the alternate view that the codicil and will together function as a single testamentary instrument. See Loveren v. Eaton, 80 N.H. 62, 63 (1921); Jackson v. Thompson, 610 S.W.2d 519, 523 (Tex. Civ. App. 1980). In determining whether the revocation of the will revokes the codicil, the general rule in these jurisdictions is:
[W]here a codicil is of such a character that it may stand independently of the will, the revocation of the will does not affect the codicil. If, however, a codicil is not so complete a testamentary instrument as to stand alone, the revocation of the will to which it is appurtenant automatically revokes it.
79 Am. JUR. 2d Wills § 476 (2002).
We agree with this rule. Although the petitioner’s suggested bright-line rule would simplify probate procedures, we fear that it may contravene the testator’s intent. When a testator executes both a will and codicil adopting a minor change to that will, and then destroys that will, it is clear that the testator’s intent is to revoke his entire testamentary scheme. The rule proposed by the petitioner would dictate a contrary result. Moreover, the petitioner’s rule forces a testator who has drafted numerous minor codicils to destroy every one of these codicils, in addition to the original will, to accomplish a revocation.
In this case, the 1997 codicil made only a few minor amendments and was not sufficiently complete as a testamentary document to stand alone without the original will. Thus, the revocation of the 1994 will would have also revoked the 1997 codicil. For this reason, we agree that the evidentiary hearing was necessary for the probate court to determine what happened to the original 1994 will.
*232Alternately, the petitioner argues that the codicil incorporates the 1994 will by reference. We have stated:
To incorporate by reference into a will a document which is not physically a part of the will: (1) the document referred to in the will must exist at the time of the will’s execution; (2) the document referred to must be specifically identified in the will so that it is clear to what document the reference is made; and (3) words of intent to incorporate the document must be on the face of the reference.
In re Estate of Came, 129 N.H. 544, 548 (1987).
Because the codicil’s incorporation of the 1994 will meets all of these requirements, the petitioner argues that it reincorporates the will. This argument, like the foregoing one, relies upon the assumption that the codicil is an independent testamentary instrument equivalent to the will that it amends. We thus reject it for the same reason.
The petitioner next argues that it overcame the presumption of revocation at the evidentiary hearing. The probate court’s factual findings are final after an evidentiary hearing unless they are so plainly erroneous that they could not be reasonably made. RSA 567-A:4 (1997). However, because the trial court used an incorrect legal standard, admitted improper hearsay evidence, and did not consider certain testimony, we reverse and remand.
We first note that the trial court’s holding misinterpreted the presumption of revocation. When a duly executed will is last known to be in the testator’s possession, but cannot be found after his death, a presumption arises that the testator destroyed the will with the intent to revoke it. See Fuller, 119 N.H. at 135. The trial court concluded that the petitioner had not overcome the presumption of revocation. In New Hampshire, however, “presumptions only take the place of evidence, and when evidence appears, if only to the extent that an inference may be drawn from it, the presumption vanishes.” Id. (quotation omitted).
In this case, the presumption vanishes because there was evidence to the contrary. The existence of the 1997 codicil and copy of the 1994 will alone was legally sufficient to rebut the presumption. The respondents note that the codicil and copy were found in the possession of Douglas’ attorney rather than among Douglas’ personal effects, a fact that distinguishes this case from decisions where courts have found that the will and codicil overcame the presumption of revocation. See Will of Herbert, 391 N.Y.S.2d 351, 352 (N.Y. Surr. Ct. 1977); In re Estate of *233Kuszmaul, 491 So. 2d 287, 288 (Fla. Dist. Ct. App. 1986). Under New Hampshire law, however, evidence need only support an inference against revocation to rebut the presumption. Compare Fuller, 119 N.H. at 135, with Kuszmaul, 491 So. 2d at 287 (“the presumption may only be overcome by competent and substantial evidence”).
The fact that Douglas’ attorney retained a copy of the will along with a codicil supports such an inference in two ways. F'irst, if Douglas wished to revoke his entire testamentary scheme by destroying his original will, it is likely that he would have contacted Peterson and had the copy and codicil destroyed as well. Second, the fact that Douglas had only a copy of the will at the consultation with Peterson supports an inference that he had lost the original will before the execution of the codicil. Thus, the evidence was sufficient to overcome the presumption of revocation.
If the proponent overcomes the presumption of revocation, the court must decide whether the proponent has proven by a preponderance of the whole evidence that the will was more likely lost rather than destroyed. Fuller, 119 N.H. at 135. The probate court erred in not applying this test. Contrary to the probate court’s holding, the petitioner need not establish what happened to the 1994 will.
Second, the trial court erred in admitting the testimony of the three adult children regarding their conversation with Andrew Reyes, who told them that Douglas had made “specific provisions” for them. This testimony was inadmissible hearsay.
The proffered testimony involves two layers of hearsay: first, the statement from Douglas to Reyes, and second, the statement from Reyes to Rebecca, Rachel, and Jason. Each layer must fall within an exception to the rule against hearsay for the statement to be admissible. See N.H. R. Ev. 805. We will assume that the first layer of hearsay, like much of the testimony in this case, falls within the exception allowing statements by the deceased in certain circumstances when the trial court finds that the statement “was made in good faith and on decedent’s personal knowledge.” N.H. R. Ev. 804(b)(5).
The second layer does not fit within any specific exception. Instead, the trial court admitted the statement under New Hampshire’s “catch-all” exception, which allows a court to introduce the testimony of an unavailable declarant when, inter alia, the statement has “circumstantial guarantees of trustworthiness” equivalent to those in other recognized hearsay exceptions. N.H. R. Ev. 804(b)(6); see also Perry v. Parker, 101 N.H. 295, 297 (1958) (“The fundamental inquiry is not the name or number of the exceptions to the hearsay rule but whether under the circumstances *234[the hearsay] satisfies the reasons which lie behind the exceptions.” (quotations and citation omitted)).
Although the “catch-all” rule affords the trial court considerable discretion, we find that the admission of Reyes’ statement was an unsustainable exercise of discretion. See Town of Weave v. Paquette, 121 N.H. 653, 659 (1981); State v. Lambert, 147 N.H. 295, 297 (2001). Reyes’ absence denied the petitioner any opportunity to test the hearsay risks of insincerity, faulty perception, imperfect memory, and ambiguity present in his out-of-court statement about Douglas’ testamentary plans. See 5 J. Weinstein & M. Berger, Weinstein’s Federal Evidence § 802.02[3] (J. McLaughlin ed., 2d ed., 1999). No “circumstantial guarantees of trustworthiness” sufficient to counter these risks exist in the unsubstantiated statement of a declarant who is a missing person sought by law enforcement agencies and suspected by all parties of embezzling over a million dollars. Cf. Chinburg v. Chinburg, 139 N.H. 616, 618-19 (1995) (finding no error when trial court denied admission of decedent’s hearsay statements under Rule 804(b)(6) despite decedent’s repeated and consistent statements over a twenty-two-month period regarding hi's intentions for life insurance policy); United States v. Singleton, 125 F.3d 1097, 1106 (7th Cir. 1997) (listing “the character of the witness for truthfulness and honesty, and the availability of evidence on the issue” as a factor in determining sufficient guarantees of trustworthiness).
Nor can we find any authority in our case law or our Rules of Evidence to support the trial court’s contention that the absence of a jury permits the admission of this statement. See N.H. R. Ev. 1101(b) (“These rules apply generally to all civil and criminal proceedings unless otherwise provided by the constitution or statutes of the State of New Hampshire or these rules.”(emphasis added)). As such, the rule against hearsay mandates that the testimony about Reyes’ statement is inadmissible. See N.H. R. EV. 802.
Finally, we are unable to determine whether the trial court found the testimony of two crucial witnesses, John Zelonis and Donna Stafford, to be accurate and credible. Because witness credibility is a question of fact, we remand this matter to the trial court. See Red Hill Outing Club v. Hammond, 143 N.H. 284, 290 (1998).
The petitioner introduced testimony from three witnesses who claimed that Douglas had affirmed the existence of his will within one year of his death. The testimony of one witness, attorney Douglas Hatfield, was relied upon by the trial court and is not contested by the respondents. Hatfield testified that Douglas discussed the 1994 will in September 1999, and described the will as a “band-aid” that would remain in effect until *235something else was done. This statement weighs against an inference that Douglas would have preferred intestacy to his 1994 will.
The court, however, did not discuss the testimony of Zelonis and Stafford in its order. Zelonis testified that Douglas told him, on August 31, 2000, that “he had his affairs in order for years” in regards to his testamentary plans. Stafford testified that Douglas told her that his will was in effect as of June 2000, and again around September 2, 2000, after the Kings’ last major argument.
These conversations, if credible and accurate, establish that Douglas did not revoke his will up to the week before his unexpected death. If so, the petitioner has demonstrated that the will was more likely lost rather than destroyed, because the record contains no evidence after the September 2, 2000 argument indicating that Douglas’ relationship with his family changed. See, e.g., Fuller, 119 N.H. at 135-36; In re Hoffman’s Estate, 290 P.2d 669, 673 (Cal. Dist. Ct. App. 1955) (holding that a missing will must be presumed lost when the testator “at all times referred to his will as being in existence and at no time made any indication of a desire to revoke or destroy it”).
If, conversely, the trial court finds that some or all of this testimony was incredible or inaccurate, it must determine whether it is more likely than not that Douglas revoked his will after the September 1999 conversation with Hatfield. See Dionne v. Dionne, 129 N.H. 638, 639 (1987) (credibility and weight of evidence is normally a question of fact).
In making this determination, the probate court should first consider the presence of the copy and codicil in Peterson’s office and the conversation with Hatfield as some evidence of Douglas’ intentions. In the absence of any direct evidence of revocation or a new will, the probate court should then consider whether the respondents’ evidence indicates that the relationship between Douglas and his family changed to such an extent during his last year that Douglas most likely revoked his will in contravention of the plan he described to Hatfield. As we have stated:
[Ejvidence that the deceased maintained the same relationship to the beneficiaries as [he] had at the time the will was executed, and that nothing occurred which would be likely to change [his] mind with regard to the legatees is evidence that tends to negate revocation and tends toward nonrevocation.
Fuller, 119 N.H. at 135.
Having found error in the probate court’s order, we therefore reverse its decision and remand for proceedings consistent with this opinion.
Reversed and remanded.
*236BROCK, C.J., and Nadeau and Dalianis, JJ., concurred; BRODERICK, J., concurred specially. Broderick, J.,concurring specially. I agree with the majority that this case should be reversed and remanded. However, I would do so solely because the probate court admitted the “double hearsay” testimony involving Andrew Reyes. Since it is apparent that this testimony contributed to the probate court’s factual finding that “it is as probable as not that DougQas] destroyed the 1994 will,” the probate court should reweigh the evidence absent this impermissible testimony.
The majority correctly observes that the probate court did not make findings on the credibility and accuracy of the testimony of certain witnesses. While such findings would ease appellate review and may, in fact, have been advisable, I disagree that the probate court was required to expressly “discuss the testimony of Zelonis and Stafford in its order.” Nor do I believe that the probate court was obligated to furnish individual assessments of their credibility.
There is no indication that the probate court failed to properly perform its obligation to review and evaluate the evidence in its entirety and, in the course of that exercise, consider witness credibility and accuracy. On appeal, our inquiry should be whether a reasonable fact finder, based on the entire record, could reach the conclusion expressed by the probate court here. See RSA 567-A:4 (1997); In re Estate of Locke, 148 N.H. 754, 755 (2002) (supreme court will not disturb probate court’s decree unless it is unsupported by the evidence or plainly erroneous as a matter of law).
I do not support the majority’s conclusion that, as a matter of law, if the testimony of John Zelonis and Donna Stafford is favorably credited by the probate court on remand, it will automatically mean that “the petitioner has demonstrated that the will was more likely lost rather than destroyed.” While that conclusion seems much more probable, especially with the “double hearsay” testimony of Reyes excised, the ultimate fact-finding role should remain with the probate court. In fact, the probate court may well have accepted their testimony in deciding this case.
The issue is not whether King destroyed his will before September 2, 2000, but whether he destroyed it before he died on September 10, 2000. The evidence revealed that King and his wife had a very bitter argument on September 2, 2000, after which “[King] wrote a missive to himself lamenting that his ‘life had come to an end.’” Again, while the likelihood of non-destruction increases with credible testimony from Zelonis and Stafford, it does not reach the point where this court should be making the ultimate factual finding as a matter of law.
*237Finally, the majority states that the probate court used an incorrect legal standard in evaluating whether the petitioner had met its burden of proof that the will was more likely lost rather than destroyed. See In re Estate of Fuller, 119 N.H. 132, 135 (1979). I think that the majority has parsed the probate court’s decision too finely. I believe the language of the court’s order, cited by the majority, is merely an observation that the petitioner was not able to establish one way or the other, what had become of the 1994 will. I do not believe a “wrong test” was applied. Since it was the petitioner’s burden to establish that the will was more likely lost rather than destroyed, the court merely noted that it had not been able to meet its burden, directly or through inference. In sum, it could not account for the will’s non-production.