---
author: TED
date: '2026-04-05'
guest: ''
layout: post.njk
source: https://www.youtube.com/watch?v=93ocNhHyGVo
speaker: TED
tags:
  - eyewitness-testimony
  - memory-contamination
  - wrongful-convictions
  - forensic-psychology
  - legal-reform
title: 证人记忆的真相：从不可靠到可靠的关键科学洞见
summary: 本讲座深入探讨了证人记忆的复杂性，挑战了其普遍被认为不可靠的观点。演讲者强调，记忆的可靠性更多取决于测试方式和时机，而非其内在属性。通过阐述科学研究和真实案例，指出了早期、未受污染的记忆测试的重要性，并提出了改革法律体系以准确应用证人记忆的建议，旨在减少冤假错案。
insight: ''
draft: true
series: ''
category: psychology
area: society-thinking
project: []
people: []
companies_orgs: []
products_models: []
media_books: []
status: evergreen
---
### 证人记忆的可靠性：从绝对确定到科学审视

我们常常认为，当一个人对目击犯罪的记忆确凿无疑时，这种记忆就是可靠的。演讲者以一个设想开场：如果你确信自己目睹了犯罪，并且愿意为此在法庭上宣誓作证，你的记忆清晰、准确、无可动摇。然而，这种感觉百分之百真实的记忆，实际上可能是错误的，并可能导致无辜者被判入狱，甚至面临死刑。这就是目击证人记忆的复杂而令人心碎的世界。数十年以来，我们对于证人记忆的理解，可能本身就存在偏差。

大多数人可能都听说过关于证人证词极不可靠的警示故事。例如，**Ronald Cotton** 的案件，强奸案受害者 **Jennifer Thompson** 错误地指认了他为袭击者。Thompson 在审判中回忆道：“当我站上证人席时，我绝对、肯定、毫无疑问地确信他就是强奸我的人。没有人能改变我的想法。”陪审团基于她的证词判定 Cotton 有罪并判处终身监禁。然而，Ronald Cotton 并非真凶。真凶是当时在该社区活动的另一名已知罪犯 **Bobby Poole**。Cotton 在监狱中度过了近11年，直到 DNA 检测最终证明了他的清白，并揭示了真正的罪犯。Thompson 的证词是真诚的，但她的记忆是错误的。

这类涉及自信错误的辨认，在 DNA 翻案的案例中已发生数百次，促使人们严重质疑证人记忆的可靠性。科学研究也长期描绘了人类记忆的“毁灭性”图景。自20世纪70年代起，**Elizabeth Loftus** 等科学家便开始展示，操纵记忆是多么容易。他们在开创性研究中，向成年人植入了关于童年时在购物中心走失，或被凶猛动物袭击的虚假记忆，尽管这些事件从未发生。这些发现似乎证实了我们最深的恐惧：记忆并非录像带，它们更像是犯罪现场的证据，在被收集过程中（如同“无手套”般触摸）就会被扭曲和污染。这种来自科学界的警示，加上错误的定罪案例，共同强化了几十年的“常识”：法律体系不应信任证人记忆，因为它过于不可靠。

<details>
<summary>Original English</summary>

Imagine for a moment
that you're absolutely certain
about the person you saw commit a crime.
You're so confident,
you'd be willing to testify about it
under oath in a court of law.
Your memory is strong,
crystal clear, absolutely unshakable.
But now imagine that that same memory,
though it feels 100 percent true,
is actually false
and could send an innocent
person to prison,
maybe even to death row.
This is the complex and sometimes
heartbreaking world of eyewitness memory.
But for decades, we've been telling
ourselves a story about eyewitness memory
that itself may not be entirely true.
Most of you have probably
heard cautionary tales
about how wildly unreliable
eyewitness testimony can be.
You may have heard about famous cases
like the case of Ronald Cotton,
where Jennifer Thompson, a rape victim,
misidentified him as her attacker.
As she would later recall her testimony
from his criminal trial,
"I was absolutely, positively,
without a doubt certain
that he was the man who raped me
when I got on that witness stand.
And nobody was going
to tell me any different."
The jury understandably
found her testimony convincing.
Cotton was found guilty
and sentenced to life in prison.
But Ronald Cotton
did not rape Jennifer Thompson.
Instead, it was a known rapist
prowling her neighborhood that night.
Cotton spent almost 11 years in prison
before DNA testing
finally proved his innocence
and revealed the identity
of the true rapist,
a man named Bobby Poole.
Jennifer Thompson's testimony was sincere,
but her memory was wrong.
DNA exoneration cases just like this one,
involving confident misidentifications,
have happened literally hundreds of times,
leading many to seriously question
the reliability of eyewitness memory.
But wrongful convictions like
these are not the only reason
why most people think
eyewitness memory is unreliable.
For years, scientific research
has also painted a damning picture
of human memory.
Starting in the 1970s,
scientists like Elizabeth Loftus
began to show how shockingly easy it is
to manipulate memory.
In groundbreaking studies,
she and others implanted
false memories in adults
of having been lost
in a shopping mall as a child,
or as having been attacked
by a vicious animal,
even though these things
never actually happened.
Findings like these seem
to confirm our worst fears.
Memories are not like video recordings.
They're more like evidence
from a crime scene,
collected by people without gloves,
distorting and contaminating it
with every touch.
This message from science
reinforced the message
from the wrongful convictions,
and the conventional wisdom
was set in stone for decades.
The legal system should not
trust eyewitness memory.
It's just too unreliable.

</details>

### 记忆的评估：如何检测而非固化

故事在此迎来了意想不到的转折。如果问题的关键不在于证人记忆本身有多么不可靠，或者说，多么容易受到污染而产生虚假记忆，而更多在于我们“何时”以及“如何”去测试证人的记忆呢？将这一点与法证证据（如 DNA 或指纹）进行类比。我们都知道法证证据可能被污染，从而误导案件，正如被污染的记忆一样。但我们并不会因此就全盘否定法证证据。相反，我们会尽早收集，在它被污染之前。为何如此？因为可靠的信息来自于分析未受污染的证据，而非受污染的证据。

同样的原则也适用于记忆证据：尽早收集，在它被污染之前。目击证人对警方怀疑的对象是否是其亲眼所见之人的记忆，可以非常可靠，但前提是其记忆未被污染。科学家们现已达成共识，即使是第一次的测试，也会污染证人对特定嫌疑人的记忆。例如，如果嫌疑人是无辜的，这是目击证人第一次看到他的脸，而此时证人正积极地回忆案发情景。即使证人当时说“不，那不是他”，这个无辜嫌疑人的脸也已与其在证人记忆中产生了关联。这就是污染。你无法阻止这种情况的发生，也无法让证人的记忆恢复到污染前的状态。因此，我们应关注警方调查早期进行的第一次未受污染的记忆测试，而不是一年、两年甚至三年后在刑事审判中进行的、已被彻底污染的最后一次测试。遗憾的是，法院往往采取相反的做法，信任审判时的最后一次记忆测试，而几乎忽略了至关重要的第一次测试。这是一个严重被低估的问题。

<details>
<summary>Original English</summary>

But here's where the story
takes an unexpected turn.
What if the problem is not so much
about how unreliable eyewitness memory is,
because of how easily
contamination can create false memories,
and more about how and when
we test the witness's memory.
Think about forensic evidence
like DNA or fingerprints.
Everybody knows that forensic
evidence can be contaminated
and end up implicating an innocent person,
much like contaminated memory can.
But we don't just dismiss
forensic evidence for that reason.
Instead, we collect it as early
as possible in the police investigation
before it's contaminated.
Why do we do that?
It's because reliable information comes
from analyzing uncontaminated evidence,
not contaminated evidence.
And the exact same principle
applies to memory evidence.
Collect it early,
before it's contaminated.

</details>

### 科学新发现：初期自信辨认的高度可靠性

基于上述对测试时机的重视，让我们深入审视警方如何进行这一至关重要的、检测证人未受污染记忆的第一次测试。在案发后的几天或几周内，警方可能会找到一名嫌疑人，并希望将其展示给目击证人，以确认是否找对了人。直接将嫌疑人的照片交给证人并询问“这是他干的吗？”是不可取的，因为这会暗示证人警方认为谁是嫌疑人。为以更少暗示性的方式测试记忆，警方常会向证人展示一套包含六张照片的“六人组”照片卡。其中一张是嫌疑人的照片，其余则是警方确知无辜的、长相相似者的照片。这样，警方可以在不暴露其真实怀疑对象的情况下，向证人展示嫌疑人照片。这是一种更公平的测试记忆的方式，而当遵循其他推荐做法时，其公平性会进一步提升。

例如，告知证人，案发现场犯下罪行的罪犯可能在这些照片中，也可能不在。更重要的是，负责向证人展示照片的警官不应知道谁是嫌疑人，以避免无意中影响证人的选择。当这些步骤都被遵循时，这便成为对证人记忆的一次纯粹测试，而测试结果也变得异常有趣。大约十年前，我的实验室在有影响力的科学期刊上发表的研究首次报告：从初步照片卡中自信地辨认出嫌疑人，是高度可靠的。这让很多人大吃一惊——它并非不可靠。对于一个花费数十年时间警告法律体系证人记忆不可靠的科学领域来说，这些发现并非易于欣然接受。但几乎所有近期科学研究都表明，一次初始的、自信的辨认，其可靠性远超该领域之前的认知。当然，它并非绝对不会出错，但其可靠性绝对不容忽视。

<details>
<summary>Original English</summary>

An eyewitness’s memory
of whether or not the police suspect
is the person who they saw
commit the crime
can be highly reliable,
but only if the witness's
memory is uncontaminated,
not after it's been contaminated.
And scientists now agree
that even the first test contaminates
the witness's memory for a given suspect.
If the suspect’s innocent, for example,
it's the first time
the witness is seeing his face,
and this is happening at a time
when the witness is actively
thinking about the crime.
So even if the witness says,
"No, that's not the guy who did it,"
this is how the innocent suspect’s face
first becomes associated with the crime
in the witness's memory.
That's contamination.
You can’t keep that from happening,
and you can’t put the witness’s
memory back the way it was.
So focus on the first
uncontaminated memory test
early in the police investigation,
not the last, thoroughly contaminated test
that happens at the criminal trial one,
two, or even three years later.
Unfortunately, courts tend
to do the reverse,
placing their faith in the last test
of memory at trial
while all but ignoring
the critical first test.
This is a seriously
underappreciated problem.
Well, with that in mind,
let's take a closer look
at how the police conduct
that all-important first test
of a witness’s uncontaminated memory.
In the days or weeks after a crime,
the police might find a suspect,
a person who they think
may have committed the crime,
and they'd like to show him to the witness
to see if they have the right guy.
They could just hand the suspect's
photo to the witness and ask,
is this the guy who did it?
The problem is, that would be suggestive,
because it would reveal
to the witness who the police think
may have committed the crime.
To test memory in a less suggestive way,
the police will often show the witness
a whole set of six photos.
It's called a six-pack photo lineup.
One photo is of the suspect,
and the others are
of similar-looking individuals
who the police know are innocent.
That way, they can still show
the suspect's photo to the witness,
but without revealing
who they think committed the crime.
It's a much fairer way to test memory,
and it becomes fairer still when other
recommended practices are followed.
Such as letting the witness know
that the perpetrator
who they saw commit the crime
may or may not be among these photos.
And the officer who's administering
these photos to the witness
should not even know who the suspect is
to avoid unintentionally influencing
the witness’s choice.
When it's done this way, it becomes
a pure test of the witness's memory,
and this is where things start
to get interesting.
About 10 years ago, work from my lab,
published in strong,
high-impact scientific journals,
first reported that a confident
identification of a suspect
from an initial photo line-up
is highly reliable.
Big surprise.
Not unreliable.
For a scientific field
that has spent decades
cautioning the legal system
about the unreliability
of eyewitness memory,
these findings were not easy
to enthusiastically embrace.
But almost all of the recent
science finds that an initial,
confident identification
is much more reliable
than the field previously thought.
Not infallible, of course,
but certainly not unreliable.

</details>

### 重新审视冤假错案：首个测试是关键

这些新的科学发现引出了一个有趣的问题：关于我之前提到的那些 DNA 翻案的案件，也就是我们知道在最后一次记忆测试中，受害者在刑事审判时自信地错误指认了无辜者，从而导致了错误的定罪，那么，这些证人在第一次记忆测试时，对同一人的反应是什么呢？深入研究会发现，他们通常在那时并没有自信地错误指认无辜的嫌疑人。问题在于，当时没有人认真对待他们的初步反馈。

还记得 Jennifer Thompson 吗？案发几天后，她在进行初步的照片卡辨认时很挣扎，在两个照片之间犹豫了几分钟，最终才锁定 Ronald Cotton 的脸，并说“我想是他”。这显然是一次不确定的辨认，充满了怀疑和犹豫。到了刑事审判时，经过大量记忆污染后，她的疑虑消失了，她变得绝对、肯定、毫无疑问地确信 Ronald Cotton 就是强奸她的男人。如果当时的人们已知晓我们现在所知的：关注第一次测试，而在本案中，你发现的是一次完全不确定的辨认，那么 Ronald Cotton 很可能一开始就不会被错误定罪。这就是问题的核心所在。

<details>
<summary>Original English</summary>

These new scientific findings
raise an interesting question
about the DNA exoneration cases
that I told you about earlier.
The ones where we know
that on the last test of their memory,
at the criminal trial,
witnesses confidently misidentified
an innocent person
contributing to a wrongful conviction.
The question is:
What did those witnesses do
the very first time their memory
was tested for that same person?
And when you look into that,
you find they usually
did not confidently misidentify
the innocent suspect at that time.
The problem is, nobody listened to them.
Remember Jennifer Thompson?
She struggled with the initial
photo lineup a few days after the crime.
Narrowing it down to two pictures,
wavering hesitantly between them
for literally minutes
before finally landing
on Ronald Cotton's face and saying,
"I think it's him."
It was obviously
inconclusive identification,
full of doubt and indecision.
By the time of his criminal trial,
after much memory contamination,
that her doubts were gone
and she became absolutely, positively,
without a doubt certain
that Ronald Cotton
was the man who raped her.
If they had known then what we know now,
focus on the first test,
where, in this case, you find
a completely inconclusive identification,
it seems likely that Ronald Cotton
never would have been wrongfully convicted
in the first place.
And that's the point.

</details>

### 证人拒绝辨认：无辜的有力证据

这里，故事又迎来了一个有趣的转折。在第一次使用照片卡进行测试时，证人常常甚至没有初步辨认出嫌疑人，不像 Jennifer Thompson 那样初步辨认了 Ronald Cotton。相反，在证人对罪犯的记忆仍如同最初一样新鲜、强烈且未受污染时，他们会看着照片（包括嫌疑人的照片），然后说：“这几个人都跟我的记忆不符。”换句话说，证人拒绝了照片卡，这提供了嫌疑人是无辜的清晰证据。

这正是关键所在：对目击证人未受污染记忆的第一次测试，可以提供指向任一方向（罪或无罪）的可靠证据，具体取决于测试结果。当证人拒绝照片卡时，它便指向嫌疑人无罪的方向。然而，许多在初次照片卡辨认中拒绝的证人，在一年或两年后的刑事审判中，可能已不再意识到自己的记忆已被污染，他们会自信地指认出之前拒绝过的同一个人。有时他们甚至不记得自己曾拒绝过，因为那已经是久远的事情了。这些被告常常因此被定罪、判处长期监禁，并锒铛入狱。与 Ronald Cotton 不同，他们可能没有 DNA 证据来证明清白。但他们拥有的，是记忆科学界传递的一个新信息，可以帮助他们证明清白——讽刺的是，这是一个彻底的观念转变：这是首次测试时，来自目击证人记忆的，指向无辜的可靠证据。

再看 **Miguel Solorio** 的案例。他于1998年因谋杀被捕。但在首次测试时，四名证人都拒绝了他的照片卡。当时没有人对此给予关注。一年多后，其中两名证人出现在他的刑事审判中，在法官和陪审团面前指认他为枪击凶手。他被定罪并判处无期徒刑，不得假释。Miguel 从19岁到44岁，共在狱中度过了25年，直到2023年底，在新科学的帮助下才最终获得平反——即关注第一次未受污染的测试。

尽管 Miguel 的故事令人悲伤，但至少新的科学成果帮助在他耗尽一生之前推翻了他的定罪。然而，许多有类似经历的人至今仍身陷囹圄，有些人甚至生命垂危。

再考虑 **Charles Don Flores** 的案例。他也于1998年因谋杀被捕。此案中有一位证人。案发当天，她告诉警方，她看到一名白人男性、留着齐肩发的嫌疑人，在案发前不久进入了邻居家。几天后，她回到警察局。警方对她进行了催眠（这总是很微妙），但她坚持自己的说法：一名留着齐肩发的白人男性。警方要求她制作了一张合成图像，她照做了，图像显示的是一名留着齐肩发的白人男性，与她所说的一致。

请记住，这是警方调查的非常早期阶段。此时，她对罪犯的记忆是最为新鲜、强烈且未受污染的。而罪犯的长相也符合这一描述。但警方怀疑的是 Charles Don Flores，一位留着短发的西班牙裔男子。出于某种莫名原因，警方将他的照片放入照片卡中，并展示给了证人。这仍然是警方调查的早期阶段。这是对证人对嫌疑人 Charles Don Flores 的未受污染记忆进行的、极其重要的第一次测试。证人查看了这张照片卡，没有看到任何留着齐肩发的白人男性，因此完全可以理解地拒绝了它，提供了 Flores 无辜的明确证据。

快进到一年后的刑事审判，你大概可以猜到这个故事的结局。此时，证人已超过100%确定，案发当天进她邻居家的是 Charles Don Flores。陪审团被说服，他被判有罪并判处死刑。他在死囚牢房中度过了25年以上，至今仍在那里。在这些年里，没有任何陪审团听到过证人在第一次测试其未受污染记忆时的行为，更不用说那强调关注第一次测试重要性的新科学共识了，因为那里才有可靠的信息。

<details>
<summary>Original English</summary>

Yet many witnesses who reject
an initial photo lineup,
after their memory becomes contaminated,
will show up at a criminal trial
a year or two later,
unaware that their memory is contaminated,
and now confidently identify
the very same person
they initially rejected.
Half the time, they don't
even remember doing that,
it was so long ago.
These defendants are often convicted,
sentenced to long prison terms,
and are now behind bars.
And unlike Ronald Cotton,
they do not have any DNA evidence
to prove their innocence.
But what they do have is a new message
from the world of memory science
that can help to do that.
And ironically,
in a complete mind flip,
it's reliable evidence of innocence
from the memory of an eyewitness
the first time they were tested.
Consider the case of Miguel Solorio.
He was arrested in 1998 for murder.
But on the first test,
four witnesses rejected his photo lineup.
Nobody paid any attention to that.
More than a year later,
two of those same witnesses
showed up at his criminal trial
and identified him as the shooter
in front of a judge and jury.
He was convicted
and sentenced to life in prison
without the possibility of parole.
Miguel spent 25 years of his life,
from age 19 to age 44,
behind bars,
before finally being
exonerated in late 2023,
with the help of the new science
that I'm telling you about today:
focus on the first uncontaminated test.
As tragic as Miguel's story is,
at least the new science helped
to overturn his conviction
before he spent his entire life in prison.
But others with a similar story
remain behind bars as we speak.
And some of them,
their lives are hanging in the balance.
Consider the case of Charles Don Flores.
He, too, was arrested for murder in 1998.
There was one witness in this case.
On the day of the crime,
she told the police that she saw
a white male with shoulder-length hair
go into her neighbor's house
shortly before the murder occurred.
A couple days later, she goes back
to the police station.
They hypnotize her.
It's always dicey,
but she stuck to her story.
She said it's a white male
with shoulder-length hair.
The police asked her to make
a composite sketch, so she did.
Came out to be a white male
with shoulder-length hair,
just like she said.
Now keep in mind, this is very early
in the police investigation.
This is a time when her memory
of the perpetrator is as fresh and strong
and uncontaminated as it will ever be.
And this is what
the perpetrator looks like.
But the police suspected
Charles Don Flores,
a Hispanic man with short hair.
And for some inexplicable reason,
put his photo in a photo lineup
and showed it to the witness.
This is still early
in the police investigation.
This is the all-important first test
of a witness's uncontaminated memory
for the suspect, Charles Don Flores.
The witness looked at this photo lineup,
didn't see any white males
with shoulder-length hair,
and quite understandably rejected it,
providing clear evidence
that Flores is innocent.
Fast forward to his criminal
trial a year later,
and you can probably guess
how this story turns out.
Now the witness is more than 100 percent
certain that it was Charles Don Flores
she saw go into her neighbor's house
on the day of that murder.
The jury was convinced,
he was found guilty
and sentenced to death.
He's been on death row
for more than 25 years,
where he remains to this day.
And in all those years,
no jury even heard what the witness did
on the first test
of her uncontaminated memory,
much less about the new
scientific consensus,
emphasizing the importance
of focusing on that first test,
because that's where
the reliable information is.

</details>

### 改革之路：科学指导下的司法公正

因此，这不仅仅是学术上的探讨。它关乎为被错误定罪者（其中一些人身处死囚牢）伸张正义，也关乎从源头上预防错误的发生。这关乎理解记忆并非一个需要被怀疑的敌人，而是一个复杂的工具，一旦得到恰当运用，便能服务于正义的宏大事业。

未来需要进行的改革已经相当清晰：第一，在警方调查和法律诉讼中，都应优先考虑对证人未受污染记忆的第一次测试，因为那里蕴藏着可靠的信息。第二，同时，应降低对证人自信的嫌疑人身份识别的重视程度，特别是当证人之前曾拒绝过同一人脸时——这很可能反映的是记忆污染，而非事实真相。第三，为了推动这些改革，需要向法律从业者——包括警察局长、辩护律师、检察官和法官——普及目击证人记忆新科学的知识。

在过去几年里，我和我的同事们一直在这样做，积极接触辩护律师和检察官，将这种新的理解带入法律体系。我们发现，起初，辩护律师们对这个新信息感到震惊，因为他们不习惯于认为证人记忆有时是可以可靠的。对他们而言，听取目击证人的陈述，即使是其未受污染记忆的第一次测试，也可能导致又一次的错误定罪。而检察官们则出于相反的原因感到警惕。对他们来说，不听取一位证人在拒绝同一张脸后，又做出自信辨认的证词，听起来像是让罪犯逍遥法外的“通行证”。因此，我们设法将双方（辩护方和控方）聚集起来，对他们来说，我们传递的新信息听起来都很危险。但这并非危险，它只是说明了记忆的工作方式。

为了更好地服务于正义，双方以及我们所有人都必须遵循科学，倾听记忆在最可靠之时所诉说的真相。如果我们能找到方法做到这一点，世界必将更加公正。谢谢大家。（掌声）

<details>
<summary>Original English</summary>

So, you see, this isn't
just an academic exercise.
It's about exonerating
the wrongfully convicted,
some of whom are on death row,
and about preventing wrongful convictions
from happening in the first place.
It's about understanding that memory
is not an enemy to be distrusted,
but a complex tool that,
when properly used,
can serve the cause of justice.
The reforms that need to be made
going forward seem pretty clear.
Number one, in both police
investigations and legal proceedings,
prioritize the first test of the witness's
uncontaminated memory,
because that's where
the reliable information is.
Number two, at the same time,
de-emphasize confident
suspect identifications
made by a witness who earlier
rejected that same face
because that likely reflects
memory contamination, not the truth.
And number three,
to get these reforms underway,
educate legal professionals,
police chiefs,
defense attorneys, prosecutors, judges
about the new science
of eyewitness memory.
Over the last couple of years,
my colleagues and I
have been doing just that,
reaching out to defense attorneys
and prosecutors alike
to bring this new understanding
to the legal system.
We find that initially, at least,
defense attorneys are alarmed
by this new message
because they're uncomfortable
with the idea that eyewitness memory
can ever be reliable.
To them, listening to what
an eyewitness has to say,
even on the very first test
of their uncontaminated memory,
is just a recipe
for another wrongful conviction.
And prosecutors,
they're equally alarmed,
but for the opposite reason.
To them, not listening to a witness
who makes a confident
suspect identification
after previously rejecting that same face
sounds like a recipe
for letting the guilty walk free.
So we've managed to bring both sides
of the aisle together in a way,
to both the defense and the prosecution
our new message sounds dangerous.
But it's not dangerous.
It's just how memory works.
To better serve the cause of justice,
both sides and all of us have to follow
the science by listening to memory
when it most reliably speaks the truth.
It will become a more just world
if we can find a way to do that.
Thank you.
(Applause)

</details>