Cognitive Bias in Criminal Investigations: Tunnel Vision and Confirmation Bias
Author: Ukrainian Psychological Hub · Published: October 11, 2026 · Editorial Policy
Ukrainian Psychological Hub | Psychology for the AI Era. Updated October 11, 2026.
An investigation begins with uncertainty. Detectives form hypotheses about what happened, who was involved, and which leads deserve attention. Those hypotheses organize necessary work, yet they can also change what investigators notice, how they interpret ambiguous details, which alternatives they test, and how confidently they describe the resulting case. Cognitive bias in criminal investigations is the study of these systematic influences on the search for, interpretation of, and integration of evidence.
Two particularly important concepts are confirmation bias—the tendency to favor information consistent with an existing belief—and investigative tunnel vision, a case-level narrowing of attention and inquiry around one explanation or suspect. The problem can arise in competent, conscientious teams without intentional dishonesty. A review of 30 criminal case-evaluation studies and a separate systematic review of 29 forensic-science studies show that the relevant mechanisms are empirically grounded, while their size, conditions, and effective remedies vary.
What Is Cognitive Bias in a Criminal Investigation?
A cognitive bias is a predictable pattern in judgment or information processing that may shift perception, memory, information search, or interpretation. In investigative settings, the question is whether decisions respond appropriately to the evidentiary value of new information or whether prior expectations exert influence beyond what that evidence warrants. Bias can affect the choice of witnesses to interview, what questions are asked, which exhibits receive testing, how inconsistencies are recorded, and how conflicting findings are reconciled.
Bias has several meanings in everyday language. Here it refers primarily to cognitive mechanisms and decision procedures, not necessarily prejudice, misconduct, or deliberate manipulation. Discrimination and unequal treatment raise additional legal and ethical concerns that require direct examination in their own right. Cognitive bias can coexist with them, but the concepts should be analyzed separately. The Meterko and Cooper review frames the problem as a combination of general human tendencies, situational conditions, and organizational factors.
Confirmation bias
Confirmation bias occurs when an investigator preferentially seeks, notices, remembers, interprets, or gives greater weight to information supporting an existing hypothesis, while applying greater skepticism to contradictory information. It can affect both evidence collection and case interpretation. For example, after identifying a suspect, a team may treat a vague witness recollection as support but demand unusually strong proof before crediting a contradictory alibi.
The defining problem is asymmetry of evaluation, not the mere fact that investigators have working hypotheses. In a sound inquiry, a favored hypothesis can be revised or rejected. Evidence should be examined for its capacity to distinguish among credible explanations, including explanations inconsistent with an initial suspicion.
Investigative tunnel vision
Tunnel vision describes the wider investigative and institutional process through which one account progressively dominates attention and resources. It may involve an early suspect focus, increasingly narrow search decisions, repeated reinterpretation of conflicting information, and reinforcement as the case proceeds through investigators, forensic laboratories, prosecutors, and courts. Findley and Scott's legal analysis explains how ordinary cognition can interact with role pressures and adversarial structures to make a case theory difficult to dislodge.
The terms overlap, but they identify different levels of analysis. Confirmation bias names a cognitive tendency; tunnel vision describes its possible cumulative expression across a case. A single biased evaluation does not establish that an entire investigation suffered tunnel vision. Likewise, a narrowly focused investigation may be justified by strong evidence, provided alternative explanations were genuinely evaluated.
How Tunnel Vision Develops: From First Lead to Fixed Narrative
1. An initial hypothesis becomes an anchor
Early witness statements, a tip, an apparent motive, or an automated match may place one person or explanation at the center of the investigation. Early judgments are useful starting points. The risk increases when the first hypothesis becomes an implicit standard against which all later evidence is judged. Studies of simulated criminal case files show that early commitment can guide later search and evaluation. In O'Brien's two studies, making an early hypothesis explicit increased confirmatory information processing; considering why a hypothesis might be wrong reduced bias in the second study.
2. The search for evidence becomes selective
Once a suspect is favored, investigators may prioritize tests and interviews most likely to confirm involvement while delaying inquiries that could identify another person. Selection matters because the material subsequently available for decision-making is partly produced by earlier investigative choices. The absence of contradictory findings may reflect a failure to look rather than genuine exclusion of alternatives. Rassin, Eerland, and Kuijpers found that the selection of follow-up investigations varied with initial guilt assessments and features of mock cases, although their participants did not generally select more incriminating than exonerating options overall.
3. Ambiguous facts are interpreted through the dominant theory
Many case facts admit competing explanations: a delayed disclosure, inconsistent estimates of time, a partial trace, a witness's uncertainty, or a person's presence near a location. Investigators can evaluate such facts differently depending on what they already believe. In the case-evaluation literature, identical or comparable evidence has sometimes received different credibility judgments when it favored versus challenged a working theory. The 2022 research review describes this problem as asymmetric skepticism and highlights differences across tasks and samples.
4. Repetition creates apparent corroboration
A theory repeated in briefings, reports, interview questions, and laboratory submissions may acquire the appearance of independent support. However, statements derived from the same initial assumption are not separate pieces of evidence. A forensic examiner's opinion informed by a detective's suspicion and a detective's confidence strengthened by that opinion may form an information-feedback loop. Analysts should track which conclusions depend on which sources and distinguish genuinely independent observations from recycled interpretations.
5. Organizational commitment increases the cost of revision
Supervisors, specialist units, public announcements, time limits, and demands for case closure can affect decisions about what to investigate next. A team may invest considerable labor in one account, making it psychologically and institutionally harder to reopen foundational assumptions. Such pressures are plausible contributors identified in the legal and empirical literature, but their occurrence and magnitude must be established case by case. The existence of strong commitment alone does not prove an investigation was erroneous.
Where Cognitive Bias Can Enter the Investigative Process
Witness identification and witness reliability
Investigators may give early identifications special importance, allow knowledge of a suspect to influence lineup administration, or interpret subsequent confidence as if it were independently formed. Biased questioning and post-identification feedback create separate risks that must be examined with eyewitness-specific research. The general lesson here is procedural: preserve each witness's original account, separate source information from later discussion, and record the conditions under which identification and confidence were first expressed.
Interviews and interrogation
A presumption of guilt may shape question selection, expectations about denials, and interpretations of hesitation, memory gaps, or changing details. Those behaviors have multiple explanations and are not reliable stand-alone proof of deception. Interviewing designed principally to confirm a belief may produce less opportunity to test innocent or alternative accounts. Research on investigative bias informs this problem, while interrogation methods, confession reliability, and specific safeguards require their own dedicated evidence base.
Crime scene reconstruction and forensic testing
A laboratory can receive a sample along with a suspect's criminal history, an officer's view of guilt, or another analyst's earlier conclusion. Some contextual details are necessary for the scientific task; others are irrelevant and potentially biasing. The Cooper and Meterko systematic review found evidence of contextual influences on forensic conclusions, particularly in studies in which practitioners were exposed to suspect-related information or previous decisions. The review does not show equal vulnerability in every forensic discipline or examination type.
The NIST human-factors report on latent-print examination recommends a systems approach to examination quality, documentation, verification, and managing sources of human error. A finding about latent fingerprints should not be automatically generalized to every laboratory method, and a forensic match alone should not be treated as proof of criminal responsibility.
Digital traces, databases, and algorithm-assisted leads
Investigators increasingly encounter search rankings, digital evidence summaries, and algorithm-assisted matches. These outputs can direct attention and introduce anchoring if a ranked candidate is mistaken for a validated identification. The correct questions concern the tool's task-specific error rates, input quality, reference population, opportunity for independent verification, and whether investigators searched for disconfirming evidence. The same caution applies when generative AI summarizes case material: fluent summaries can omit qualifiers or create apparent coherence. This is a procedural implication, not a claim that every tool has a demonstrated bias effect of the same size.
Prosecutorial review and courtroom presentation
Information can be filtered again when investigators summarize evidence for prosecutors or prepare testimony. Reports that highlight confirmatory material and compress unresolved contradictions may influence downstream decisions. Prosecutors' disclosure duties, evidentiary rules, and remedies vary by jurisdiction; cognitive-science findings do not determine whether a particular constitutional, statutory, or professional rule was violated. Scientific assessment asks how information was processed. A court determines the applicable law and its consequences.
What Does the Research Actually Show?
Evidence from reviews of criminal case evaluation
In Cognitive Biases in Criminal Case Evaluation, Meterko and Cooper reviewed 30 primary-data and intervention papers, 18 involving police participants or police documents. They found supporting evidence that professionals and laypeople can be influenced by prior expectations. Importantly, the studies differed in designs and outcomes, and only two professional-participant studies in the review were conducted in the United States. Most used vignettes, fictional or adapted case files, and controlled decisions rather than full investigations conducted under ordinary field conditions.
The review also identified six studies evaluating intervention approaches, with varying results. This matters for practice: identifying a risk does not automatically validate every proposed checklist, training session, or organizational reform. Assessments of countermeasures should measure actual search behavior, evidence quality, error patterns, and downstream outcomes where feasible.
Evidence from forensic laboratories
The 2019 systematic review by Cooper and Meterko examined 29 forensic-science studies, two of which had major methodological deficiencies. Its stronger findings concerned the influence of suspect-related context, exemplar selection, and knowledge of prior decisions in the settings tested. The review identified potential benefits of limiting task-irrelevant information, providing comparison samples appropriately, and blinded replication. Eleven of the included studies involved latent fingerprints, making extrapolation to less-studied disciplines particularly uncertain.
Evidence from experiments with investigators and laypeople
In a study involving Israeli investigators and lay participants responding to hypothetical criminal scenarios, Elaad (2022) reported stronger tunnel-vision-related responses among police participants on several measures. The full findings were more nuanced than a simple claim that investigators ignore every exculpatory fact: participants could reduce guilt assessments after exonerating information. The scenarios were hypothetical and should not be interpreted as a direct estimate of wrongful convictions in actual police practice.
Rassin and colleagues reported effects of initial guilt evaluation and case characteristics on selected investigative actions across three analogue studies. Crucially, the overall proportion of selected incriminating actions did not generally exceed half. Such mixed results improve the scientific picture: confirmation bias is a vulnerability whose expression depends on how tasks, evidence, incentives, and available alternatives are structured.
What current studies cannot establish
The literature does not provide a single reliable percentage of criminal investigations affected by tunnel vision, a universal cognitive-bias score for detectives, or an individual-case diagnostic test proving that bias caused a conviction. Many studies are small, simulated, or drawn from jurisdictions with different training and legal procedures. Observational reconstructions may identify plausible mechanisms but cannot always isolate the counterfactual outcome: what would have happened under another investigation strategy.
Evidence of a biased process is distinct from evidence that a suspect is innocent, and an exoneration does not by itself reveal the precise psychological causes of the original error. These propositions require separate records, analyses, and legal adjudication. A sound review considers competing explanations for every failure, including inadequate evidence, mistaken procedures, resource constraints, and deliberate misconduct when documented.
A Documented Example: The Brandon Mayfield Fingerprint Error
The FBI's erroneous fingerprint identification of Brandon Mayfield after the March 2004 Madrid train bombings illustrates how an initial comparison can become resistant to conflicting information. The U.S. Department of Justice Office of the Inspector General review describes how examiners relied on apparent similarities, interpreted ambiguous details by reference to a known print, and failed to sufficiently reconsider their conclusion after Spanish authorities questioned the match. The FBI withdrew the identification, and Mayfield was released.
The official review identified examination errors and systemic problems. It also found that information about Mayfield learned later in the investigation likely contributed to a failure to reassess the result after the Spanish police reached a different conclusion. The review did not find that Mayfield's religion caused the examiners' initial fingerprint identification. This distinction matters: documented sequence, identified mechanism, and an investigator's subjective motive are different levels of claim.
The case supports reforms in documentation, independent verification, and control of contextual information. It is not a statistical demonstration of the frequency of such errors, and it does not establish that every forensic discipline behaves like fingerprint examination. Follow-up DOJ OIG reporting and NIST's human-factors recommendations offer concrete institutional context.
Why Expertise and Good Intentions Are Not Enough
Experience helps investigators interpret many forms of evidence and recognize patterns. It can also strengthen familiar expectations, particularly when a pattern is incomplete or new evidence is ambiguous. Training in professional ethics remains indispensable but does not remove the basic cognitive tendency to use prior information as an interpretive frame. The remedy therefore involves methods that make errors observable and correctable.
An investigative team can act in good faith and still miss a viable alternative; conversely, a team may pursue one leading theory because the evidence has legitimately excluded the others. The differentiating question is whether the process preserved fair tests of the competing explanations and recorded how contrary findings changed the assessment. An independent reviewer needs access to the decision trail rather than just the final narrative.
How Investigative Teams Can Reduce Confirmation Bias
Preserve multiple working hypotheses with explicit tests
At defined points in a case, investigators can write down the leading explanation, at least one serious alternative, and the observations that would make each explanation less plausible. Hypotheses should be evaluated according to evidence that genuinely discriminates among them. Merely listing alternative suspects without funding or performing meaningful checks produces the appearance of open-mindedness without testing it.
Ask what would falsify the favored explanation
An effective review identifies facts that would challenge the current theory: verified locations, timestamps, absence or presence of trace evidence under appropriate conditions, independent accounts, or incompatible timelines. O'Brien's experiment found that considering why an initial hypothesis might be incorrect reduced confirmatory processing in a mock-file task. This supports a promising technique, while not proving that a single prompt eliminates bias in every real investigation.
Keep exculpatory and contradictory leads visible
A case management system should preserve reports of inconsistent witness statements, laboratory exclusions, alternate-source leads, unresolved alibis, and reasons for closing or reopening an avenue of inquiry. Investigators should document the status of each lead and the evidentiary basis for decisions. An inconsistent detail is not automatically exculpatory; its value depends on the alternatives and circumstances. Transparent retention permits meaningful reassessment.
Separate task-relevant evidence from contextual suggestion
Forensic examiners should receive the information required to perform and interpret their particular task. Irrelevant claims about suspect character, confessions, investigator expectations, or unrelated history may be withheld until the relevant examination is completed. Linear Sequential Unmasking–Expanded and its casework implementation guidance propose ordering information by relevance, objectivity, and potential biasing power. These are research-informed procedural frameworks; field effectiveness and local operational feasibility need continuing evaluation.
Use genuinely independent verification
A second review is most informative when the verifier can reach a conclusion independently and does not know the original analyst's conclusion or an investigator's desired result. Merely repeating an examination with the same context can reproduce the same error. The NIST report and forensic-bias review address verification, documentation, and the role of blinding. Independence must be defined according to the method and stage of the examination.
Record confidence, uncertainty, and evidence lineage
Investigative notes should show what an officer knew when a judgment was made, what evidence prompted a change, what remained uncertain, and which statements depended on earlier assumptions. Where a technique has known limitations, those should accompany the conclusion. A transparent record helps reviewers distinguish contemporaneous reasoning from hindsight and prevents circularly sourced propositions from being counted as independent corroboration.
Build structured challenge into supervisory review
Supervisors and multidisciplinary teams can periodically assign a reviewer to test the strongest alternative account and examine neglected leads. Reviews should be scheduled before key irreversible steps where feasible and should protect dissent from interpersonal or organizational penalties. Quality assurance requires evaluating whether the challenge produces useful information rather than measuring whether reviewers agree with the original team.
Evaluate countermeasures instead of assuming they work
Policy makers should compare outcomes before and after intervention, use independent quality assessments where possible, and look for unintended effects. Better documentation, lower rates of reversals, and more complete follow-up of contradictory leads can provide useful process indicators, while actual correctness remains difficult to observe in many cases. The Meterko and Cooper review emphasizes the limited and mixed intervention evidence.
A Practical Investigative Bias Audit
A case review can ask: What were the initial hypotheses and when were they adopted? Which evidence was gathered before a suspect was selected? Were tests and interviews chosen because they could distinguish explanations, or primarily because they could confirm one? What exculpatory or inconsistent information was collected, and how did it affect the theory? Which laboratory or witness conclusions depended on investigator-supplied context? Was verification meaningfully independent? Could a new team reconstruct the sequence of knowledge and decisions from preserved records?
These questions are a procedural audit, not a standardized psychometric instrument. They cannot assign a validated numeric risk of wrongful conviction or determine legal guilt. Their value is to expose identifiable decision points for professional analysis, independent testing, and documented review.
How Bias Relates to Wrongful Convictions
Tunnel vision can help explain how separate investigative weaknesses interact. An early suspect hypothesis may shape a witness interview; a witness's uncertain statement may guide a laboratory request; an ambiguous laboratory result may then appear to corroborate the interview. The result can become more persuasive than its underlying pieces warrant, particularly when their shared dependence on a prior assumption is obscured. Findley and Scott analyze this cumulative case-building problem.
Wrongful-conviction evidence documents the seriousness of these pathways, but exoneration databases do not yield an unbiased population estimate of investigative tunnel vision. Cases discovered and corrected are a selected subset of all investigations. Moreover, wrongful convictions can involve multiple documented causes, and assigning a single cognitive mechanism without a contemporaneous record risks hindsight bias. Case-specific evaluations should distinguish factual findings, inference about processes, and legal determinations.
Legal Standards, Expert Testimony, and Professional Ethics
Cognitive psychology supplies scientific findings about information processing. Criminal procedure determines which investigative actions are permitted, which materials must be disclosed, how expert evidence is assessed, and what consequences follow from errors. Those rules differ among U.S. states, between federal and state proceedings, and across countries. A psychological account of confirmation bias cannot establish a constitutional violation, a Brady disclosure failure, evidentiary inadmissibility, or actual innocence without analysis of the governing law and case record.
In forensic psychological practice, the American Psychological Association Specialty Guidelines for Forensic Psychology stress impartiality, adequate scientific foundations, consideration of rival hypotheses, role clarity, and accurate reporting of limitations. They are professional guidance, not one universally binding law. Expert opinions should distinguish behavioral observations, psychological mechanisms, psychometric results where relevant, assumptions, and the legal question presented.
A psychological expert may assess the plausibility of a bias mechanism or explain methods for reducing contextual influence. Determining whether that mechanism operated in a particular investigation ordinarily requires reviewing documentation, testimony, and the timing of information exposure. The expert should not equate a finding of biased reasoning with a verdict of guilt or innocence.
How to Interpret the Concepts Without Common Myths
Myth: Only inexperienced or dishonest investigators show confirmation bias
Bias is a general feature of human cognition and can influence trained professionals. The relevant risks often arise through apparently reasonable decisions made under uncertainty. Expertise, integrity, and robust procedures contribute to good practice; systematic checks remain necessary.
Myth: Any early suspect focus proves tunnel vision
Investigations must prioritize leads. A narrow focus can be evidence-based when alternatives have been proportionately tested and excluded. Tunnel vision becomes a concern when the narrowing is insulated from meaningful contrary evidence.
Myth: One debiasing training session solves the problem
The intervention literature is limited and heterogeneous. Structural safeguards, contemporaneous documentation, independent review, and task-specific validation offer more auditable protection than confidence in awareness alone. Their effects should continue to be measured.
Myth: Confirmation bias proves wrongful conviction
A biased process may undermine reliability without establishing the ground truth of the alleged offense. The reverse is also possible: a correct conclusion may be reached through a poor decision process. Scientific criticism, evidentiary weight, and legal findings require separate analysis.
Myth: Forensic results are either objective or subjective
Forensic disciplines differ in measurement structure, error rates, and degree of interpretive judgment. Context controls and blind verification can improve some processes, but do not replace method validation, proficiency testing, clear uncertainty reporting, or substantive evidentiary assessment.
Frequently Asked Questions
What is an example of confirmation bias in a police investigation?
A detective identifies a suspect early, interprets ambiguous statements as incriminating, but dismisses comparable contradictions in the suspect's favor without equivalent examination. The key feature is unequal treatment of evidence based on whether it supports the working hypothesis.
Are confirmation bias and tunnel vision the same?
They overlap. Confirmation bias is a cognitive tendency in searching for or evaluating information. Tunnel vision describes the wider narrowing of a case around one theory, potentially involving several people, institutions, and successive decisions.
Can trained police officers be affected?
Yes. Experiments and reviews involving professional participants document vulnerabilities. The effect is conditional, and the current literature does not justify claiming that all officers or all investigations are affected to the same degree.
Does tunnel vision mean the suspect is innocent?
No. It describes concerns about a decision process. Determining guilt or innocence requires an evaluation of the case evidence under the applicable legal standards.
What is the difference between bias and misconduct?
Cognitive bias can be unintentional. Misconduct concerns behavior evaluated under professional, institutional, or legal rules and may be intentional or otherwise culpable depending on those rules. One finding does not automatically establish the other.
How can forensic laboratories limit contextual bias?
They can restrict irrelevant case information, manage the sequence of disclosures, document examinations before viewing comparison material when methodologically appropriate, and use independent blinded verification. The precise procedure must be adapted and tested for each discipline.
Can a biased investigation still produce valid evidence?
Yes. Individual evidence items must be evaluated for collection integrity, analytic validity, independence, and relevance. A process concern should prompt careful scrutiny rather than automatic acceptance or rejection of all results.
Can you calculate a confirmation-bias score for an investigation?
There is no universally validated score that translates case notes into a probability that confirmation bias caused an error or conviction. Structured quality reviews can identify deviations and conditions requiring additional scrutiny.
Does the research apply equally in every U.S. state and country?
General cognitive mechanisms have broad relevance, but training, procedural rules, institutional incentives, evidence systems, and legal remedies vary. Empirical results obtained in one setting need jurisdiction- and task-specific interpretation.
What should a defense attorney or case reviewer examine?
A chronology of what investigators knew, when the suspect became the dominant focus, the handling of contradictory leads, witness-interview records, laboratory submissions, verification procedures, and how the final narrative cites its sources. Legal actions must follow the local procedural framework.
Conclusion
Cognitive bias in criminal investigations is best understood as a problem of how hypotheses shape evidence production and evaluation across time. Confirmation bias can favor an established interpretation; tunnel vision can turn that tendency into an increasingly self-reinforcing case narrative. Empirical research documents vulnerabilities and supports carefully designed information controls, open testing of alternatives, independent verification, and transparent decision records. The strongest investigative practice preserves the ability to change a conclusion when the evidence changes.
For broader context, see Investigative Psychology: What It Is and How It Differs From Criminal Profiling.
Related Articles
References
American Psychological Association. (2013). Specialty guidelines for forensic psychology. American Psychologist, 68(1), 7–19. Professional guidelines.
Cooper, G. S., & Meterko, V. (2019). Cognitive bias research in forensic science: A systematic review. Forensic Science International, 297, 35–46. https://doi.org/10.1016/j.forsciint.2019.01.016.
Dror, I. E., & Kukucka, J. (2021). Linear Sequential Unmasking–Expanded (LSU-E): A general approach for improving decision making as well as minimizing noise and bias. Forensic Science International: Synergy, 3, 100161. https://doi.org/10.1016/j.fsisyn.2021.100161.
Elaad, E. (2022). Tunnel vision and confirmation bias among police investigators and laypeople in hypothetical criminal contexts. SAGE Open. https://doi.org/10.1177/21582440221095022.
Findley, K. A., & Scott, M. S. (2006). The multiple dimensions of tunnel vision in criminal cases. Wisconsin Law Review, 2006, 291. University of Wisconsin Law School repository.
Meterko, V., & Cooper, G. (2022). Cognitive biases in criminal case evaluation: A review of the research. Journal of Police and Criminal Psychology, 37, 101–122. https://doi.org/10.1007/s11896-020-09425-8.
National Institute of Standards and Technology. (2012). Latent print examination and human factors: Improving the practice through a systems approach (NISTIR 7842). Official publication.
National Research Council. (2009). Strengthening forensic science in the United States: A path forward. National Academies Press. Official report.
O’Brien, B. (2009). Prime suspect: An examination of factors that aggravate and counteract confirmation bias in criminal investigations. Psychology, Public Policy, and Law, 15(4), 315–334. https://doi.org/10.1037/a0017881.
Quigley-McBride, A., Dror, I. E., Roy, T., Garrett, B. L., & Kukucka, J. (2022). A practical tool for information management in forensic decisions: Using Linear Sequential Unmasking-Expanded (LSU-E) in casework. Forensic Science International: Synergy, 4, 100216. https://doi.org/10.1016/j.fsisyn.2022.100216.
Rassin, E., Eerland, A., & Kuijpers, I. (2010). Let's find the evidence: An analogue study of confirmation bias in criminal investigations. Journal of Investigative Psychology and Offender Profiling, 7, 231–246. https://doi.org/10.1002/jip.126.
U.S. Department of Justice, Office of the Inspector General. (2006). A review of the FBI's handling of the Brandon Mayfield case. Official report.