We must decide whether the testimony of two experts on child sexual abuse impermissibly intruded on the jury’s role of assessing the credibility of the child complainants. In Commonwealth v. Dockham, 405 Mass. 618 (1989), we held that expert testimony on the general behavioral characteristics of sexually abused children was properly admitted where the defendant had argued that such testimony should have been admitted only in response to hypothetical questions based on the facts in evidence. Here, the defendant takes the antipodal position, arguing that the testimony given in response to hypothetical questions closely tailored to the facts amounted to *845an impermissible endorsement of the credibility of the complaining child witness. We conclude that the challenged expert testimony of one of the experts was improper in several respects, and that admission of the testimony was not harmless error.
1. In December, 1988, a Superior Court jury convicted the defendant on seven indictments charging rape of a child under sixteen years and three indictments charging assault and battery on a child under the age of fourteen years. Federico filed a motion for a new trial, which was denied in August, 1994. He simultaneously appealed from his convictions, arguing that the judge improperly allowed two experts to bolster the credibility of the complainants’ claims of sexual abuse. The Appeals Court reversed, Commonwealth v. Federico, 40 Mass. App. Ct. 616, 618-619 (1996), ruling that the testimony of both experts impermissibly intruded on the jury’s role of assessing the credibility of the complainants, and that the admission of the testimony was prejudicial error.1 We granted the Commonwealth’s application for further appellate review.2 We reverse the judgment of the Superior Court.
2. The jury could have found the following facts. In 1981, a mother, whom we shall call Maria, and her two daughters, whom we shall call Julie (then aged eleven years) and Sharon (then aged nine years), began living with the defendant shortly before the defendant and Maria were married. That same year the defendant began sexually abusing both sisters; the abuse continued chronically for the next six years.3 Neither sister revealed the abuse to anyone until March, 1988, when first Sharon, and subsequently Julie, told their mother about the abuse. A few days later Maria took Sharon, and then Julie, to a *846physician for an examination. A police officer and a detective also interviewed the sisters at home.
At the time of the trial in 1988, Julie and Sharon were nineteen and sixteen years old, respectively. They testified in detail about the sexual abuse to which they had been subjected by the defendant. The Commonwealth called as fresh complaint witnesses the detective and police officer who had interviewed the sisters, as well as their mother. The Commonwealth also called two experts to testify, Dr. Renee Brant, a child psychiatrist, and Dr. Jan Paradise, a pediatric gynecologist. Neither expert had treated either of the children. The prosecutor asked Dr. Brant a series of hypothetical questions in which she was asked to assume certain facts; in each case one of the facts she was asked to assume was that sexual relations had occurred between one or two girls and an adult male. She was then asked to opine as to whether the assumed facts were “consistent with” children who had been having sexual relations with an adult male (stepfather) over a number of years. In response to each hypothetical question, Dr. Brant opined that the assumed facts were “consistent with” children who had been having sexual relations with an adult.
There was evidence that there was no physical trauma to the genital area of either child. Dr. Paradise testified in response to a series of hypothetical questions. In each case she was asked to assume certain facts, one of which was vaginal penetration of a girl by an adult male, and another was the absence of physical signs of trauma to the genital area of the girl. The remaining assumed facts (different in each hypothetical question) concerned the ages of the girls, frequency and the timing of intercourse, and the timing of the physical examination of the girls in relation to the most recent act of intercourse. Dr. Paradise was then asked whether the assumed facts were “inconsistent” with each other. She opined that the assumed facts were not “inconsistent.”4
3. The defendant claims that the admission of the expert testimony of both Dr. Brant and Dr. Paradise impermissibly intruded on the jury’s role of assessing the credibility of the child complainants and “would inescapably have the same impact on the jury as a direct reference to and comparison with the child witnesses in the case.” Commonwealth v. Perkins, 39 *847Mass. App. Ct. 577, 583 (1995). The Commonwealth responds that the testimony was admissible because courts previously have allowed experts to testify about typical symptoms of sexually abused children, information which is beyond the knowledge of jurors. The Commonwealth not only asks that we affirm the convictions, but also argues that Commonwealth v. Perkins, supra, has been “wrongly decided.”5 In this case the expert testimony was particularly important, says the Commonwealth, because the two adolescent sisters did not disclose the sexual abuse for years, they remained at home with their abuser while the abuse was continuing, and they developed symptoms of anxiety and stress only after they had disclosed the abuse. This behavior, left unexplained, might unfairly suggest to a lay person that no abuse had occurred.6 See Commonwealth v. Hudson, 417 Mass. 536, 540 (1994) (expert testimony on typical symptoms of sexually abused children allowed because information is beyond common knowledge of jurors and may be of assistance in assessing witness’s credibility).
It hardly needs repeating that a trial judge has broad discretion to determine whether to admit expert testimony. See Commonwealth v. Colin C., 419 Mass. 54, 59 (1994); Commonwealth v. Dockham, supra at 628. Expert testimony “is admissible whenever it will aid the jury in reaching a decision, even if the expert’s opinion touches on the ultimate issues that the jury must decide.” Commonwealth v. Dockham, supra, quoting Simon v. Solomon, 385 Mass. 91, 105 (1982). See Commonwealth v. Colin C., supra at 59; Commonwealth v. Pikul, 400 Mass. 550, 553 (1987). See also Proposed Mass. R. Evid. 702 (“If scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education, may testify thereto in the form of an opinion or otherwise”). Under this principle, we have held that testimony on the general behavioral characteristics of sexually abused children may properly be the subject of expert testimony because behavioral and emotional characteristics common to these victims are “beyond the jury’s common knowledge and *848may aid them in reaching a decision.” Commonwealth v. Colin C., supra at 60. See Commonwealth v. Dockham, supra at 630. See also Commonwealth v. Day, 409 Mass. 719, 724 (1991) (expert testimony concerning “battered child syndrome” admissible because condition is not matter of common knowledge); Commonwealth v. Mamay, 407 Mass. 412, 421 (1990) (rape trauma syndrome is “beyond the jury’s common knowledge,” and therefore admissible). Such expert testimony may assist the jurors in assessing a victim witness’s testimony and credibility:
“While jurors may be capable of personalizing the emotions of victims of physical assault generally, and of assessing witness credibility accordingly, tensions unique to the trauma experienced by a child sexually abused by a family member have remained largely unknown to the public. As the expert’s testimony demonstrates the routine indicia of witness reliability — consistency, willingness to aid the prosecution, straightforward rendition of the fact — may, for good reason, be lacking. As a result jurors may impose standards of normalcy on child victim/ witnesses who consistently respond in distinctly abnormal fashion.”
Commonwealth v. Dockham, supra, quoting State v. Middleton, 294 Or. 427, 440 (1983) (Roberts, J., concurring).7
Such evidence must, however, be confined to a description of the general or typical characteristics shared by child victims of sexual abuse. Deference must be preserved for the role of the jury as the final judge of credibility; “[e]valuations of credibility are, of course, within the exclusive province of the trier of fact,” Commonwealth v. Montanino, 409 Mass. 500, 504 (1991), quoting Commonwealth v. Ianello, 401 Mass. 197, 202 (1987), and “witnesses may not offer their opinions regarding the credibility of another witness.” Commonwealth v. Mon*849tonino, supra at 504. See Commonwealth v. Sires, 413 Mass. 292, 304 (1992).8
We have identified several broad areas of expert testimony that are most likely to run afoul of this prohibition. First, we prohibit an expert from directly referring to or comparing the behavior of the specific child complainant to general characteristics of sexual abuse victims because “[sjuch testimony impermissibly intrudes on the jury’s province to assess the credibility of the witness.” Commonwealth v. Trowbridge, 419 Mass. 750, 759 (1995) (“[ajlthough expert testimony on the general behavioral characteristics of sexually abused children is permissible, an expert may not refer or compare the child to those general characteristics”).9 Commonwealth v. Colin C., supra at 60. See Commonwealth v. Richardson, 423 Mass. 180, 185 (1996) (harmless error to allow expert to testify, after child victim had given inconsistent testimony on sequence and dates of abuse, that such inconsistency was not unusual among child victims); Commonwealth v. Dockham, supra at 628, 630 (expert testimony admissible where there were “no references or comparison to the child witness”). The risk of improper comparisons between any general behavioral characteristics of sexually abused children and a particular complaining child witness is most acute when the expert witness has examined or treated the child. Testimony on the general characteristics of sexually abused children by such experts has been disallowed. See Commonwealth v. Trowbridge,, supra at 759-760; Commonwealth v. Rather, 37 Mass. App. Ct. 140 (1994). See also Commonwealth v. Swain, 36 Mass. App. Ct. 433, 444-445 (1994).
Second, the expert may not directly opine on whether the alleged victim was in fact subject to sexual abuse. Commonwealth v. Colin C., supra at 61 (expert testimony inadmissible where opinion that children were sexually abused “was tantamount to *850an expert opinion that the . . . claims . . . were likely true”); Commonwealth v. Trowbridge, supra (error to admit expert testimony that child’s behavior during gynecological visit was consistent with that of sexually abused children).
Third, with respect to the accused, the expert may not provide profiles or testify as to the typical attributes or characteristics of the perpetrators of child abuse. Commonwealth v. Day, supra at 724 (testimony concerning “battered child syndrome” admissible, but testimony concerning “profile” of child batterers is not). See Commonwealth v. Goetzendanner, 42 Mass. App. Ct. 637, 644 (1997).10 We examine the testimony of the two experts in this case with those principles in mind.
In this case, Dr. Brant, a child psychiatrist, testified in response to a series of hypothetical questions, the assumed facts of which were based on the facts in evidence, as a hypothetical question must be.11 In the case of each hypothetical question, however, Dr. Brant was asked to “assume” that the children had sexual relations with their stepfather; she was then asked to opine on whether the “facts are consistent with the children who have had sexual relations at those ages with an adult.” Dr. *851Brant opined that the behavior described in each case was consistent with the behavior of children who had been sexually abused.
Dr. Brant could hardly have done otherwise, for the hypothetical question was tautological: Dr. Brant was asked to opine on whether sexual relations between a child and an adult was “consistent” with an assumed fact that sexual relations between the child and the adult stepfather had occurred. It could well be that the prosecutor intended the jury to understand that each of the other assumed facts (for example, the delay in reporting the abuse) could co-exist with sexual abuse. But there is a significant risk that the jury could have understood Dr. Brant’s testimony to mean that sexual abuse had in fact occurred in this case. That is impermissible. See Commonwealth v. Colin C., supra at 60-61. Because the only evidence against the defendant were the allegations of the complainants, at the very least Dr. Brant’s testimony tended to support their allegations and to buttress their credibility. We conclude that the testimony in response to the flawed hypothetical questions asked of Dr. Brant should not have been admitted.
In the absence of evidence of physical injury,12 a medical expert may be able to assist the jury by informing them that the lack of such evidence does not necessarily lead to the medical conclusion that the child was not abused.13 Commonwealth v. Allen, 40 Mass. App. Ct. 458, 465-466 (1996). See Commonwealth v. Dockham, supra at 621 (expert testimony that absence of trauma to genital area not inconsistent with sexual *852abuse admitted without challenge or comment on plenary review). Medical experts, like other experts, may not opine as to whether a particular child has been sexually abused. Commonwealth v. Colin C., supra at 60 (we have never allowed “an expert to testify that an alleged victim was in fact sexually assaulted”). See Commonwealth v. Montmeny, supra at 528 (direct opinion by medical expert that rape had occurred would have been “beyond the witness’s appropriate province as an expert”); Commonwealth v. Gardner, 350 Mass. 664, 665-667 (1966) (error to elicit from medical witness that there had been “forcible entry” of alleged victim of rape).
With respect to the admission of the expert medical testimony by Dr. Paradise, the hypothetical questions put to her did not suffer from the same defect as the hypothetical questions put to Dr. Brant. In each case Dr. Paradise was asked to “assume” that a child was vaginally penetrated on a regular basis, and was then asked to opine on whether this was “inconsistent” with the absence of physical trauma to the genital areas of the child. The jury could not have understood her testimony to mean that the abuse did occur, but only that it was possible that the abuse had occurred.
4. Harmless error analysis. We turn to consider whether the admission of the testimony of Dr. Brant was harmless error. Defense counsel objected to the admission of the testimony of Dr. Brant.14 An error is nonprejudicial only when we are sure that the error “did not influence the jury, or had but very slight effect .... But if one cannot say, with fair assurance, after pondering all that happened without stripping the erroneous action from the whole, that the judgment was not substantially swayed by the error, it is impossible to conclude that substantial rights were not affected.” Commonwealth v. Peruzzi, 15 Mass. *853App. Ct. 437, 445 (1983), quoting Kotteakos v. United States, 328 U.S. 750, 764-765 (1946). See Commonwealth v. Flebotte, 417 Mass. 348, 353 (1994).
Addressing Dr. Brant’s testimony, the hypothetical questions were deeply flawed and the opinion testimony of Dr. Brant could have been understood by the jury as testimony that sexual abuse did in fact occur in this case. Where, as here, the complainants’ allegations were the only evidence against the defendant, we cannot conclude that Dr. Brant’s expert testimony did not play a significant role in the jury’s determination and that the error was harmless. The admission of the expert testimony of Dr. Brant that was elicited through hypothetical questions was reversible error. Accordingly, the judgments are reversed, the verdicts set aside, and the case is remanded for a new trial.
So ordered.
Justice O’Connor participated in the deliberation on this case, but retired before the opinion was issued.
Appendix.
A. An example of one of the hypothetical questions posed by the prosecution to Dr. Brant follows:
Q.:“I would ask you to assume for purposes of this hypothetical, Doctor, that two sisters living with their mother and stepfather had sexual relations with the stepfather on a regular basis since the ages of twelve and nine. Further assume that neither sister disclosed those relations for six years while it was ongoing. I would ask you further to assume that ■ at age fifteen the younger sister disclosed the sexual activity to her mother the following day on which the stepfather had intercourse with her in the afternoon and in the evening scolded her, yelled at her, and lectured her for failing to fold towels and finish dishes in the sink.
“I would ask you further to assume that the morning after that incident, the fifteen-year-old girl became upset and told her mother the Defendant had been having sexual relations with her for years. Do you have an opinion, to a reasonable degree of medical certainty, as to whether or not those assumed facts are consistent with the girl who had been having sexual relations with an adult since the age of approximately nine-and-a-half?”
Based on that hypothetical question, Dr. Brant opined as follows:
*854“Well, I have described the Circumstances in which the sexual relations can be maintained as a secret for a very long time. Then, comes the question about what the circumstances are under which a child who has had ongoing sexual relations will tell someone about it.
“Those circumstances can include, first of all, a child growing up and growing into adolescence and having a changed awareness of what their situation is, that more and more they might have wishes to have relationships outside the family. To the extent that they are limited in doing that and having sexual relationships within the family, that can become more and more bothersome to a child and create more conflict.
“It is often in the context of a family argument and anger that a child will, for the first time, tell someone about what has happened. Also, I think that what you described in terms of the child having sexual relationships at one point in time and then being reprimanded for not folding clothes in the proper sort of way, I think there can be a lot of strong feelings, a lot of conflict about, on the one hand, being treated in a more adult fashion; on the other hand, being reprimanded as a child or a servant, that can give rise to anger. The anger in some way can counteract the fears that have led a child to keep the sexual relations secret.
“I think that by virtue of the argument, the anger, a child going into adolescence, that those are the kinds of circumstances in which it is very typical that a child might for the first time disclose a secret that they have been keeping for many years.”
B. An example of a hypothetical question posed by the prosecution to Dr. Paradise follows:
Q.: “And, Doctor, I would ask you to assume for purposes of this question that a fifteen-year-old girl was vaginally penetrated by an adult male penis several times weekly on a regular basis from the ages of eleven to age fifteen. I would ask you to further assume that the most recent penetration was five days prior to a physical examination conducted on this girl which revealed no physical signs of trauma to the genital area of the girl.
“Based on your training and experience, do you have an opinion to a reasonable degree of medical certainty as to whether or not those three assumed facts are inconsistent?”
Dr. Paradise opined as follows:
“Because you described the girl who has been experiencing intercourse for four years, from the ages of eleven to fifteen, and trauma means injury, fresh injury, like bruising or bleeding or a laceration.
“Had there been some injury produced by that intercourse, it might have occurred very early on during this continuing series of events. *855After four years, even if there had been intercourse within a few days before the examination, there would be no reason to presume that there would be any evidence of injury at the time of the examination.”
We note that at the time of trial in 1988, in deciding whether to admit the challenged expert testimony, the trial judge did not have the benefit of appellate cases that have since elaborated on the admissibility of expert testimony in child sexual abuse cases. See, e.g., Commonwealth v. Dockham, 405 Mass. 618 (1989); Commonwealth v. Perkins, 39 Mass. App. Ct. 577 (1995).
The defendant also alleged error in (1) the conduct of the plea hearing; (2) the admission of testimony of witnesses that commented on the veracity of one of the complainants; and (3) the jury instructions on fresh complaint. With respect to these arguments, the Appeals Court concluded that there was no error. See Commonwealth v. Federico, 40 Mass. App. Ct. 616, 619-621 (1996). The Appeals Court correctly decided these issues.
The sexual abuse started with illicit touching, and eventually included oral sex and sexual intercourse. The incidents initially occurred several times a week, with increasing frequency over the years, culminating in nearly daily occurrence of sexual intercourse with one or the other sister.
For examples of the hypothetical questions posed by the prosecution to Dr. Brant and to Dr. Paradise, see Appendix.
We denied the Commonwealth’s application for further appellate review on February 29, 1996. See Commonwealth v. Perkins, 422 Mass. 1104 (1996).
At trial, the behavior of the complainants that could be viewed as inconsistent with their allegations of sexual abuse by their stepfather was the subject of forceful cross-examination by defense counsel.
SexualIy abused children may be hesitant witnesses, especially when confronting their abusers. Their allegations of sexual abuse may be considered unreliable, or there may be concerns that child complainants are susceptible to techniques of questioning or therapy that might cause false recollections or confabulation. See, e.g., Ceci, How Suggestible are Preschool Children? Cognitive and Social Factors, 36 J. Am. Acad. Child Adolescent Psychiatry, 948 (July, 1997); American Psychological Association, The Suggestibility of Children’s Recollections (John Dorris ed. [1991]).
This is true whether the testimony stems from an expert witness or a lay witness. We have been careful to note that “[a]n expert may not render an opinion on the credibility of a witness because the jury is capable of making that assessment without the aid of an expert.” Commonwealth v. Ianello, 401 Mass. 197, 202 (1987).
We prohibit the expert witness from offering comparative testimony, reasoning that, while it falls short of directly rendering an opinion on the credibility of the specific witness, “[i]t would be unrealistic to allow this type of expert testimony and then expect jurors to ignore it when evaluating the credibility of the complaining child,” Commonwealth v. Ianello, supra at 202.
In addition, we have observed that where expert testimony is admitted it is good practice for the trial judge to give carefully limiting instructions to the jury regarding the proper function and use of such evidence. See Commonwealth v. Dockham, supra at 629.
The Commonwealth informed the judge that Dr. Brant would respond to hypothetical questions “regarding common clinical phenomena seen in children who have been sexually abusedf,] such as secrecy, delayed disclosure and retraction . . . [and] common behaviors seen in children who have been sexually abused which may appear strange or bizarre to a lay person and even inconsistent with a claim of sexual abuse.” Expert testimony regarding such common clinical phenomena is, of course, admissible. See Commonwealth v. Hudson, 417 Mass. 536, 540 (1994); Commonwealth v. Dockham, 405 Mass. 618, 629 (1989). However, the actual hypothetical questions posed by the prosecutor to Dr. Brant did not assume “common clinical phenomena”; they assumed specific facts admitted in evidence through the testimony of the two child complainants. See PJ. Liacos, Massachusetts Evidence § 7.10.2, at 418 (6th ed. 1994) (“A question based on assumptions not in evidence is improper”). It is not necessary, of course, for counsel to use hypothetical questions to elicit an expert’s opinion when the opinion is not based on the expert’s personal observations. Proposed Mass. R. Evid. 705 (an expert may give an opinion without prior disclosure of the underlying facts or data unless the court requires otherwise), approved in Department of Youth Servs. v. A Juvenile, 398 Mass. 516 (1986), is intended to eliminate the requirement for posing a hypothetical question to expert witnesses. See Department of Youth Servs., supra at 532; Liacos, supra at § 7.10.3, at 420-421.
Evidence of physical injury of abuse is not always available in child sexual abuse cases, but is admissible when available. See, e.g., Commonwealth v. LeFave, 407 Mass. 927, 930-931 (1990) (physical evidence of sexual abuse showing that all female victims suffered vulvitis, redness around their labia, with one child having cracking and Assuring around her labia with bump on her hymen admitted without discussion on review). See also Commonwealth v. Pikul, 400 Mass. 550, 552 (1987) (extensive physical evidence of sexual abuse of murder victim admitted without challenge); Commonwealth v. Montmeny, 360 Mass. 526, 528-529 (1971) (physical evidence of rape properly admitted).
The jury may be under the mistaken understanding that certain types of sexual abuse always or nearly always cause physical injury or scarring in the victim. Particularly where the defendant argues that the absence of any scarring or injury is evidence that the abuse did not occur, expert medical testimony to the contrary may assist the jury in properly evaluating the evidence. See Myers, Expert Testimony in Child Sexual Abuse Litigation, 68 Neb. L. Rev. 1, 35-36 n.122 (1989).
We agree with the analysis of the Appeals Court, 40 Mass. App. Ct. 616, 618 n.2 (1996), that the defendant adequately preserved the issue for appeal. The Commonwealth correctly notes that, in the years intervening between the defendant’s conviction and his appeal, we have had occasion to clarify the law regarding the use of expert witnesses in child abuse cases. It suggests that the defendant should not have the benefit of what are now more stringent requirements for the use of such expert witnesses. While we recognize that ordering a new trial may come with a heavy cost to the complainants and their family, particularly where, as here, so many years have lapsed since the trial, we adhere to our view that, on direct appeal, a defendant has the benefit of intervening decisional law. See Commonwealth v. D’Agostino, 421 Mass. 281, 284 (1995).