The witness does not hesitate.
“I'm one hundred percent certain.”
The statement sounds powerful because certainty is persuasive.
People naturally place more trust in someone who appears sure of what they remember than someone who says, “I think that's him.”
Inside a courtroom, that difference can be enormous.
A hesitant identification invites doubt.
A confident identification can feel like the end of the argument.
But eyewitness research has forced courts and investigators to confront a more complicated question:
When did the witness become so certain?
A witness who immediately identifies someone under well-controlled conditions and expresses high confidence presents one evidentiary situation.
A witness who initially says, “He looks familiar, but I'm not sure,” and then becomes completely certain after seeing the same suspect repeatedly presents another.
The courtroom may make those witnesses look identical.
Their identification histories are not.
Research summarized by the National Institute of Justice continues to examine how jurors evaluate eyewitness confidence, including the distinction between confidence expressed during the initial identification and confidence displayed much later in court.
That distinction matters because confidence can change.
And memory can change with it.
Confidence Is Persuasive Because We Use It Everywhere Else
Imagine asking two people for directions.
The first says:
“I think the courthouse is two blocks that way, but I'm not positive.”
The second says:
“I know exactly where it is. Follow this street for two blocks.”
Most people follow the second person.
Confidence acts as a shortcut for knowledge.
Usually that is useful.
Someone who has performed a job for twenty years often sounds more certain than someone attempting it for the first time.
Someone who clearly remembers an event may speak more confidently than someone whose memory is vague.
The problem is that confidence and accuracy are not interchangeable.
A person can be uncertain and correct.
A person can be confident and wrong.
Eyewitness evidence makes that distinction particularly consequential because the jury is not merely deciding whether to trust someone's directions.
It may be deciding whether another person committed a crime.
There Is an Important Nuance: Initial Confidence Can Matter
The lesson from eyewitness research is sometimes oversimplified into:
Confidence means nothing.
That is not an accurate description of the modern research.
Recent research has found a meaningful relationship between accuracy and a witness's confidence at the time of an initial identification, particularly when the identification is obtained under appropriate conditions.
A 2025 NIJ-supported nationwide survey of law-enforcement practices noted that research has demonstrated a strong predictive relationship between initial confidence and identification accuracy, while also finding that more than half of surveyed agencies were unaware of that recent evidence.
The timing matters enormously.
The witness's first confidence statement can contain useful information.
The witness's confidence months later may have a much more complicated history.
The Most Important Statement May Have Happened at the Police Station
Suppose a witness examines a six-person photo lineup.
The witness pauses.
Looks again.
Points to photograph number four.
The investigator asks:
“How certain are you?”
The witness responds:
“Maybe 60 percent. He looks the closest.”
That statement matters.
Now imagine the case proceeds for eighteen months.
The witness learns police arrested number four.
The witness sees the defendant at hearings.
The prosecutor discusses the case.
The witness thinks repeatedly about the event.
Then comes trial.
The prosecutor asks:
“Do you see the person who robbed you in this courtroom?”
The witness points at the defendant.
“Absolutely. I'll never forget his face.”
If the jury hears only the final statement, it receives a very different picture from the actual identification history.
This Is Why Researchers Want the First Confidence Statement Preserved
The National Academies examined decades of research on eyewitness identification and recommended that law enforcement record the witness's confidence verbatim at the time of the initial identification.
Not:
“Witness was confident.”
Not:
“Positive ID.”
The witness's actual words.
The committee explained that later confidence can be affected by feedback and other information encountered after the identification.
That recommendation recognizes something fundamental.
Confidence is evidence too.
And like other evidence, its condition when first collected can matter.
“I Think That's Him” Can Slowly Become “That's Him”
The change does not have to involve dishonesty.
The witness may never consciously decide to exaggerate.
Imagine the sequence.
First identification:
“I think that's him.”
Police arrest the person.
The witness thinks:
I must have picked correctly.
A detective later mentions:
“We found some other evidence too.”
Now the witness thinks:
I did pick correctly.
The witness sees the defendant at a preliminary hearing.
There he is again.
Months pass.
The face becomes increasingly familiar.
By trial, the witness is not pretending to be certain.
The witness is certain.
The difficult question is whether that certainty reflects the original memory or everything that happened afterward.
Confirmation Can Be Powerful
Suppose a witness chooses someone from a lineup and an investigator says:
“That's the person we suspected.”
Those few words introduce new information.
The witness now knows the police independently suspected the same person.
That can feel like confirmation.
Even subtle feedback can matter.
“Good.”
“You picked the suspect.”
“That's who the other witness identified.”
None of those statements changes what the witness originally saw.
But they can affect how the witness subsequently thinks about the identification.
This is one reason modern identification procedures try to separate the witness's memory from information possessed by investigators.
A Blind Administrator Solves a Surprisingly Human Problem
One reform recommended by the National Academies is double-blind lineup administration: neither the witness nor the administrator conducting the identification procedure should know which person is the suspect.
Why prevent the administrator from knowing?
Because people communicate without realizing it.
Imagine an investigator knows number three is the suspect.
The witness studies number one.
Nothing.
Number two.
Nothing.
Then number three.
Perhaps the investigator becomes slightly more attentive.
The witness hesitates.
“Take your time,” the investigator says.
Nobody intends to influence anybody.
But the interaction is no longer completely neutral.
If the administrator does not know who the suspect is, those suspect-specific cues become much harder to give.
A Fair Lineup Should Not Contain an Obvious Answer
Imagine police are looking for a man described as having a beard.
The lineup contains six photographs.
Five men are clean-shaven.
One has a beard.
The witness selects the man with the beard.
How much did the witness recognize the face?
How much did the lineup itself identify the expected choice?
The problem becomes even more obvious if the witness was told beforehand that police had arrested someone.
Now the witness may reasonably assume the person must be among the photographs.
The lineup becomes less like:
“Do you recognize anyone?”
and more like:
“Which of these people did it?”
Those are very different cognitive tasks.
The Perpetrator May Not Be There at All
One of the National Academies' recommendations is deceptively simple:
Tell witnesses that the perpetrator may or may not be in the lineup and that the investigation will continue whether or not they make an identification.
That instruction changes the witness's task.
Without it, the witness may compare the faces against one another.
Who looks most like the person I remember?
But the person who looks most like the perpetrator is not necessarily the perpetrator.
If the actual perpetrator is absent, every choice is wrong.
The correct response is:
“None of them.”
Familiarity Can Masquerade as Recognition
This is another danger that can develop over time.
A witness sees a stranger commit a crime.
Later, police show the witness a photograph.
The witness is uncertain.
Several weeks later, another identification procedure includes the same person.
That face is now familiar.
Why?
There are at least two possibilities.
The witness remembers the person from the crime.
Or the witness remembers the person from the earlier photograph.
The subjective experience can be similar:
I have seen this face before.
The difficult part is identifying where that familiarity came from.
Repeated exposure can therefore complicate an identification rather than strengthen it.
NIJ-sponsored research has found potential problems with repeated identification tasks involving the same witness.
The Defendant Becomes Extremely Familiar Before Trial
By the time a criminal case reaches trial, the witness may have encountered the defendant repeatedly.
Photographs.
News reports.
Court hearings.
Conversations with investigators.
Conversations with prosecutors.
Perhaps social media.
Then the witness enters court.
The defendant is sitting beside defense counsel.
The witness is asked to identify the perpetrator.
This is nothing like the first identification.
The defendant may now be one of the most familiar faces associated with the event.
That is why the original identification procedure can tell us something the dramatic courtroom identification cannot.
Courtroom Confidence Is Especially Persuasive
The problem is that jurors see the witness now.
They do not experience the witness eighteen months earlier.
They watch a person take an oath, look across the courtroom and point.
Research summarized by NIJ has found that potential jurors can place excessive weight on confidence expressed in court compared with confidence expressed immediately after the original identification.
That creates a peculiar evidentiary problem.
The jury may be most impressed by the confidence statement that occurred after the greatest opportunity for outside influence.
The Courtroom Can Hide the Evolution of the Memory
Consider these two presentations.
Version One
The witness testifies:
“I recognize him. I'm absolutely certain.”
That sounds extremely strong.
Version Two
The jury learns the entire sequence:
Immediately after the crime, the witness could provide only a general description.
Two days later, the witness selected the defendant's photograph but said, “I think it might be him.”
Police then told the witness the person had become their suspect.
The witness later saw the defendant's photograph in news coverage.
The witness saw him at two court hearings.
At trial, the witness says, “I'm absolutely certain.”
Same witness.
Same courtroom identification.
Very different evidentiary picture.
Video Recording the Lineup Can Preserve What a Police Report Cannot
A written report might say:
“Witness identified suspect number four.”
That sentence sounds definitive.
A recording might show something else.
The witness studies the photographs for two minutes.
Rejects number four initially.
Returns to it.
Says:
“Maybe this one.”
Asks whether the person has been arrested.
Says:
“I don't know. He looks similar.”
Then signs the identification form.
The report is not necessarily false.
The witness did select number four.
But it cannot communicate the entire process.
The National Academies therefore recommended video recording eyewitness identification procedures.
A recording preserves hesitation, timing, instructions, feedback and the witness's actual words.
Decision Speed Can Add Context
Researchers have also examined how quickly witnesses make identifications.
A person who immediately recognizes a face presents a different situation from someone who studies a lineup for several minutes and eventually settles on one photograph.
That does not create a universal stopwatch rule.
Human recognition is more complicated than that.
But NIJ-supported research has examined how identification speed and confidence interact and how jurors interpret those signals.
Again, the central point is that the identification is a process.
The final answer is only one part of it.
Jurors May Not Recognize a Biased Lineup
Some problems seem obvious once they are explained.
That does not mean ordinary jurors naturally detect them.
NIJ-supported research published in 2022 found that biased lineups were often not perceived as biased by study participants. When participants did recognize the bias, however, they placed less trust in the resulting suspect identification.
This helps explain why simply showing jurors a lineup may not be enough.
A procedure that looks normal to someone unfamiliar with identification science may contain features capable of influencing the witness.
Cross-Examination Has Limits
The traditional legal response to questionable testimony is cross-examination.
Ask the witness about the lighting.
Distance.
Stress.
Initial description.
Lineup.
Confidence.
Prior statements.
That remains an important safeguard.
But eyewitness research has raised questions about whether ordinary courtroom tools are always sufficient to expose mistaken identifications.
An NIJ-sponsored review of eyewitness research concluded that jurors can be insensitive to some factors affecting identification accuracy and examined expert psychological testimony as a way of helping jurors evaluate those factors.
The problem is easy to understand.
Cross-examination is excellent at exposing a liar.
A mistaken eyewitness may not be lying.
“Are You Sure?” Can Actually Strengthen the Witness
Imagine cross-examination:
“You only saw the man for fifteen seconds?”
“Yes.”
“It was dark?”
“Yes.”
“You were frightened?”
“Yes.”
“And you're telling this jury you're certain that's the man?”
The witness looks directly at the defendant.
“I have absolutely no doubt.”
The lawyer intended to weaken the identification.
The final answer may instead become the most memorable moment of the testimony.
A sincere but mistaken witness can be a formidable witness precisely because there is no lie to expose.
Memory Errors Do Not Look Like Lies
A lying witness may contradict known facts.
Avoid questions.
Change explanations.
Have an obvious motive.
A mistaken witness can be entirely different.
Cooperative.
Consistent.
Emotional.
Sincere.
Certain.
The witness may have spent years reliving the event.
That can make the testimony extraordinarily compelling.
The jury is therefore being asked to make a distinction people rarely have to make in ordinary life:
Do I believe this person is telling the truth while also believing the person is wrong?
Those propositions can coexist.
DNA Exonerations Made That Possibility Impossible to Ignore
Eyewitness misidentification has played a significant role in wrongful-conviction cases later exposed through DNA testing. NIJ describes eyewitness error as one of the factors associated with wrongful convictions and continues to support research aimed at improving identification evidence.
DNA changed the conversation because it sometimes provided an independent answer to identity.
The witness could remain certain.
The biological evidence could establish that someone else committed the crime.
That forced courts and investigators to confront a fact that psychology had long warned about:
Subjective certainty cannot make an inaccurate memory accurate.
Confidence Can Grow Because the Case Grew Around the Witness
Suppose the witness makes a tentative identification.
Then police discover that the suspect owns a similar vehicle.
The suspect lives nearby.
The suspect is arrested.
The prosecutor files charges.
The case survives preliminary proceedings.
To the witness, all of this can look like independent confirmation.
Why would police arrest the person if I were wrong?
Why would prosecutors bring the case?
Why would the case still be going forward?
Those institutional decisions can reinforce the witness's belief.
Yet some of those decisions may themselves have depended heavily on the original identification.
The evidence begins to form a circle.
Two Witnesses Can Reinforce Each Other
Imagine two people witness the same crime.
Police interview them separately.
Good.
Later, however, they talk.
“I picked number three.”
“So did I.”
Now both witnesses know the other made the same selection.
That can increase confidence.
If one witness had been uncertain, agreement may feel validating.
This is why separating witnesses early in an investigation can matter.
Independent memories are most valuable while they remain independent.
News Coverage Can Become Part of the Memory Environment
A suspect is arrested.
The person's photograph appears on television.
The witness sees it.
Again.
And again.
By trial, the defendant's face is deeply associated with the crime.
This does not mean the witness will necessarily become mistaken.
But it makes the origin of familiarity more difficult to separate.
Was the face memorable because of the original event?
Or because it has been encountered dozens of times since?
The witness may not be able to answer that question consciously.
Social Media Makes Repeated Exposure Easier Than Ever
Modern cases add another complication.
A witness can search for the suspect.
Look at photographs.
Read comments.
Watch videos.
See what other witnesses are saying.
None of that information existed inside the witness's original memory.
Once absorbed, however, separating sources becomes difficult.
Investigators cannot control everything a witness encounters after a crime.
That makes documenting the initial memory even more important.
An Early Mistake Can Become a Stable Memory
Memory does not necessarily become progressively less confident as it ages.
Sometimes the opposite happens.
A person repeatedly retrieves the same reconstructed version of an event.
Each retrieval reinforces it.
The story becomes familiar.
The details become stable.
The witness becomes more comfortable telling it.
Consistency can therefore be misleading.
A story repeated the same way for years may feel reliable.
But if the original reconstruction contained an error, repetition can preserve the error too.
This Is Why the First Description Matters So Much
Immediately after the event, a witness says:
“He had a round face and no facial hair.”
The defendant has a narrow face and a beard.
Months later, the witness says:
“I remember the beard clearly.”
That inconsistency deserves attention.
The initial description was produced before the defendant became familiar.
It may therefore provide important evidence about what the witness actually encoded during the event.
Later detail is not automatically false.
Memory can be refreshed by legitimate information.
But new specificity should not simply erase old uncertainty.
Courts Have Had to Decide When Identification Procedures Become Constitutionally Problematic
The Supreme Court's eyewitness-identification cases distinguish between unreliable evidence generally and identification evidence affected by improper law-enforcement conduct.
In Perry v. New Hampshire, the Court held that the Due Process Clause does not require a preliminary judicial reliability inquiry merely because an identification occurred under suggestive circumstances; the constitutional screening rule at issue is triggered by unnecessarily suggestive circumstances arranged by law enforcement.
That means not every questionable identification is automatically excluded before trial.
Traditional trial protections—including cross-examination, evidentiary rules, counsel and jury instructions—remain important.
Different states can also provide additional protections under their own law.
Admissible Does Not Mean Infallible
This distinction appears throughout evidence law.
A judge allowing the jury to hear an identification does not necessarily mean the judge has declared it correct.
The jury may still have to decide what weight it deserves.
That is why instructions, expert testimony and evidence about the identification procedure can matter.
The legal system frequently admits evidence containing uncertainty.
The fact-finder's task is to evaluate that uncertainty.
Modern Research Has Changed Some Older Assumptions
Eyewitness science continues to develop.
One good example involves sequential and simultaneous lineups.
An older theory suggested that showing photographs sequentially—one at a time—might be categorically superior to showing them simultaneously.
The modern picture is more complicated.
The 2025 NIJ-supported national survey noted that recent research has not established sequential arrays as simply superior to simultaneous arrays, and the IACP model policy recognizes both formats as valid.
That is an important reminder about science itself.
Reforms should follow evidence, not slogans.
As research improves, procedures may need to change again.
The Goal Is Not to Distrust Eyewitnesses
Eyewitnesses can provide essential evidence.
Sometimes they are the only people who saw what happened.
They can identify perpetrators accurately.
They can provide details no camera captured.
They can help investigators find evidence that otherwise would never be discovered.
NIJ has long recognized both sides of the issue: eyewitness evidence can be crucial, but even sincere witnesses can make identification mistakes.
The goal is therefore not to teach jurors:
“Never believe an eyewitness.”
It is to ask better questions about how the identification was produced.
The Better Question Is Not Simply “How Certain Are You?”
A more complete set of questions would be:
What did the witness originally see?
For how long?
Under what conditions?
What description did the witness initially give?
How was the suspect first presented?
Were appropriate fillers used?
Was the witness told the perpetrator might not be present?
Did the administrator know which person was the suspect?
What did the witness say immediately after choosing?
Was that statement recorded verbatim?
Did anyone provide confirming feedback?
How many times did the witness see the suspect afterward?
Did the witness encounter photographs or news coverage?
How did confidence change over time?
That history tells us far more than one number given from the witness stand.
A Witness Who Says “Seventy Percent” May Be Giving Better Evidence
There is something psychologically uncomfortable about uncertainty.
Jurors want answers.
Investigators want answers.
Victims want answers.
But an honest expression of uncertainty can be valuable evidence.
“I think it's him, but I'm not certain.”
That may accurately describe the state of the witness's memory.
Pressure to convert uncertainty into certainty does not improve the underlying memory.
It merely makes the testimony sound stronger.
A justice system interested in accuracy should preserve uncertainty when uncertainty genuinely exists.
The Police Report Should Not Upgrade the Witness's Words
Consider the difference.
Witness:
“I think number five looks the most like him.”
Report:
“Witness identified number five as the perpetrator.”
The second statement may technically describe what occurred, but it loses something crucial.
Uncertainty.
This is why verbatim confidence statements are so valuable.
They prevent institutional language from unintentionally strengthening evidence.
“Possible identification” and “positive identification” can sound worlds apart to a later reader.
The witness's own words are better.
A Jury Should Be Able to See the Identification Before It Became a Courtroom Performance
The ideal evidentiary record preserves the original event as closely as possible.
The witness's first description.
The lineup instructions.
The lineup itself.
The identification procedure.
The witness's response.
The witness's confidence.
Any feedback.
If those things are recorded, a jury can evaluate the identification historically.
Without them, the jury may see only the finished product:
A witness who has had months or years to become familiar with the defendant and certain about the case.
Confidence Should Have a Timestamp
This may be the simplest way to understand the problem.
“I'm 95 percent certain” is incomplete information.
When?
Five minutes after the crime?
During the first properly administered lineup?
After police said the suspect's fingerprints were found?
After seeing the defendant's photograph on television?
After attending three hearings?
At trial two years later?
Confidence is not a permanent property attached to a memory.
It exists at a particular moment in the history of that memory.
That moment matters.
The Strongest Witness May Be the One Whose Uncertainty Was Preserved
A witness who says:
“I was 80 percent certain when I first identified him, and that's still the best description of my confidence”
may sound less impressive than someone declaring absolute certainty.
But from an evidentiary standpoint, the first witness may be giving the court something valuable:
An honest account of the limits of memory.
Certainty sounds powerful.
Accuracy matters more.
The Courtroom Sees the End of a Long Process
By trial, the witness's memory has a history.
The original event.
The first interview.
The first description.
The lineup.
The police response.
The arrest.
News coverage.
Court appearances.
Conversations.
Preparation for testimony.
Then, finally, the witness takes the stand.
The jury encounters the memory at the end of that journey.
The central challenge is determining how much of what the witness now believes came from the original event and how much accumulated afterward.
That is why the most dramatic moment in an eyewitness case may not be the witness pointing across the courtroom.
The more important moment may have occurred months earlier in a quiet interview room, when the witness first looked at six photographs, paused over one, and said:
“I think that's him—but I'm not completely sure.”