dissenting:

I.

Wherever the truth may lie, this case represents a family tragedy. It may also constitute a grave injustice. Allegations of sexual misconduct by one family member against another pose exceptional difficulties for the judicial system. In many instances, no one but the alleged perpetrator and the alleged victim is able to testify regarding what did or did not happen. The cases are exceedingly troublesome for all concerned, including judges and jurors. On the one hand, we are becoming increasingly aware of the frequency with which sexual abuse occurs and remains hidden. On the other, we have learned, painfully, that the passions and pressures that can arise within a family sometimes cause troubled individuals to make false accusations. Judges and juries must make decisions with the gravest consequences for the fives of those involved on the basis of limited information, the reliability of which is often open to serious question.

These problems are especially pronounced where the alleged victim is a child. The sexual molestation of children is unfortunately just a facet of child abuse in general, a facet that appears to represent an all too common phenomenon. While the difficult problem of determining whether the victim of alleged sexual misconduct consented to the act involved does not arise in cases involving children, other unique complications exist. Some studies show that young children are more vulnerable to suggestion than adults and that when describing past “events” they are often unable to separate fantasy from reality. Maryland v. Craig, 497 U.S. 836, 868 , 110 S.Ct. 3157, 3175 , 111 L.Ed.2d 666 (1990) (dissenting opinion of Justice Scafia joined by Justices Brennan, Marshall, and *1350 Stevens) (citing Lindsay & Johnson, Reality Monitoring and Suggestibility: Children’s Ability to Discriminate Among Memories from Different Sources, in Children’s Eyewitness Memory 92 (S. Ceci, M. Toglia & D. Ross eds. 1987); Christiansen, The Testimony of Child Witnesses: Fact, Fantasy, and the Influence of Pretrial Interviews, 62 Wash.L.Rev. 705, 708-11 (1987)); see also Cohen & Hamick, The Susceptibility of Child Witnesses to Suggestion, 4 Law & Hum.Behav. 201 (1980). 1 Children are also more susceptible to manipulation than are adults — particularly by their parents and others in positions of authority or trust. In some cases a molester may use this weakness to his or her advantage and try to keep the truth from coming out. However, in others, the result may be that an innocent person is falsely labelled a molester. Some of the investigative techniques experts use to bring the facts of child molestation to light can themselves be a source of unreliability if not properly employed. 2 The unfortunate fact is that, because of their greater suggestibility, young children sometimes testify to events that never happened.

If courts make the wrong decisions in child molestation cases, the consequences are horrendous. In civil cases judges are frequently called upon to decide whether a child should be taken away from a parent. Leaving the child in the custody of a molester may damage the child irreparably. More and more, we are finding that persons who commit egregious crimes when they are adults were themselves the victims of child abuse, frequently at the hands of one or both of their parents. The pattern is self-perpetuating. However, depriving a child of a parent, and a parent of a child on the basis of false charges manufactured by a vindictive third-party is among the cruelest and most unjust acts society can perform. In criminal cases, the consequences are at least as drastic. The defendant’s life is ordinarily ruined as the result of a conviction, and the risks to his physical well-being if he is incarcerated are significant. Even hardened criminals do not readily tolerate child abusers. Yet, an acquittal of a guilty person will in many cases mean that the abuser will resume abusing his or her victim, and possibly other children as well.

Because the consequences of an erroneous decision are so grave, courts must exercise the greatest possible care in their efforts to reach the correct decision in child molestation cases. There is a particular duty imposed on courts — if only by virtue of the state’s special obligation to children — to use every appropriate technique to ensure that the truth is ascertained. Under some circumstances, this may involve providing for psychological or physical examinations of the child or the alleged molester at public ex *1351 pense, and even for extensive studies in a proper medical setting. When a jury trial is involved, the court should insure that the jury has the benefit of whatever aid or assistance the court can properly provide the jurors in arriving at the right result. At the very least it should make certain that the jury is instructed in the fullest, fairest and most informative manner possible, and that it gives whatever instructions would help the jurors to ascertain the truth.

II.

The case before us is one in which we will probably never know the truth with certainty. We can only hope that on this occasion the jury system worked as intended and that the jury returned the correct result. However, the degree of confidence we can have in any such verdict depends in part on the extent to which the instructions provided the jurors with the guidance and assistance they required, and to which they were entitled. Here, that confidence is, unfortunately, substantially diminished. For, in my opinion, the trial court failed to afford the necessary guidance to the jurors — guidance that would have enabled them to make a decision that would have had a significantly greater chance of matching the truth. Perhaps the jury reached the correct result anyway. As one who bears a responsibility for the ultimate consequences, I certainly hope so.

Thomas McGravey was accused of sexually molesting his nephew, Michael, when the boy was seven and again when he was nine. McGravey was convicted, in Guam, by a jury. McGravey maintains that Michael fabricated the charges at the instigation of his mother, McGravey’s sister, with whom McGravey has had a bitter family feud. At trial the existence of the feud was revealed as well as the fact that the boy had once before lied when his mother’s interests were at stake. He had falsely told school officials that strap marks inflicted on him by his mother were caused by a fall from a ladder.

There was no corroborating evidence supporting Michael’s story that he was molested by his uncle or even that he was molested at all. There is no indication that he was ever examined by any professional, let alone one trained to deal with the sexual molestation of children, actual or alleged. 3 No experts testified at the trial — medical, psychological, or otherwise. With the exception of a police officer who testified that she could not recall any details regarding the complaint Michael had filed along with his mother several months earlier and Michael’s mother’s boyfriend who testified as to the close relationship between Michael and his mother, everyone who testified at the trial was a member of the McGravey family. Not one of the family members offered any evidence directly relating to the incident, except that Michael’s mother testified that Michael told her about the events in question at some point after the second alleged occurrence. She also described the history of estrangement from her brother in some detail. McGra-vey’s brothers and father testified that Michael had some behavioral problems and that McGravey was not the type of person who would engage in acts of homosexual child molestation. While all of this evidence may have sketched a partial and partisan picture of some aspects of McGravey family life, it cast little light on the central question of whether McGravey had sexually molested Michael.

The only question before the jury was whether to believe the boy, then ten years old, or McGravey. The jury had to determine whether a heinous crime had in fact occurred solely on the basis of its assessment of the respective credibility of McGravey and his nephew. It is hard to imagine a case that would place more strenuous demands on our adversary system of criminal justice or a circumstance in which meticulous instruction of the jury would be more critical. Yet only the most routine pattern jury instructions were given. No special effort was made to ensure that the jury arrived at the correct result. No special instructions were tailored to the type of case involved. To the contrary, the defendant’s requests that special *1352 instructions pertinent to a case of this nature be given were rejected in their entirety. 4

The majority concludes that the standard instructions that were given to the jury were of sufficient assistance in this case — that they sufficiently protected McGravey’s rights. I strongly disagree. As an appellate court we should be vigilant in our efforts to ensure that juries are given whatever instructions are necessary to minimize the possibility that unjust verdicts will be rendered. We can do that here by requiring that in child molestation cases the court give instructions that are designed to deal with the unique problems that are presented by such cases. I would hold, in the exercise of our supervisory powers, see United States v. Rubio-Villareal, 967 F.2d 294, 297-98 (9th Cir.1992) (en banc), that in child molestation cases a trial judge should ordinarily give the jury an appropriately worded child-witness instruction — a special cautionary instruction of the type discussed in Section III infra. Moreover, where such a prosecution rests on the testimony of a single witness, I believe that the jury should be instructed to examine the evidence with particular care. 5 Because neither of these protective instructions was given here, I would reverse McGravey’s convictions.

III.

A.

The majority holds that the trial court’s refusal to give a child-witness instruction along the lines of the one requested by McGravey was not error. The relevant portion of the requested instruction reads:

The majority offers two reasons why the trial court’s failure to give the child-witness instruction was not erroneous. I disagree with both.

First, the majority claims that the essentials of the child-witness instruction were covered by the other witness credibility instructions given to the jury. This is simply incorrect. By definition general credibility instructions cannot replace a specific instruction that children are likely to be more suggestible than adults and may lack a full appreciation of the consequences of the charges they make. The majority also notes that the Manual of Model Criminal Jury Instructions for the Ninth Circuit recommends that a specific child-witness instruction not be given. 9th Cir.Crim. Jury Instr. 4.14 comment, at 52 (1992). Such a recommendation is not binding on us. We are free to hold that an instruction is required in a particular case even if it is not recommended in that informal guide. In fact, we recently reiterated that we may find an instruction erroneous, notwithstanding its presence in our pattern jury instruction book. United States v. Warren, 984 F.2d 325 , 327 n. 3 (9th Cir.1993). Here, the recommendation contained in the manual is not only not binding, it is wrong. 6

*1353 B.

In other situations we have not hesitated to hold that jurors must be given special cautionary instructions — in addition to general credibility instructions — to help them weigh the testimony of witnesses who fall into certain categories in eases where there is no corroborative evidence to support the testimony of those witnesses. We have held that it is error to fail to give a special credibility instruction regarding the testimony of witnesses granted governmental immunity; United States v. Bernard, 625 F.2d 854, 858-59 (9th Cir.1980); witnesses who are accomplices or/and government informers; United States v. Bosch, 914 F.2d 1239, 1247 (9th Cir.1990); Guam v. Dela Rosa, 644 F.2d 1257, 1259-60 (9th Cir.1980); United States v. Martin, 489 F.2d 674 , 677 n. 2 (9th Cir. 1973), cert. denied, 417 U.S. 948 , 94 S.Ct. 3073 , 41 L.Ed.2d 668 (1974); or witnesses who were drug addicts; United States v. Ochoa-Sanchez, 676 F.2d 1283, 1289 (9th Cir.), cert. denied, 459 U.S. 911 , 103 S.Ct. 219 , 74 L.Ed.2d 174 (1982); Dela Rosa, 644 F.2d at 1261 . We have recognized that these witnesses present special credibility problems and that juries should be given added guidance as to how to evaluate the testimony.

There is at least as compelling a reason for a cautionary instruction pertaining to child witnesses in sexual molestation cases. For the reasons I have explained above, I would direct that ordinarily such an instruction be given in all such cases. Nevertheless, I would permit a trial judge to decline to give the instruction in certain circumstances, if he or she has good cause for doing so. In this connection I note that as a child approaches maturity, there is less reason to be concerned about the dangers of suggestibility and the possible confusion of imagination and reality. Also, to the extent that physical or other corroborative evidence exists there may be less reason for concern over the genesis of the child’s contentions. 7 All in all, however, the giving of the instruction along with the necessary balancing ones, can do little harm and may in particular cases make a difference between the jury’s returning a just and an unjust verdict.

The majority seeks to divest itself of its responsibility to ensure that justice is done by uttering the shibboleth “we leave it to the sound discretion of the trial judge”. How many times have appellate judges failed to do their job by citing this phrase? Countless, I would suspect. Sometimes, courts of appeals must do just what the majority suggests. Other times, not. This is one occasion on which we clearly err in failing to review the district court’s determination with sufficient care.

The district judge followed an erroneous Ninth Circuit pattern jury instruction.. See discussion IIIA and n. 4, supra. In our Comments to our Model Instructions we told the district courts that no special instruction should be given. See 9th Cir. Crim. Jury Instr. 4.14 comment (1992). While the district judge was not compelled to follow our “recommendation”, it could hardly have es *1354 caped his notice that two members of our court (the same number as constitutes a majority today) had officially proclaimed that district judges should not give the instruction requested by the defendant. In fact, the prosecutor represented to him that the “(d)efendant’s proposed ‘Child as a Witness’ instruction is expressly disapproved by the Ninth Circuit Court of Appeals”. Immediately afterwards, he denied the defendant’s request.

We can hardly now pretend that all the district court did was to exercise its own sound discretion. To the contrary, the district judge applied the erroneous rule we foisted upon him. To now seek to escape responsibility for that error by claiming that we are simply relying on the “sound discretion of the trial judge” is to reveal a certain blind spot — either intellectual or moral. The argument is too reminiscent of the bureaucratic and jurisprudential fumbling, bungling and buck-passing that led to the execution of Private Slovik. 8 In reality, the responsibility for the defendant’s fate lies with us. No invocation of pat phrases or standards of review will allow us to pass it on to the district judge. Essentially, he was merely following orders — or, in- this case, “recommendations”.

C.

Turning to the facts of this case, I would hold that the judge’s failure to give a child witness instruction constituted error. There are no circumstances that suggest a reason for omitting the instruction here. To the contrary, there is even more reason to give one, since it is undisputed that the child previously lied on behalf of the very person who the defendant suggests induced him to testify to sexual molestation. Thus, there was no justification for refusing to give the requested instruction.

IV.

I believe the trial judge committed an additional error in this ease as well. The government’s proof of sexual molestation depended entirely upon the testimony of one witness. That requires a special instruction, wholly apart the fact that the witness was a young boy. Because the success of the government’s case depended wholly upon the testimony of a single witness, the jury should have been instructed to review the evidence against the defendant with particular care. This rule is a settled proposition under California law. Because much of Guam’s statutory law is modeled upon California’s, California cases are persuasive where, as here, there is a paucity of authoritative Guam deci-sional law. See Hemlani v. Guerrero, 902 F.2d 1412 , 1414 n. 3 (9th Cir.1990); Lucero v. Stewart, 892 F.2d 52, 54 (9th Cir.1989); Guam v. Ojeda, 758 F.2d 403, 406 (9th Cir. 1985).

In California v. Rincon-Pineda, the California Supreme Court unanimously held that a special instruction should be given sua sponte by the trial court to protect a defendant’s rights in cases in which the prosecution is based upon the uncorroborated testimony of a single witness. 14 Cal.3d 864 , 123 Cal.Rptr. 119, 133-34 , 538 P.2d 247, 261 (1975); California v. Pringle, 177 Cal.App.3d 785 , 223 Cal.Rptr. 214, 217 (1986), disapproved on other grounds by California v. Gammage, 2 Cal.4th 693 , 7 Cal.Rptr.2d 541 , 828 P.2d 682 (1992). That instruction, now CALJIC No. 2.27, currently reads as follows:

[hereinafter “single-witness instruction”] The majority cites a raft of boiler-plate credibility instructions in support of its conclusion that the instructions actually given were adequate. 9 However, none emphasizes the need *1355 for particularly careful review where there is only one witness. General credibility instructions are no substitute for the single-witness instruction. The Rincón-Pineda, court made clear that the single-witness instruction was an additional procedural protection for the defendant that served an important function above and beyond any general jury instructions on witness credibility. See 123 Cal.Rptr. at 133-34 , 538 P.2d at 261-62 .

A trial court’s failure to give the single-witness instruction sua sponte is error. California v. Adams, 186 Cal.App.3d 75 , 230 Cal.Rptr. 588, 590 (1986), disapproved on other grounds by California v. Gammage, 2 Cal.4th 693 , 7 Cal.Rptr.2d 541 , 828 P.2d 682 (1992); Pringle, 223 Cal.Rptr. at 217 . In this case no single-witness instruction was given and none of the other instructions, either separately or as a whole, admonished the jury to review Michael’s testimony with particular care because the government’s proof of the sexual molestation was entirely dependent upon a single witness. As in Adams and Pringle the erroneous omission of the single witness instruction was compounded by the judge’s giving a legally correct instruction that stated that “[t]he testimony of a victim of sexual conduct need not be corroborated if that victim is believed beyond a reasonable doubt.” [hereinafter “non-corroboration instruction”] 10 While it is proper to give both the non-corroboration instruction and the single-witness instruction, California v. Gammage, 2 Cal.4th 693 , 7 Cal.Rptr .2d 541, 828 P.2d 682 (1992), 11 it is improper to give the non-corroboration instruction without a single-witness instruction. Adams, 186 Cal.App.3d 75 , 230 Cal.Rptr. at 590 ; Pringle, 223 Cal.Rptr. at 217 . To achieve a proper balance between the rights of the defendant and those of the complaining witness, both instructions must be given. California v. Hollis, 235 Cal.App.3d 1521 , 1 Cal.Rptr.2d 524, 527 (1991), review granted — Cal.4th-, 4 Cal.Rptr.2d 764 , 824 P.2d 570 review dismissed — Cal.4th-, 9 Cal.Rptr.2d 92 , 831 P.2d 317 (1992). The trial court erred in not giving a single-witness instruction here.

Y.

The only remaining question is whether the trial court’s errors were prejudicial or harmless. In our en banc decision in Rubio-Villareal we stated that the government must prove that errors with respect to jury instructions are harmless beyond a reasonable doubt. 967 F.2d at 297 n. 3. I believe the trial judge’s errors were not harmless under this standard.

The evidence of guilt was not overwhelming but rather consisted solely of the testimony of a child-witness who had been known to lie for his mother’s sake in the past. There was nothing inconsistent or implausible in McGravey’s testimony that he never sexually molested his nephew and that the boy was lying at his mother’s behest. On the basis of the written record, there is no way of determining who was telling the truth — McGravey or Michael. The entire ease for conviction rested squarely upon Michael’s credibility. The trial court’s failure to give a child-witness instruction and a single-witness instruction went to the core of the jurors’ determination of guilt or innocence. Thus, errors in instructions as to how the jury should determine that credibility were highly prejudicial. I would reverse and remand for a new trial *1356 in which the omitted instructions would be given to the jury.

I respectfully dissent.

. The majority seeks to cast doubt on the theory that children are more suggestible than adults. It is difficult to understand the majority’s argument in light of recent well-publicized cases where sensational charges, later proved to be false, were made by children after they were interviewed by therapists who employed aggressive and leading questions. In the McMartin case (jury trial, no published opinion), Ray Buck-ey and Peggy McMartin Buckey, owners of the McMartin pre-school in Manhattan Beach, CA, were charged with multiple counts of rape and sodomy, acts which the children alleged took place over many months and were accompanied by the slaughtering of animals and elaborate satanic rituals. Despite wide and arguably prejudicial pre-trial publicity, both defendants were acquitted by a jury on 52 counts of child molestation; the jury was unable to reach a verdict on 13 other counts. See David Shaw, McMartin and the Media: Time McMartin Coverage was Biased, Critics Charge, Los Angeles Times, January 22, 1990 at Al. In the Akiki case (juiy trial, no published opinion), another pre-school teacher, Dale Akiki, was acquitted by a juiy on all counts, but only after spending more than two years in jail awaiting trial. Allegations in this case, again made after interviews with a large number of children, included the ritual slaying inside the pre-school of a giraffe and an elephant. See Reaction to Akiki Verdict: Many are Critical of D.A. and the Prosecutor, San Diego Union Tribune, November 24, 1993 at B7. I should state yet again that none of this is meant to sirgue that child molestation does not exist, or is not a serious problem. There is, however, also a significant threat to the liberty interests of some alleged molesters created by the suggestibility of some alleged child victims. Simply put, I am aware of no trial since those held in Salem in which a group of adults suggested occurrences similar to those alleged in the McMartin and Akiki cases.

. See, e.g., Christiansen, supra at 711, 714-15; H. Humphrey, Report on Scott County Investigation (1985).

. Whether a court should order such examinations in cases of alleged sexual abuse against children is a question not raised by the parties and, therefore, not before us.

. To a large extent this was not the trial judge's fault. The judge declined to give McGravey's requested child-witness instruction on the basis of an erroneous recommendation in our Manual of Model Criminal Jury Instructions for the Ninth Circuit. See discussion in Part III.A infra.

. I would balance these instructions with others that ensure that the testimony of the child-witness is not improperly disregarded or given insufficient weight. See discussion in Section IV infra. What is required in child molestation cases is not only adequate protection for the rights of the defendant but also fuller and more specific guidance for the juiy as to how the crucial credibility determinations should be made.

.The recommendation in the Model Criminal Jury Instructions is based upon a misreading of Pocatello v. United States, 394 F.2d 115, 116-17 (9th Cir.1968) (per curiam), a case which is simply irrelevant to the issue presented here. The question in Pocatello was whether a five year old child was competent to testify. Pocatello explicitly considered only whether a child-witness' testimony was admissible, an issue we left to the sound discretion of the trial judge. The case does not in any way address whether special credibility instructions should be given to the jury regard *1353 ing the weight that should be given the testimony of a child witness. See also Barnes v. United. States, 600 A.2d 821, 825 (D.C.App.1991) ("There is an important distinction between the competency of a child to testify and the assessment of a child witness' credibility. The former is a question of law to be decided by the court. The latter is a function for the jury.” (internal citation omitted)).

. I believe that the trial judge should have considerable latitude in formulating the exact language of a child-witness instruction in a manner that fits the particular circumstances of a given case, Barnes, 600 A.2d at 821 ; Commonwealth v. Avery, 14 Mass.App.Ct. 137 , 437 N.E.2d 242, 246 (1982). However, any child-witness instruction should convey the fact that children are more suggestible than adults. It should also note that children may have difficulties separating recollection of an imaginary event from a real memory. An appropriate child witness instruction might read:

. William Bradford Huie (1954), The Execution of Private Slovik.

. I have no disagreement with anything the majority states in Part I (or Part II) of its opinion. The majority simply fails to address the fact that the jurors were not given an instruction that told them to review the evidence with particular care because the government’s proof of sexual moles *1355 tation was entirely dependent on the uncorroborated testimony of one witness.

. The Guam non-corroboration instruction was upheld in Guam v. Welch, 1990 Crim. No. 90-00008A, 1990 WL 320365 (Guam App. Oct. 30, 1990). Welch does not answer the question whether a single witness instruction is required under Guam law in the circumstances of the present case. The opinion does not indicate whether a single-witness instruction was given or not and in Welch there was corroboration. Moreover, the Welch court relied upon California v. McIntyre, 115 Cal.App.3d 899 , 176 Cal.Rptr. 3 (1981), which upheld a non-corroboration instruction where a single-witness instruction was given. Thus, the majority errs when it states its decision here comports with Welch ; that decision was based upon cases which require the giving of a single. witness instruction.

. Gammage disapproved statements in Adams and Pringle that only the single-witness instruction should be given. Gammage reaffirmed that both the single-witness instruction and the non-corroboration instruction should be used.