Skip to content
Search resources Open menu

Child-CERT

Unfounded Is Not False: Outcome Statistics in Child Sexual Abuse Allegations

Published

When a child sexual abuse (CSA) case goes to trial, the factfinder will often hear that false allegations are rare. That statement may be accurate as far as it goes, but it answers a narrow and somewhat artificial question. Investigations and the legal system do not ordinarily set out to show that an allegation was false. They ask whether abuse can be established, and when it cannot, the inquiry usually stops at “unfounded” or “unsubstantiated.” A formal finding that an allegation was false is unusual, so the number of allegations labeled false is small almost by design.

The more informative figure is the unfounding rate: the proportion of allegations that an investigation could not substantiate. In the studies most often cited on this subject, the unfounding rate runs from about 44 to 57 percent of reports, while the share judged false or fictitious runs from about 3 to 6 percent. Quoting only the false rate, without the unfounding rate, leaves the impression that nearly every allegation is true. The research does not support that impression.

This paper explains the difference between “unfounded” and “false,” presents the unfounding rates and false rates side by side for the major studies, corrects several figures that are commonly repeated from secondary sources, and discusses how the distinction bears on an expert’s claims of experience and on motion practice. Each figure below was checked against the original article.

The Bottom Line

At the time a child is forensically interviewed, every allegation is unproven. Interviewers seldom learn how their cases end, and when they do, the outcome usually reflects what could be proven rather than what happened.

In the major studies that have counted all outcomes, roughly half of the allegations were unfounded, that is, not substantiated; the Canadian national data exceed half when cases that remained “suspected” are counted as not substantiated. In the five studies summarized below, allegations judged false or fictitious were a small fraction, between about 3 and 6 percent of all reports. The unfounded group outnumbers the false group by roughly eight to eighteen times.

The unfounding rate is not a false rate. An unfounded case may involve abuse that could not be proven, a concern that turned out not to be abuse, or, occasionally, a fabrication. What the unfounding rate measures is how often the truth was never established. That is precisely the information a false rate leaves out.

“False allegations are rare” and “most allegations are true” are therefore different claims. A low rate of proven-false allegations says nothing about how many allegations are proven true.

An expert who relies on experience with a large number of interviews is relying on a caseload in which about half of the allegations, if those cases resemble the studies, were never substantiated.

Three Outcomes, Not Two

The legal system and child protective services (CPS) agencies decide whether an allegation can be proven, not whether it is true. A CPS agency “substantiates” a report when the evidence meets its administrative standard, and labels the rest “unsubstantiated” or “unfounded.” In most systems an unfounded finding means that maltreatment could not be established; it does not mean that the agency found the report to be false. In Arizona, a Department of Child Safety finding is substantiated only when it is supported by a preponderance of the evidence (A.R.S. § 8-811). Law enforcement agencies use the same word differently. Under FBI crime-reporting rules, a police report is “unfounded” only when investigation shows that no offense occurred or was attempted, so a police “unfounded” and a CPS “unfounded” are not equivalent, and figures from the two systems should not be compared without checking which meaning applies.

Criminal verdicts work the same way. A conviction requires proof beyond a reasonable doubt (In re Winship, 1970). An acquittal means the government did not meet that burden. Juries do not find that a defendant is innocent or that an accusation was false, and investigating agencies rarely make a formal finding of falsity. Because the system so seldom determines that an allegation was false, the number of allegations recorded as false is bound to be small, and that number cannot be read as the complement of the true allegations.

Much of the disagreement over false-allegation rates is a disagreement over definitions. Research on false allegations of adult rape has shown that criminal justice professionals and researchers often mean different things by a “false allegation,” so their estimates inevitably diverge (Saunders, 2012). The same problem runs through the CSA literature, where “unfounded,” “unsubstantiated,” “inconclusive,” “erroneous,” “fictitious,” and “false” are used inconsistently.

Why the Unfounding Rate Is the Better Question

There are two ways to ask how often allegations fail. The first is to ask what proportion of allegations were unfounded: investigated and not substantiated. Every investigating agency produces that figure as a matter of course, because every case it closes is classified one way or the other. The second is to ask what proportion of allegations were false. That figure exists only where someone went further than the investigation required and made an affirmative determination of falsity, such as a recantation the investigator credited, a confession by a coaching adult, or a fabrication that was documented in the file.

Because investigations seldom take that further step, a reported false rate counts only the cases in which falsity happened to be documented. It does not count unfounded cases that were in fact false but were closed without a finding, and it cannot, because no one examined them for that purpose. A false rate is therefore a floor, not an estimate of how many allegations were untrue.

The unfounding rate has its own limitation: it includes true allegations that could not be proven, and it cannot be read as the proportion of allegations that were untrue. Its value is that it is complete. It accounts for every investigated case and shows how large the zone of uncertainty actually is. When a witness quotes a false rate alone, the appropriate response is to ask for the unfounding rate in the same study.

What the Studies Show

The table summarizes the studies most often cited on this question. Percentages are of all cases in each sample. The unfounding rate is every case that was not substantiated. It is broken down into cases left unresolved and cases judged not to involve abuse, and the share judged false or fictitious is shown separately.

[[TABLE]]

Five points stand out. First, the unfounding rate is consistently near half: 47, 44, 50, 57, and 55.5 percent. Second, every study has a large unresolved group within that unfounded half. Third, the share of reports judged false or fictitious is small, between about 3 and 6 percent of all reports in these samples. Fourth, the categories are not interchangeable: Jones and McGraw’s agency labeled 47 percent of reports “unfounded,” but only 6 percent were judged fictitious; Everson and Boat’s caseworkers substantiated 56 percent of cases but judged only about 5 percent false. Jones and McGraw’s “fictitious” category was also defined to include misperceptions and confused interpretations of innocent events as well as deliberate falsifications, and Everson and Boat’s caseworkers viewed only 59 percent of the false reports as deliberate fabrications by the child. Fifth, the false group is small precisely because so few cases are ever resolved as false; most cases that are not substantiated simply remain unresolved.

Other data show the same pattern. In a Welsh study of all 410 children referred for suspected sexual abuse over four years, in an area where reporting was not mandatory, 350 referrals were fully investigated; abuse was substantiated for 197 children (48 percent of all referrals and 56 percent of those investigated), and 153 (44 percent of those investigated) were unsubstantiated. Investigators judged 63 of the investigated cases (18 percent) to be false allegations, 30 of them (8.5 percent) both false and malicious, a higher share than in the U.S. studies above, under criteria the authors set out in an appendix (Anthony & Watkeys, 1991). In the Canadian national incidence study of 7,672 maltreatment investigations, 42 percent of investigations were substantiated, a further 23 percent remained suspected, and 35 percent were unsubstantiated; counting suspected cases as not substantiated, as most U.S. studies do, 58 percent were not substantiated. Only 4 percent of all investigations were considered intentionally fabricated, rising to 12 percent where there was a custody or access dispute (Trocmé & Bala, 2005). These figures cover all forms of maltreatment, not only sexual abuse. Nationally, CPS agencies in the United States identified 532,228 child victims of all forms of maltreatment in fiscal year 2024 among 2,990,234 children who received an investigation or alternative response, or fewer than one child in five, about 18 percent (U.S. Department of Health and Human Services, 2026). The rest include children who received an alternative, non-investigative response, for which no substantiation finding is made, as well as children whose reports were investigated and not substantiated. That figure covers all types of maltreatment, not only sexual abuse, and it measures substantiation, not truth.

A review of the early literature found false-allegation rates of 2 to 8 percent among referrals to child abuse clinics and 6 percent among emergency room referrals, with higher rates of 36 to 55 percent in two small samples of allegations arising in custody disputes (Mikkelsen et al., 1992). Those two custody samples contained 11 and 18 cases. The larger custody study by Thoennes and Tjaden (1990), which drew on court records covering more than 9,000 families in custody or visitation disputes and a detailed sample of 169 cases, 129 of them with a determination about whether abuse had occurred, found that allegations in that setting were no more likely to be ruled unfounded than CSA reports generally.

A Closer Look at the Carnes Study

The multisite study by Carnes and colleagues (2001) is often cited, and often misread. It evaluated an extended, multi-session forensic evaluation model developed by the National Children’s Advocacy Center. Professionals at 18 agencies in 12 states enrolled 147 children whose first investigative interview had not produced a clear statement; 10 cases were excluded for inadequate documentation, leaving 137.

Evaluators judged 61 children (44.5 percent) to have made credible disclosures. They judged abuse unlikely in 26 cases (19 percent), including 4 cases (3 percent) in which an adult appeared to have coached the child to make a false statement, and the outcome remained unclear in 50 cases (36 percent). Follow-up information was available for 49 of the 61 credible-disclosure cases. Of those 49, 43 (88 percent) had at least one corroborating factor, and 73 percent were “supported in the legal arena.” That legal support combined several kinds of outcome measured at the end of the study period: confessions in 26 percent of the cases, criminal convictions in 10 percent, protective rulings by family or civil courts in 35 percent, indictments with trials pending in 20 percent, and diagnostic medical evidence in 12 percent.

Read correctly, the study shows three things. A criminal conviction had been documented, at the time of follow-up, for roughly five of the 137 children. More than half of the children (55 percent) completed the evaluation without a credible disclosure, which leaves their cases unresolved or judged unlikely, not proven false. And the study’s own authors acknowledged that ground truth could not be known and that the outcome classifications rested on the evaluators’ judgment using a credibility guide with no known psychometric properties (Carnes et al., 2001). The study’s unfounding rate, in the sense used here, is 55.5 percent. Two common restatements of this study are wrong: the 73 percent figure is not the proportion “proven in a court of law,” and the 55 percent without a credible disclosure is not the proportion found “unfounded or false.” Because the sample was limited to children whose first interview was unclear, none of these rates can be generalized to forensic interviews as a whole.

The Expert’s Experience Base

Experts who testify about how abused children behave, disclose, delay, or recant often rest those opinions in part on their own experience, sometimes many hundreds of forensic interviews. That experience is real, but its evidentiary value depends on knowing which of those children were abused. The studies above show that a substantial portion of any interviewer’s caseload will consist of children whose allegations were never resolved, and a further portion will consist of children who were judged not to have been abused. Interviewers rarely receive systematic feedback on case outcomes, and the outcomes they do learn reflect provability. The same limitation applies to any expert, retained by either side, whose opinion rests on uncontrolled clinical experience: an expert who has evaluated many cases the expert regards as false allegations faces the same question of how those cases were confirmed.

The research literature recognized this problem decades ago. Horowitz and colleagues (1995) explained that no single reliable and valid method exists for establishing what actually happened to an alleged victim, that field studies must state explicit and convergent criteria for assigning children to abused, non-abused, and false-allegation groups, and that researchers who fail to do so “risk confusion rather than insight” (p. 49). They also warned that studies limited to the small proportion of cases that can be absolutely confirmed risk being unrepresentative. An individual expert’s caseload has the same problem as a poorly specified study: unless the expert can say which cases were confirmed, the expert cannot say whether the patterns observed came from abused children, non-abused children, or both.

The issue can be framed accurately with questions such as these:

  1. When you interview a child, has the allegation been proven at that point?
  1. Of the children you have interviewed, in how many do you know the final outcome of the case?
  1. Of those, how many were proven by conviction or confession? How many were unfounded, that is, never substantiated? How many were determined to be false?
  1. Did anyone investigate the unfounded cases to determine whether they were false?
  1. So for a substantial number of the children you have interviewed, you do not know whether abuse occurred?
  1. When you describe how abused children behave based on your experience, which children are you counting?

The fair point is not that none of the expert’s cases were true. It is that the expert cannot separate the true cases from the others, so experience alone cannot establish what abused children, as opposed to children who made allegations, typically do.

The same questions apply, with equal force, to an expert called by the defense. An expert who testifies that allegations arising in custody disputes, or allegations with particular features, are often false, and who rests that opinion on cases from his or her own practice, should be able to say how many of those cases were determined to be false by a court, an investigation, or a recantation that was itself tested, rather than unfounded or simply believed by the expert to be false; whether anyone outside the evaluation confirmed that conclusion; and how the cases were selected. An expert whose experience consists mainly of cases referred by one side sees a sample shaped by that referral source. Uncontrolled clinical experience does not establish which allegations were false any more than it establishes which were true, and a factfinder should weigh such opinions by the same standard whichever party offers them.

“Erroneous” and “Malicious” Allegations

A common formulation in expert testimony divides false allegations into two types. Erroneous allegations are made in good faith and, on investigation, are found to have another explanation or to lack sufficient evidence that abuse occurred. Malicious allegations are knowingly made to achieve some goal and are described as rare. The studies usually cited for this view include Jones and McGraw (1987), Thoennes and Tjaden (1990), Oates and colleagues (2000), and Trocmé and Bala (2005). Usage varies across these studies. Oates and colleagues kept inconclusive cases (21 percent) separate from erroneous concerns; their category of erroneous accounts by children (2.5 percent) comprised three allegations made in collusion with or after coaching by a parent, three children who misinterpreted an innocent event, and eight deliberate fabrications (1.5 percent), while good-faith over-reactions by parents, community members, and professionals (62 cases, about 11 percent) and nine fabrications by adults were counted within “not sexual abuse” (Oates et al., 2000, pp. 151–152).

The distinction is legitimate and is supported by the research. Its significance for cross-examination lies in what the “erroneous” category contains. By its own terms it includes allegations for which there was insufficient evidence that abuse occurred, which is the large unresolved group in every study above. An expert who says “false allegations are rare” while meaning only malicious allegations has defined away most of the cases that did not result in proven abuse. The appropriate follow-up is to ask the expert how common erroneous allegations are and what the unfounding rate was in each study the expert relies on. In each of the studies listed, the answer is far larger than the malicious figure.

Using These Statistics in Court

Statistical evidence about allegation outcomes is most useful in pretrial litigation. In a motion under Federal, Military, or Arizona Rule of Evidence 702, or a motion in limine directed to an expert’s testimony about the rarity of false allegations or about patterns drawn from personal experience, the figures in this paper can show that the expert’s statements omit the largest category of outcomes and that the expert’s experience cannot be validated against ground truth.

At trial, courts are wary of numbers. The Court of Appeals for the Armed Forces has held that testimony quantifying the likelihood that a complainant is truthful is the functional equivalent of vouching (United States v. Brooks, 2007; United States v. Mullins, 2010), and Arizona bars expert testimony quantifying the proportion of complainants who tell the truth (State v. Lindsey, 1986, 149 Ariz. at 475–76). That principle applies in both directions. It limits an expert’s testimony that false allegations are rare, and it equally limits any suggestion that a non-substantiation or acquittal rate shows that a particular allegation is false. Where percentages are excluded, the same point can often be made in general terms, such as “about half of investigated cases are unfounded” or “investigations ordinarily stop at unfounded; they do not go on to determine whether an allegation was false.”

The distinction should be stated with care. Unproven means unknown. It is not evidence that an allegation is false, and an argument that treats every unproven case as false makes the same error as an argument that treats every case not proven false as true. Overstating the point, for example by describing unresolved cases as false, invites a correction that can cost counsel credibility with the court.

Limits of the Evidence

Most of these studies are old. The Denver studies, the North Carolina caseworker survey, and the custody study were conducted between 1983 and the late 1980s, and the second Denver study in 1992, before recorded, protocol-based forensic interviews and Children’s Advocacy Centers became standard. The samples differ: CPS reports, clinic referrals, custody cases, and children selected because their first interview was unclear are not the same populations, and the national CPS data cover all forms of maltreatment. Classifications rest on caseworker or clinician judgment, which is itself imperfect. Finally, no study can establish whether a particular allegation is true, unresolved, or false; rates describe groups of cases defined in particular ways.

Getting the Numbers Right

Several figures in circulation come from secondary summaries rather than the original studies. The review by Mikkelsen and colleagues (1992) is a useful source but misreports two primary studies. It describes Jones and McGraw’s Denver data as 53 percent confirmed, 17 percent “suspicious but not conclusive,” 24 percent insufficient evidence, and 2 percent “definite false allegations.” The original table shows 49 percent reliable and 4 percent recanted (together the agency’s 53 percent founded), 24 percent insufficient information, 17 percent unsubstantiated suspicion, and about 6 percent fictitious, of which 5 percent were made by adults and 1 percent by children (Jones & McGraw, 1987). Mikkelsen and colleagues also give Everson and Boat’s false-allegation rates by age as 6, 7, 4.3, and 8 percent; the original table gives 1.6 percent for children under 3, 1.7 percent for ages 3 to 5, 4.3 percent for ages 6 to 11, and 8.0 percent for adolescents (Everson & Boat, 1989). For the Carnes study, the correct figures are 12 states and 18 agencies, 44.5 percent credible disclosures among 137 valid cases, and criminal convictions in 10 percent of the credible-disclosure cases with follow-up, as described above. Jones and McGraw themselves cautioned that their figures should be used as a basis for further study, not as a basis for proving that a particular case is or is not true, and noted that their study had already been misused in court for that purpose (Jones & McGraw, 1987). Citing the primary sources accurately avoids giving an opposing expert an easy correction.

References

Anthony, G., & Watkeys, J. (1991). False allegations in child sexual abuse: The pattern of referral in an area where reporting is not mandatory. Children & Society, 5(2), 111–122.

Carnes, C. N., Nelson-Gardell, D., Wilson, C., & Orgassa, U. C. (2001). Extended forensic evaluation when sexual abuse is suspected: A multisite field study. Child Maltreatment, 6(3), 230–242. https://doi.org/10.1177/1077559501006003004

Everson, M. D., & Boat, B. W. (1989). False allegations of sexual abuse by children and adolescents. Journal of the American Academy of Child & Adolescent Psychiatry, 28(2), 230–235. https://doi.org/10.1097/00004583-198903000-00014

Horowitz, S. W., Lamb, M. E., Esplin, P. W., Boychuk, T. D., Reiter-Lavery, L., & Krispin, O. (1995). Establishing ground truth in studies of child sexual abuse. Expert Evidence, 4(2), 42–51.

Jones, D. P. H., & McGraw, J. M. (1987). Reliable and fictitious accounts of sexual abuse to children. Journal of Interpersonal Violence, 2(1), 27–45. https://doi.org/10.1177/088626087002001002

Mikkelsen, E. J., Gutheil, T. G., & Emens, M. (1992). False sexual-abuse allegations by children and adolescents: Contextual factors and clinical subtypes. American Journal of Psychotherapy, 46(4), 556–570. https://doi.org/10.1176/appi.psychotherapy.1992.46.4.556

Oates, R. K., Jones, D. P. H., Denson, D., Sirotnak, A., Gary, N., & Krugman, R. D. (2000). Erroneous concerns about child sexual abuse. Child Abuse & Neglect, 24(1), 149–157. https://doi.org/10.1016/S0145-2134(99)00108-8

Saunders, C. L. (2012). The truth, the half-truth, and nothing like the truth: Reconceptualizing false allegations of rape. British Journal of Criminology, 52(6), 1152–1171. https://doi.org/10.1093/bjc/azs036

Thoennes, N., & Tjaden, P. G. (1990). The extent, nature, and validity of sexual abuse allegations in custody/visitation disputes. Child Abuse & Neglect, 14(2), 151–163. https://doi.org/10.1016/0145-2134(90)90026-P

Trocmé, N., & Bala, N. (2005). False allegations of abuse and neglect when parents separate. Child Abuse & Neglect, 29(12), 1333–1345. https://doi.org/10.1016/j.chiabu.2004.06.016

U.S. Department of Health and Human Services, Administration for Children and Families, Administration on Children, Youth and Families, Children’s Bureau. (2026). Child maltreatment 2024. https://acf.gov/cb/data-research/child-maltreatment

Legal Authorities

In re Winship, 397 U.S. 358 (1970).

State v. Lindsey, 149 Ariz. 472, 720 P.2d 73 (1986).

United States v. Brooks, 64 M.J. 325 (C.A.A.F. 2007).

United States v. Mullins, 69 M.J. 113 (C.A.A.F. 2010).

Ariz. Rev. Stat. Ann. § 8-811.

Ariz. R. Evid. 702; Fed. R. Evid. 702; Mil. R. Evid. 702.