with whom MACK, Senior Judge, joins, dissenting:
For the reasons expressed in the vacated division opinion, In re Melton, 565 A.2d 635 (D.C.1989), vacated, 581 A.2d 788 (1990), I would reverse and remand the ease to the trial court with instructions to vacate Melton’s civil commitment order.
I.
The first issue is how a trial court is to determine whether the “facts or data ... upon which an expert bases an opinion or inference ... [is] of a type reasonably relied upon by experts in the particular field_” Fed.R.Evid. 703. Everyone agrees that the court may not defer entirely to the proffered experts’ own assessments of reliable hearsay.1 So how does the court test those underlying sources?
In the division opinion, Melton, 565 A.2d at 643-45, we noted two extreme views. According to the Third Circuit, considered the so-called liberal admissibility view, the trial court “must make a factual inquiry and finding as to what data experts in the field find reliable. There is no discretion to forbear from making this inquiry and finding.” In re Japanese Elec. Prods. Antitrust Litig., 723 F.2d 238, 277 (3d Cir.1983), rev’d on other grounds, 475 U.S. 574, 106 S.Ct. 1348, 89 L.Ed.2d 538 (1986). Under this approach, the court only considers whether experts customarily rely on the proffered data; no analysis is made of the reliability of the underlying data itself. The court leaves vigorous examination of the assumptions that form the basis for the opinion — including scrutiny of the underlying data — to cross-examination. See id.
At the other extreme, the “more restrictive view requires the trial court to determine not only whether the data are of a type reasonably relied upon by experts in the field, but also whether the underlying data are untrustworthy for hearsay or other reasons.” In re “Agent Orange” Prod. Liab. Litig., 611 F.Supp. 1223, 1244 (E.D.N.Y.1985), aff'd, 818 F.2d 187 (2d Cir. 1987), cert. denied, 487 U.S. 1234, 108 S.Ct. 2898, 101 L.Ed.2d 932 (1988); see Soden v. Freightliner Corp., 714 F.2d 498, 505 (5th Cir.1983) (Rule 703 requires trial court to examine reliability of expert’s sources).
Consistent with the division opinion, the majority and I agree with Judge Weinstein, taking an intermediate position, that the trial court not only must make findings as to what data experts in the field find reliable but also must do something more to assure such reliance is reasonable. According to Judge Weinstein:
[T]he court may not abdicate its independent responsibilities to decide if the bases [used by a qualified expert] meet minimum standards of reliability as a condition of admissibility. See Fed.R.Evid. 104(a). If the underlying data are so lacking in probative force and reliability that no reasonable expert could base an opinion on them, an opinion which rests entirely upon them must be excluded. Fed.R.Evid. 401, 402.
“Agent Orange, ” 611 F.Supp. at 1245; see ante at 902, 904-905.
The question, therefore, is what more the trial court must do to assure that expert reliance on particular data is reasonable— i.e., to assure that “minimum standards of reliability,” 611 F.Supp. at 1245, have been met? The majority opinion appears to incorporate Judge Weinstein’s views in “Agent Orange” and leaves it at that. Based on the cases Judge Weinstein cited in support of his “minimum standards of reliability” test,2 his view is closer to the *910more restrictive, Fifth Circuit approach in Soden than to the liberal Third Circuit view in Japanese Electronic Products.
The division opinion — with which I still agree — supplied the following, more specific test for determining that the “minimum standards of reliability” have been met:
(2) the judge [must] conclude[ ] that the information (if not admissible for its truth under an exception to the hearsay rule) is of a type for which the underlying reliability of the data can be sufficiently explored through cross-examination of the testifying expert.
Melton, 565 A.2d at 645 (citing S. Saltzburg & K. Redden, Federal Rules of Evidence Manual 671 (4th ed. 1986)). Under this test for the most appropriate and efficient way of evaluating whether an expert’s data meets “minimum standards” and thus is “reasonably relied upon,” Fed. R.Evid. 703, the court must assure that the Third Circuit’s premise in Japanese Electronic Products, 723 F.2d at 277, is sound as applied: that the expert knows enough about his or her sources of information, including how the data has been gathered and analyzed, that cross-examination can effectively explore the bases — including deficiencies — of the expert opinion.
The majority’s (and the government’s) concern about this “cross-examination” test to assure “minimum standards of reliability” apparently is attributable to the following two paragraphs in the division opinion:
Where experts rely on lay observations of behavior or events, however, cross-examining experts about the reasonableness of their reliance on this data may be more difficult. Absent the ability of an expert to testify concerning both the manner in which a lay observer has perceived and recorded certain facts and whether that manner accords with methods of fact-gathering ordinarily relied upon, cross-examination may provide little or no insight into the reliability of the basis for the expert opinion. In such circumstances a trial judge, therefore, may reasonably conclude that a jury will be incapable of performing its function of “determining the appropriate weight to be given to the testimony,” Japanese Electronic Products, 723 F.2d at 278, and exclude the expert opinion. See Note, Hearsay Bases [of Psychiatric Opinion Testimony: A Critique of Federal Rule of Evidence 703, 51 S.Cal. L.Rev. 129, 145 (1977)] (“[A]n opinion can be no better than the hypothesis or assumption upon which it is based.”).
We are not stating categorically that expert reliance on hearsay observations by laypersons will always fail to satisfy the second prong of the test set forth supra at page 645. But, given the possibility that lay observers in some contexts (ie.g., a family with a mentally ill member) may not be disinterested, and given, further, that the expert’s inability to vouch for the lay observer's reliability will not necessarily undermine the expert’s own aura of authority, a trial court must be very careful to evaluate whether the proffered expert testimony and its underlying sources can be effectively scrutinized before the jury.
Melton, 565 A.2d at 646 (footnotes omitted). This discussion in Melton does not say that expert opinion based in part on “lay observations of behavior or events,” id., is automatically excluded. It is not even intended to mean that expert opinion based entirely on such lay observations is automatically excludable, although it will be suspect (and very likely excludable) unless the expert can satisfactorily explain why under the circumstances such reliance is reasonable — surely not too much to expect. The point is this: before an expert testifies, the court must be able to know that counsel, on cross-examination, will be able to get to the heart of the expert’s analysis and that the jury will hear a more *911reasonable basis for reliance on hearsay than merely “trust me, I’m an expert.”
Judge SCHWELB concludes that “[t]he panel majority’s second prong would effect a substantial and in our view unnecessary intrusion upon the expert’s sources.” Ante at 906. I do not understand why that is true, or indeed why the division test is more intrusive than the verification-of-reliability efforts reflected in the cases Judge Weinstein cites in “Agent Orange,” see supra note 2, which the majority apparently accepts in simply adopting the Weinstein approach without elaboration.
Presumably, the majority requires the trial court to probe the experts to some extent about the reliability of the underlying data. Presumably, too, if the experts are not immediately convincing, the court will insist on deeper probing, perhaps with the help of counsel; for anything less would abdicate the trial court’s independent responsibility. See “Agent Orange, ” 611 F.Supp. at 1245. Such probing, as needed, is no less intrusive — or, as far as I can tell, materially more difficult — than the test in the vacated division opinion requiring the trial court to assure that “the underlying reliability of the [hearsay] data can be sufficiently explored through cross-examination of the testifying expert.” Melton, 565 A.2d at 645. Thus, either the majority effectively agrees with the division opinion on this issue (despite denying it) or does not come to grips with the need to assure the “minimum standards of reliability” (as it purports to do). I cannot tell from Judge Sohwelb’s opinion where the majority really stands.3
II.
The second issue is whether the testifying psychiatrists, Dr. Byrd and Dr. Cornet, were qualified to render expert opinions on Melton’s dangerousness to himself or to others. I agree that psychiatrists commonly testify about a person’s dangerousness — and properly so. See Barefoot v. Estelle, 463 U.S. 880, 896-903, 103 S.Ct. 3383, 3396-3399, 77 L.Ed.2d 1090 (1983); ante at 898-899. I also agree that, under applicable caselaw, a proffered expert need not be a specialist in order to testify concerning matters requiring special expertise. See Ornoff v. Kuhn and Kogan Chartered, 549 A.2d 728, 732 (D.C.1988); Baerman v. Reisinger, 124 U.S.App.D.C. 180, 181, 363 F.2d 309, 310 (1966); ante at 897-898. But the flipside is this: whatever the expert’s formal qualifications, the trial court must make sure that the proffered expert is qualified by training, knowledge, and experience to express an opinion on the subject about which he or she is called to testify. See United States v. Davis, 772 F.2d 1339, 1342-44 (7th Cir.) (psychiatrist’s testimony properly excluded where she lacked expertise on compulsive gambling), cert. denied, 474 U.S. 1036, 106 S.Ct. 603, 88 L.Ed.2d 581 (1985); see generally Melton, 565 A.2d at 647-49. We have recently stressed this point: “ ‘it is the actual qualifications of the witness that count, rather than his [or her] title.... Just as the wrong title may mean that the witness is nevertheless qualified, the right title will not suffice if the witness does not have the qualifications required by the facts of the case.’ ” District of Columbia v. Anderson, 597 A.2d 1295, 1299 (D.C.1991) (quoting 3 J. Weinstein & M. Berger, Weinstein’s Evidence ¶ 702[04] at 702-51 (1990)).
In this case, the trial court refused to consider Melton’s counsel’s objections to the psychiatrists’ qualifications (saying “[w]e’ve been through that before”), did not purport to exercise discretion, see Johnson v. United States, 398 A.2d 354, 363 (D.C.1979), and thus, for all we can tell, declined to apply the criteria for admitting expert testimony. See Ibn-Tamas v. United States, 407 A.2d 626, 632-33 (D.C.1979) *912(quoting three-part test for expert admissibility in Dyas v. United States, 376 A.2d 827, 832 (D.C.), cert. denied, 434 U.S. 973, 98 S.Ct. 529, 54 L.Ed.2d 464 (1977)); Japanese Elec. Prods., 723 F.2d at 277 (trial court has “no discretion to forbear from making ... inquiry and finding” as to “what data experts in the field find reliable”); see generally Melton, 565 A.2d at 648-49. The majority, however, makes its own findings from the record. Judge Schwelb’s opinion relies on Dr. Byrd’s and Dr. Cornet’s qualifications, based on long relationships with Melton, to justify their testifying about his dangerousness to himself from not taking prescribed medications. See ante at 898. I have two responses. First, the trial court, not the appellate court, must rule in the first instance; this court should not find facts. Second, even if the doctors were qualified to testify about Melton’s danger to self, their opinions of his dangerousness focused as much, if not more, on Melton’s dangerousness to others — an issue at the heart of Melton’s civil commitment, an issue on which not all psychiatrists believe they are qualified to testify, see Estelle v. Smith, 451 U.S. 454, 472, 101 S.Ct. 1866, 1878, 68 L.Ed.2d 359 (1981),4 and an issue which the majority does not address as to these particular psychiatrists. As Judge Schwelb recognizes, the jury merely found Melton dangerous to self or to others. See ante at 894 n. 1. We therefore cannot be sure the jury did not focus exclusively on dangerousness to others — a possibility that makes an expert’s qualifications to testify on danger to others critically relevant.5
In a recent post-Melton probable cause hearing, D.C.Code § 21-525 (1989), the government’s psychiatrist and only witness stated that he could not say there was a 50% or 70% chance that the respondent in that case, or in any case, was likely to be dangerous to self or others. This psychiatrist testified that psychiatrists are not generally capable of predicting future dangerousness with any accuracy. See Transcript at 8, In re D.L., MH No. 86-90 (D.C.Super.Ct. Jan. 30, 1990).6 In another recent case, the government’s psychiatrist at a probable cause hearing reluctantly expressed an opinion about the respondent’s dangerousness, noting that it was a very complicated field and the subject of some controversy and that his ability to express such an opinion varied from case to case. See Transcript at 10-11, In re M.G., MH No. 155-90 (D.C.Super.Ct. Feb. 14, 1990).7
*913The legislature has, of course, decided that psychiatric evaluation will be useful and sometimes necessary in civil commitment decisions, but I see nothing in the legislative scheme indicating that every psychiatrist, by virtue of a credential as such, is qualified to testify as to a particular person’s dangerousness to self or others. The trial court accordingly abused its discretion in failing to verify that Dr. Byrd and Dr. Cornet were qualified to opine on Melton’s dangerousness.
Respectfully, therefore, I dissent as to both issues presented.
. Amicus (the United States) notes it "agree[s] with appellant that ‘the court cannot ... simply ... accept[ ] as reasonable reliance what the expert decided to rely upon.’ ’’ Amicus Brief at 5 (quoting Appellant’s Brief at 22).
. The cases cited, 611 F.Supp. at 1245, are: United States v. Esle, 743 F.2d 1465, 1474 (11th Cir.1984) (expert opinion based on untrustworthy market surveys); Barrel of Fun, Inc. v. State Farm Fire & Casualty Co., 739 F.2d 1028, 1033 (5th Cir.1984) (unreasonable for expert to rely upon voice stress analysis); United States v. Cox, 696 F.2d 1294, 1297 (11th Cir.) (affirming exclusion of expert testimony that was based on hearsay knowledge of historical events not reasonably relied upon by experts in the field), cert. denied, [464 U.S. 827, 104 S.Ct. 99, 78 L.Ed.2d 104 (1983)]; Soden v. Freightliner Corp., 714 F.2d 498, 503-06 (5th Cir.1983) (exclusion of expert opinion *910based on unreliable accident statistics): Dallas & Mavis Forwarding Co., Inc. v. Stegall, 659 F.2d 721, 722 (6th Cir.1981) (testimony of state trooper offered as an expert regarding cause of accident based on story of biased eye witness "the sort of hearsay testimony” Rule 703 meant to foreclose); Toys "R" Us, Inc. v. Canarsie Kiddie Shop, Inc., 559 F.Supp. 1189, 1205 (E.D.N.Y.1983) (exclusion of expert opinion based on surveys lacking “sufficient indicia of trustworthiness” under Rule 703).
. Judge Schwelb finds it "unnecessary and even fruitless to attempt to detail further, in the abstract, the minimal reliability standard." Ante at 906, n. 27. He then engages in appellate court fact-finding to ascertain — by mere assertion — that information provided by Melton's relatives and his hospital records “meets minimal standards of reliability.” Id. If this is so obvious, then presumably counsel could assure the court and jury of that fact through cross-examination.
. As the Supreme Court recognized in Estelle v. Smith, 451 U.S. 454, 472, 101 S.Ct. 1866, 1878, 68 L.Ed.2d 359 (1981):
Indeed, some in the psychiatric community are of the view that clinical predictions as to whether a person would or would not commit violent acts in the future are "fundamentally of very low reliability" and that psychiatrists possess no special qualifications for making such forecasts. See Report of the American Psychiatric Association Task Force on Clinical Aspects of the Violent Individual 23-30, 33 (1974); A. Stone, Mental Health and Law: A System in Transition 27-36 (1975); Brief for American Psychiatric Association as Amicus Curiae 11-17.
. Judge Schwelb responds that "[t]here is nothing in this record to suggest that the two witnesses were any less qualified to discuss” Melton’s dangerousness to others than dangerousness to himself. Ante at 900 n. 9. Accordingly, he implicitly acknowledges that the government has proffered no direct evidence that the doctors were qualified to testify about future dangerousness. Furthermore, our caselaw demands that the proffering party must show the expert has sufficient "skill, knowledge, or experience” in the field to demonstrate that his or her opinion "will probably aid" the fact-finder; the law does not require the other party to provide evidence that the expert is unqualified. Dyas, 376 A.2d at 832. Judge Schwelb, therefore, has the burden of proof reversed by suggesting — erroneously—that if an expert is qualified to address dangerousness to self, he or she must be deemed qualified to opine on dangerousness to others, absent rebuttal to the contrary.
. Dr. Donald Taylor testified:
I don’t think I have said ‘I think, 50-percent chance or 70-percent chance this person will do this or the other thing,’ and I — I'm quite willing to concede that I don’t think that that’s within the state of the art to predict over a long period of time, more than a few days of time, with statistical accuracy, that is, with anything approaching more than 50/50. Even that perhaps is doubtful, so I concede that. I think it's very difficult to predict future behavior beyond a day or two.
In re D.L., Transcript at 8.
. Dr. Andrew Schwartz testified that predicting dangerous behavior:
var[ies] very greatly, from case to case. I personally feel that the whole question of *913prediction of dangerous behavior is a very uncertain one.... [I]t’s very difficult to make predictions about dangerousness or about human behavior in general in a predictive sense. But I think that within the spectrum of uncertainty, if you will, that there are some patients with whom one can be relatively more confident than others.
In re M.G., Transcript at 11.