Skip to content
Search resources Open menu

False Confessions

Corroboration and Assessing the Reliability of a Confession

Published

Snap Shot

  • The question: What independent evidence and features of a confession bear on whether it is reliable, and what corroboration does the law require before a confession can support a conviction?
  • What the research shows: In a study of 38 DNA exoneration cases with false confessions and available records, 36 confessions reportedly contained key details about how the crime occurred, and in 27 cases officers testified that they had not disclosed facts to the suspect (Garrett, 2010); Garrett concluded the details most likely came from the investigation, and the finding describes this selected group, not confessions generally. Details are most informative when they were not disclosed, are later verified, or lead to evidence investigators did not already have; a confession can also influence how other evidence is perceived.
  • What it does not show: The research does not provide a validated test that sorts true from false confessions, and the exoneration studies examine known errors, not the large body of confessions presumed to be true. Most of the experimental work uses students or online samples in low-stakes settings.
  • The law: Federal courts require substantial independent evidence tending to establish a confession’s trustworthiness (Opper); Mil. R. Evid. 304(c), since 2016, requires independent evidence that would tend to establish trustworthiness; Arizona applies a corpus delicti rule requiring a reasonable inference that the crime occurred (Gillies, Carlson). Each standard is modest, and the remaining weight of the confession is for the factfinder.
  • Issues for each side: For prosecutors, the research favors documenting what was disclosed to the suspect and by verifying non-public details and new leads independently; for defense counsel, tracing where each detail in the confession could have come from and by testing whether the “corroboration” is truly independent of the confession.

A confession is persuasive evidence because it seems to answer the central question in a case directly. The United States Supreme Court has described a confession as probably the most probative and damaging evidence that can be admitted against a defendant (Arizona v. Fulminante, 1991, quoting Bruton v. United States, 1968, White, J., dissenting). That persuasive force is also why its reliability matters. Most confessions are presumed to be true, although the proportion that are false is unknown and no study measures it directly, and many are corroborated by physical, documentary, or witness evidence. A small number of documented cases, however, show that innocent people have given detailed confessions that investigators, prosecutors, judges, and jurors found convincing.

Reliability questions reach court in two ways. First, every jurisdiction requires some corroboration before a confession can support a conviction, and the defense may challenge whether that threshold has been met. Second, once a confession is admitted, the jury decides how much weight it deserves, and both sides offer evidence about how it was obtained, what it contains, and how well it fits the other evidence (Crane v. Kentucky, 1986).

This paper describes the research on what corroboration can and cannot show: the role of non-public details, the post-admission narrative, new information that leads to evidence, and the risk that a confession shapes other evidence. It then summarizes the federal, military, and Arizona corroboration rules. The related questions of how crime details enter a confession and how investigators’ expectations shape an investigation are addressed in separate papers in this hub, “Contamination: Non-Public Details and Post-Admission Narratives” and “Investigator Bias, Tunnel Vision, and Forensic Confirmation Bias.”

What Corroboration Means in the Research

In the research literature, corroboration refers to the relation between a confession and evidence that exists independently of it. Researchers have distinguished three questions (Leo & Ofshe, 1998; Leo, 2008). The first is whether the suspect admitted guilt. The second is what the suspect said about how the crime was committed, sometimes called the post-admission narrative. The third is whether that narrative fits the facts that can be established without the confession. A bare admission (“I did it”) says little that can be checked. A narrative that describes where, when, how, and with what, and that can be compared with crime-scene, medical, forensic, and witness evidence, can be tested.

Leo and Ofshe (1998) identified at least three indicia of reliability: whether the statement led to the discovery of evidence unknown to the police, whether it identified highly unusual elements of the crime that had not been made public, and whether it accurately described mundane details of the crime scene that are not easily guessed and had not been reported publicly. On this view, a narrative that contains errors about facts a perpetrator would know, or that matches only facts already known to investigators, supplies weaker corroboration. These criteria are a framework for organizing the evidence rather than a validated instrument; no study has measured how accurately they classify confessions when applied by different evaluators.

The same reasoning underlies long-standing investigative practice. The Reid manual advises investigators to withhold non-public crime details from suspects so that a confessor’s knowledge of them can be tested, a practice the 2025 scientific review credits (Inbau et al., 2013, as cited in Kassin et al., 2025). Investigative-interviewing approaches likewise emphasize obtaining a full account and checking it against the evidence. When this practice is followed and documented, a confession that contains accurate withheld details, or that leads investigators to evidence they did not know about, carries substantial weight.

Non-Public Details in Known False Confessions

The research that has most influenced current thinking comes from DNA exoneration cases. Garrett (2010) examined the first 250 people exonerated by postconviction DNA testing in the United States; 40 had falsely confessed, and records were obtained for 38. In 36 of the 38, the confession reportedly contained key details about how the crime occurred. In 27 of the 38, officers testified at trial that they had not disclosed facts to the suspect, and in some trials prosecutors emphasized in closing argument that the details were non-public or matched the crime-scene evidence. Because DNA later excluded these confessors, Garrett concluded that the details most likely came from the investigation itself. A follow-up analysis of 66 cases found accurate non-public facts in 62, or 94% (Garrett, 2015, as reported in Kassin et al., 2025).

These cases show that the presence of accurate details, standing alone, does not establish that the details originated with the suspect. They do not show how details entered any particular confession. Kassin and colleagues (2025) describe two routes: deliberate disclosure in some documented cases of misconduct, and more often inadvertent disclosure, as when investigators ask leading questions, show photographs, visit the scene with the suspect, or mention facts they assume the suspect already knows. Innocent suspects may also learn details from news reports, community talk, or witnesses. The exoneration data come from a selected set of cases, mostly homicides and rapes, in which DNA was available and later tested; they describe how errors occurred in those cases, not how often detailed confessions are false.

Experimental work supports the inadvertent route. In a mock-crime study with student participants, people playing investigators passed crime details to both innocent and guilty mock suspects without any incentive to do so, and false confessions later contained accurate details; in a second experiment with online participants, listeners who heard only the confession showed a bias toward guilt, but hearing the full interview reduced that bias for innocent confessors (Alceste et al., 2020). The setting was low-stakes and the investigators were not trained police officers, which limits direct generalization, but the finding illustrates why a complete recording is the most reliable way to establish what was disclosed.

Known false confessions also contain features that ordinarily signal credibility. In an analysis of 20 false confessions, all contained visual and auditory details of the crime and scene, most referred to other people, many quoted the victim, and some expressed remorse; in an accompanying mock-jury study, a detailed confession produced more convictions than a bare admission (Appleby et al., 2013). Because false confessions can contain such features, vividness and emotion should not by themselves be treated as signs of reliability; the study did not compare them with true confessions.

New Information, Errors, and Fit

The prosecution’s most persuasive corroboration is often information that the investigators did not have before the interview and that is later verified: the location of a weapon, body, or stolen property; the identity of a witness; a fact later confirmed by forensic testing; or a detail that explains something investigators had not understood. Such information cannot have been supplied by the investigators, and its later verification is independent of the confession itself (Leo & Ofshe, 1998). The value of this kind of corroboration depends on documentation showing that the information was in fact new at the time and that the later discovery followed from the statement.

Errors cut in both directions. A guilty suspect may minimize his role, misremember peripheral details, or lie about some facts while admitting others, so inconsistencies do not by themselves show a confession is false. At the same time, a confession that is wrong about central facts a perpetrator would know, such as the method of a killing or the location of an injury, raises a legitimate question about its source. The research does not establish a threshold at which errors outweigh accurate details; that judgment depends on which facts are wrong, how central they are, and whether the accurate facts were available to the suspect from another source.

Factfinders appear to make some use of fit. In a series of mock-juror studies, jurors were marginally more likely to convict when a confession was consistent with the case facts, did not respond to consistency in a study with jurors reporting for service, and did use consistency when the content of the confession itself varied, with expert testimony increasing that sensitivity on the verdict measure (Henderson & Levett, 2016). In an online study of 181 jury-eligible adults, interrogation questions showing that the detective had disclosed confidential details, or an expert video explaining contamination, lowered ratings of the likelihood of guilt, and the questions lowered ratings of the confession’s accuracy; the two together had no added effect (Vornhagen & Hart, 2026). Juror research is summarized more fully in “How Jurors Respond to Confession Evidence.”

When the Corroboration Is Not Independent

Corroboration is useful only if it is independent of the confession. Laboratory and archival research suggests that knowledge of a confession can influence witnesses and lay evaluators, and possibly examiners; whether it did so in a particular case depends on what the witness or examiner knew and when. In a laboratory study, after witnessing a staged theft and viewing a lineup that did not contain the thief, 61% of participants who had chosen someone changed their identification when told another lineup member had confessed, and half of those who had made no choice then selected the confessor (Hasel & Kassin, 2009). Lay participants who were told that a defendant had confessed were more likely to judge handwriting samples to be from the same writer (Kukucka & Kassin, 2014). Kassin, Dror, and Kukucka (2013) described this pattern as a forensic confirmation bias.

Archival data show a related pattern. In DNA exoneration cases, false-confession cases were more likely than eyewitness cases to contain additional errors, most often invalid or improper forensic science, mistaken identifications, and informant testimony, and where several errors were present the confession was usually obtained first (Kassin et al., 2012). The authors argued that these findings bear on the law’s assumption that pieces of evidence are independent. The archival data do not show that the confession caused the later errors in a given case, and the laboratory studies used participants who were not trained forensic examiners. Still, they support the practice, recommended in the 2025 scientific review, of shielding forensic examiners and lay witnesses from knowledge of a confession where possible (Kassin et al., 2025). That practice protects the value of corroboration for the prosecution as much as it protects the accused.

Limits of the Evidence

The strongest field evidence on non-public details comes from DNA exonerations, a small, selected group of cases in which the confession is known to be false. Those cases show that detailed, apparently corroborated false confessions occur and how they have been explained at trial. They cannot show how often detailed confessions are false, because no study has drawn a representative sample of confessions and determined which were true. The share of exonerations involving false confessions is a share of known wrongful convictions, not a rate of false confessions among confessions generally, and no general false-confession rate is known.

The experimental studies use college students or online participants, low-stakes mock crimes, and vignettes or audio clips rather than full trials. Their strength is that the truth is known; their weakness is distance from real investigations, real stakes, and real deliberation. The Leo and Ofshe (1998) indicators are a reasoned framework, not a validated classification tool. Few of the key studies analyzed differences by sex or gender: the exoneration analyses did not test them, and Vornhagen and Hart (2026), whose sample was about evenly divided between men and women, did not report sex differences. Group research identifies features that make a confession more or less informative; it cannot establish that a particular confession is true or false.

Questions to Ask About This Research

When the DNA exoneration studies (Garrett, 2010, 2015) are cited to argue that the details in a confession are meaningless, ask whether there is evidence in this case of how those details could have reached the suspect, what the recording shows was disclosed, and whether any detail led to evidence investigators did not already have; the studies show that details can be supplied, not that they were supplied here. When accurate details are cited as proof that the suspect is the perpetrator, ask which details were public, which were in documents or photographs the suspect saw, which were mentioned in the interview, and whether the interview was recorded from the start. When the confession-influence studies (Hasel & Kassin, 2009; Kukucka & Kassin, 2014; Kassin et al., 2012) are cited, ask whether the examiner or witness in this case actually knew of the confession before reaching a conclusion. When juror studies are cited, ask who the participants were and whether the materials resembled a trial. These questions are explained in “Reading Research Claims: A Guide for Attorneys” on this site.

Corroboration and the Law

Federal and Military Standards

Federal law has long required that a confession be corroborated before it can support a conviction. In Opper v. United States (1954), the Supreme Court held that the government must introduce substantial independent evidence that would tend to establish the trustworthiness of the statement; the independent evidence need not prove the offense on its own, and it is sufficient if it supports the essential facts admitted enough to justify a jury inference of their truth. Smith v. United States (1954) applied the same rule and explained its purpose as guarding against convictions based on untrue confessions alone. Once a confession is admitted, the jury may hear evidence of the circumstances in which it was obtained, because those circumstances bear on its credibility as well as its voluntariness (Crane v. Kentucky, 1986; see 18 U.S.C. § 3501(a)).

In courts-martial, Mil. R. Evid. 304(c) provides that an admission or confession may be considered on the question of guilt only if independent evidence, direct or circumstantial, has been admitted that would tend to establish its trustworthiness. Other uncorroborated admissions cannot supply that evidence; not every fact in the statement must be independently proven; the independent evidence need raise only an inference of the statement’s truth; its amount and type go to the weight the factfinder gives the statement; and the military judge alone decides when corroboration is adequate (Mil. R. Evid. 304(c)(1)–(5), 2024 ed.). Statements made before or at the time of the act, and statements offered under other evidence rules, do not require corroboration (Mil. R. Evid. 304(c)(3)). The rule was revised in 2016 largely in response to United States v. Adams (2015), which had required corroboration of the essential facts admitted (Criminal Law Department, 2025). Under the revised rule, the Court of Appeals for the Armed Forces described the required quantum as slight and the weight of the corroboration as a matter for the factfinder (United States v. Whiteeyes, 2022). Under Rule 304(g), when a statement is admitted the defense may present evidence on its voluntariness, and members are instructed to give it the weight it deserves.

Arizona

Arizona applies a corpus delicti rule rather than the federal trustworthiness doctrine. Before a confession may be considered, the State must present evidence supporting a reasonable inference that the charged crime was actually committed; circumstantial evidence suffices, and the proof may come at any time before the State rests (State v. Gillies, 1983; State v. Carlson, 2015). In Carlson, the Arizona Supreme Court noted that it had never adopted the trustworthiness doctrine and agreed with State v. Morgan (2002) that independent evidence establishing one crime may help corroborate closely related crimes in the same confession. Arizona’s threshold therefore concerns whether a crime occurred, not whether the defendant committed it, and once it is met the reliability of the confession is for the jury. In Carlson, the Arizona Supreme Court upheld the exclusion of an expert’s testimony relaying the defendant’s own claim that his confession to a television reporter was false, and of risk-factor testimony by an expert who had not examined the defendant and whose testimony went to a general propensity to lie rather than to mental or physical circumstances affecting the statement; the court did not hold that expert testimony on confession research is inadmissible. Under A.R.S. § 13-3988, the jury hears evidence on voluntariness and gives the confession such weight as it deserves. Arizona confession law is addressed further in “Arizona Confession Law.”

Special Issues in Legal Practice

The research described above is the same whichever party relies on it. The science operates the same way regardless of who is asking the questions, and the research summarized here does not by itself favor either side. What differs between the parties is not the science but the vantage point — the particular places where these issues tend to surface, and the particular errors each party is best positioned to guard against. The two summaries below identify those recurring issues. They are offered for orientation, not as litigation strategy; in every case the controlling question is what the evidence in that case will, and will not, support, and the expert’s role is to inform the trier of fact rather than to advocate for either side.

Special Issues for Prosecutors

The research supports the prosecution’s use of corroboration that is well documented and genuinely independent. A confession that led to evidence investigators did not have, or that contains accurate withheld details the recording shows were never disclosed, is the kind of confession the research treats as most informative (Leo & Ofshe, 1998), and the Reid manual’s own advice to withhold non-public facts serves the same end (Inbau et al., 2013, as cited in Kassin et al., 2025). A complete recording, a written list of holdback details prepared before the interview, records of what photographs or reports the suspect saw, and forensic analysis completed or confirmed without knowledge of the confession all strengthen the evidence and answer the questions the exoneration research raises. The legal thresholds in Opper, Rule 304(c), and Carlson are modest; confirmed, volunteered details are the kind of independent evidence the research treats as most informative about reliability, and no research permits an expert to say that a particular admission is true or false.

The research also counsels caution. Accurate details are persuasive only to the extent their source can be shown, and the exoneration cases include trials in which investigators testified that they had not disclosed details that, given the later DNA results, most likely came from the investigation (Garrett, 2010). A detailed or emotional narrative is not itself corroboration (Appleby et al., 2013), and forensic or eyewitness evidence obtained after examiners or witnesses learned of a confession may be challenged as not independent (Hasel & Kassin, 2009; Kassin et al., 2012). Meeting the legal threshold does not settle the weight the factfinder will give a confession whose details could have come from the interview.

Special Issues for Defense

The research supports a careful, detail-by-detail analysis of where the facts in a confession could have come from. Defense counsel are often best positioned to compare the confession with everything the suspect could have learned before and during the interview: news coverage, conversations, documents, photographs, a visit to the scene, and the interview itself. The DNA exoneration studies (Garrett, 2010, 2015) and the contamination experiment (Alceste et al., 2020) support asking whether details were disclosed, especially where the interview was only partly recorded. Errors about central facts, the absence of any new verified information, and corroborating evidence developed after a confession was known are subjects the research makes relevant to weight, just as documented holdback details and independent verification are for the prosecution, and they remain relevant after the corroboration threshold is met (Crane; Mil. R. Evid. 304(c)(4), 304(g); A.R.S. § 13-3988).

The research also counsels caution. Most confessions are presumed to be true, although the proportion that are false is unknown; independently verified evidence that a confession led to, or withheld details that a complete recording shows were not disclosed, are the kind of independent evidence the research treats as most informative about reliability, whether the record shows that for a given detail is for the factfinder, and no research permits an expert to say that a particular admission is true or false. Minor inconsistencies are common in true confessions and do not by themselves indicate falsity. The exoneration studies describe a selected group of known errors and do not permit an inference that a particular detailed confession is false, and in Carlson the Arizona Supreme Court upheld the exclusion of an expert’s testimony relaying the defendant’s own claim that his confession to a television reporter was false, and of risk-factor testimony by an expert who had not examined the defendant and whose testimony went to a general propensity to lie rather than to mental or physical circumstances affecting the statement; the court did not hold that expert testimony on confession research is inadmissible. A focused challenge to specific details whose source is uncertain is better supported than a general claim that confessions are unreliable.

This paper is for educational purposes only and is not legal advice. The author is a psychologist, not an attorney; summaries of law are general and must be verified against current authority or with a licensed attorney.

References

Alceste, F., Jones, K. A., & Kassin, S. M. (2020). Facts only the perpetrator could have known? A study of contamination in mock crime interrogations. Law and Human Behavior, 44(2), 128–142. https://doi.org/10.1037/lhb0000367

Appleby, S. C., Hasel, L. E., & Kassin, S. M. (2013). Police-induced confessions: An empirical analysis of their content and impact. Psychology, Crime & Law, 19(2), 111–128. https://doi.org/10.1080/1068316X.2011.613389

Criminal Law Department, The Judge Advocate General’s Legal Center and School, U.S. Army. (2025). Criminal law deskbook (2025 update).

Garrett, B. L. (2010). The substance of false confessions. Stanford Law Review, 62(4), 1051–1118.

Garrett, B. L. (2015). Contaminated confessions revisited. Virginia Law Review, 101(2), 395–454.

Hasel, L. E., & Kassin, S. M. (2009). On the presumption of evidentiary independence: Can confessions corrupt eyewitness identifications? Psychological Science, 20(1), 122–126. https://doi.org/10.1111/j.1467-9280.2008.02262.x

Henderson, K. S., & Levett, L. M. (2016). Can expert testimony sensitize jurors to variations in confession evidence? Law and Human Behavior, 40(6), 638–649. https://doi.org/10.1037/lhb0000204

Kassin, S. M., Bogart, D., & Kerner, J. (2012). Confessions that corrupt: Evidence from the DNA exoneration case files. Psychological Science, 23(1), 41–45. https://doi.org/10.1177/0956797611422918

Kassin, S. M., Cleary, H. M. D., Gudjonsson, G. H., Leo, R. A., Meissner, C. A., Redlich, A. D., & Scherr, K. C. (2025). Police-induced confessions, 2.0: Risk factors and recommendations. Law and Human Behavior, 49(1), 7–53. https://doi.org/10.1037/lhb0000593

Kassin, S. M., Dror, I. E., & Kukucka, J. (2013). The forensic confirmation bias: Problems, perspectives, and proposed solutions. Journal of Applied Research in Memory and Cognition, 2(1), 42–52. https://doi.org/10.1016/j.jarmac.2013.01.001

Kukucka, J., & Kassin, S. M. (2014). Do confessions taint perceptions of handwriting evidence? An empirical test of the forensic confirmation bias. Law and Human Behavior, 38(3), 256–270. https://doi.org/10.1037/lhb0000066

Leo, R. A. (2008). Police interrogation and American justice. Harvard University Press.

Leo, R. A., & Ofshe, R. J. (1998). The consequences of false confessions: Deprivations of liberty and miscarriages of justice in the age of psychological interrogation. Journal of Criminal Law and Criminology, 88(2), 429–496. https://doi.org/10.2307/1144288

Vornhagen, J., & Hart, K. J. (2026). How evidence of confession contamination affects perceptions and judgments in a mock case. Psychology, Crime & Law. Advance online publication. https://doi.org/10.1080/1068316X.2026.2647334

Legal Authorities

Arizona v. Fulminante, 499 U.S. 279 (1991).

Bruton v. United States, 391 U.S. 123 (1968).

Crane v. Kentucky, 476 U.S. 683 (1986).

Opper v. United States, 348 U.S. 84 (1954).

Smith v. United States, 348 U.S. 147 (1954).

State v. Carlson, 237 Ariz. 381, 351 P.3d 1079 (2015).

State v. Gillies, 135 Ariz. 500, 662 P.2d 1007 (1983).

State v. Morgan, 204 Ariz. 166, 61 P.3d 460 (App. 2002).

United States v. Adams, 74 M.J. 137 (C.A.A.F. 2015).

United States v. Whiteeyes, 82 M.J. 168 (C.A.A.F. 2022).

18 U.S.C. § 3501.

Ariz. Rev. Stat. Ann. § 13-3988.

Manual for Courts-Martial, United States, Mil. R. Evid. 304 (2024 ed.).

Related topic papers: Contamination: Non-Public Details and Post-Admission Narratives · Investigator Bias, Tunnel Vision, and Forensic Confirmation Bias · Electronic Recording of Interrogations · How Jurors Respond to Confession Evidence · Arizona Confession Law · Confessions in Courts-Martial: Article 31(b), Mil. R. Evid. 304 and 305, and Expert Assistance

Related topics from Arizona Forensic Services: Child Abuse Allegations · Online Sex Offense