Foval Learning Institute
Courses / Memory / Lesson 5

What a confident memory is worth

85 min

Listen: this lesson as a conversation

Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.

In this lesson you will learn to
  • State what the traditional position on eyewitness confidence was, and why it was reasonable
  • State the five pristine conditions and what follows from them, in the synthesis's own terms
  • Explain why "initial" is the load-bearing word in the whole claim

This is the best-evidenced thing in the course and the hardest lesson to write, because the honest answer has a condition attached to every part of it.

What everybody knows

A confident witness is no more likely to be right than an unconfident one. That is the position this lesson starts from, and the synthesis describes it as one the legal system increasingly accepts.1 This course has not read a textbook or a judicial direction stating it, so where it is written down is something the lesson does not claim.

The synthesis this lesson is built on opens by agreeing that it was reasonable. In the authors' own words: "The U.S. legal system increasingly accepts the idea that the confidence expressed by an eyewitness who identified a suspect from a lineup provides little information as to the accuracy of that identification. There was a time when this pessimistic assessment was entirely reasonable because of the questionable eyewitness-identification procedures that police commonly employed."1

Read that as a model of how to disagree with somebody. Not "everyone was wrong". The position was correct about the evidence that produced it, and the evidence changed.

What changed

Wixted and Wells published the synthesis in 2017.1 This course has read its Summary, its Box 3, the section introducing the five recommendations, and one field-study passage. The fifty pages of evidence review were not opened, so what follows is the paper's own account of its evidence rather than this course's weighing of it.

Their question, in their words: after thirty years of research, "the time seems ripe to ask how eyewitness confidence informs accuracy under more pristine testing conditions (e.g., initial, uncontaminated memory tests using fair lineups, with no lineup administrator influence, and with an immediate confidence statement)."1

And their answer, in their words: "Under those conditions, mock-crime studies and police department field studies have consistently shown that, for adults, (a) confidence and accuracy are strongly related and (b) high-confidence suspect identifications are remarkably accurate."1

With the other half of the same sentence, which must travel with it: "However, when certain non-pristine testing conditions prevail (e.g., when unfair lineups are used), the accuracy of even a high-confidence suspect ID is seriously compromised."1

The five conditions

Box 3 of the paper, verbatim and complete.1

  1. Include only one suspect per lineup.
  2. The suspect should not stand out in the lineup.
  3. Caution that the offender might not be in the lineup.
  4. Use double-blind testing.
  5. Collect a confidence statement at the time of the identification.

What each one is protecting against is worth a sentence, and the five sentences below are this course's reasoning from the conditions rather than the paper's own account of them.2 The paper's explanations, if it gives any, are in the fifty pages this course did not read.

One suspect. A lineup where everybody is a suspect is a test with no wrong answer available. The other people are there so that a witness who is guessing has somewhere to guess wrong.

Not standing out. If the suspect is the only one who matches the description the witness gave, the lineup asks "which of these looks most like your description?" rather than "is your man here?"

The caution. Without it, a witness reasonably assumes the offender is present and picks whoever is closest.

Double-blind. An administrator who knows the answer can signal it without meaning to.

Confidence at the time. Everything that happens afterwards can change how confident somebody feels, and lesson 4 is one mechanism by which.

What follows, in the authors' words: "if the pristine conditions listed here (summarized in Box 3) are followed, then a low-confidence ID implies low accuracy, and a high-confidence ID implies high accuracy."1

Predict first

Before the figures. Under those five conditions, how accurate do you expect a high-confidence identification to be? Write a percentage down. And write down what you expect for a low-confidence one.

Show the answer

Both guesses are usually wrong and in opposite directions.

People guess too low for high confidence. In one police-department field study the synthesis analyses, of 72 suspect identifications made with high confidence at the identification itself, 96 percent were correct, and a signal-detection model fitted to those data "estimated high-confidence suspect-ID accuracy to be approximately 97% correct".1

And too high for low confidence. The same model estimated low-confidence accuracy at "closer to 50% correct".1

That second number is the one that matters more, and it is the one nobody expects. A witness who says they aren't sure is, on this estimate, about as likely to be wrong as right. The traditional position was that confidence tells you nothing. This says it tells you a great deal, and most of what it tells you is on the low-confidence side, where it is a warning.

Note the scope before you carry any of it away. One field study, analysed inside a synthesis this course has read four sections of, on adults, on initial confidence statements. What this course read does not say which of the five conditions that police department met, and the figures are worth little without them. This lesson gives them with that attached every time.

The word the whole thing rests on

"Initial". It is in the paper's own statement of what it set out to ask, and it is the first word of the list of pristine conditions.

The claim is about confidence expressed at the moment of the identification. Not confidence in court. Not confidence after being told the suspect was charged, or after months of being treated as the witness whose identification mattered.

The word is theirs and it is in the sentence stating the question: the pristine conditions are "initial, uncontaminated memory tests using fair lineups, with no lineup administrator influence, and with an immediate confidence statement".1 Two of the three things in that list are about timing, once as "initial" and once as "an immediate confidence statement", which is how much weight the point is carrying.

A lesson that dropped the word would be teaching the opposite of the paper, because the courtroom certainty everybody has seen is exactly the kind the claim doesn't cover.

What this course did not read is the paper's own discussion of the courtroom case, so what it can say is that the claim is scoped to the initial statement, and not what the authors think ought to follow at a trial.2

Check yourself

Somebody reads this and concludes that eyewitness identification is reliable after all, and that the reform campaigns were overblown. What is wrong with that?

Show the answer

Almost everything, and taking it apart is the point of the lesson.

The conclusion is conditional and the conditions are the reforms. The finding is that identifications are accurate under the five procedures, and the five are reforms rather than the status quo. The synthesis is an argument for the reforms, not against them. Which of the five any particular campaign was for is not something this course read, and the point does not need it.

The authors say so themselves, and this is the sentence most likely to be dropped: "Unfortunately, some jurisdictions have not yet made reforms that would create pristine testing conditions and, hence, our conclusions about the reliability of high-confidence identifications cannot yet be applied to those jurisdictions."1

And the other half of their own summary rules out the comfortable reading: when unfair lineups are used, "the accuracy of even a high-confidence suspect ID is seriously compromised."1

There is a second error in the conclusion and it is subtler. The finding isn't that eyewitness identification is reliable. It's that confidence is informative, which cuts both ways: initial high confidence is very accurate and initial low confidence is about a coin toss. A system that took this seriously would convict on some identifications and throw out others, which is a different thing from trusting witnesses more.

So the honest sentence, and it is a mouthful: under five stated procedural conditions, for adults, an initial confidence statement carries a great deal of information about whether an identification is correct, in both directions, and those conditions aren't met everywhere.

Three things people get wrong about this

"Confidence means nothing." The position this synthesis revises, and it was reasonable about the procedures that produced it. It isn't what the evidence under pristine conditions shows.

"This proves eyewitnesses are reliable." It shows confidence is informative under conditions. Low confidence under the same conditions is a warning, and the paper is explicit that its conclusions don't reach jurisdictions without the reforms.

"A confident witness in court is a confident witness." The claim is about initial confidence, and the courtroom is the case it specifically doesn't cover.

Practice

Check a procedure against the five

Take 25 minutes.

Find an account of a real identification procedure. A news report of a trial, a documentary summary, a police force's published policy, a wrongful-conviction case write-up. Anything that describes how somebody was picked out.

Go through Box 3 one condition at a time, and for each write one of three things: met, not met, or the account doesn't say.

Then three lines.

  1. The tally, out of five.
  2. How many were "does not say", which is usually most of them.
  3. What the synthesis would let you conclude about a confident identification made under what you can establish, and what it would not.

"Does not say" is the commonest answer and it is the finding. An account that describes a confident identification without describing the procedure has told you nothing the synthesis can work with.

Write the opposing case

Take 25 minutes. This one is harder than it looks.

Write the strongest case for the traditional position, in about 200 words: that eyewitness confidence tells you little about accuracy. Use the synthesis's own words where you can, because it makes that case itself in its first paragraph.

Then write the strongest case against it, same length, same rule.

Then two lines.

  1. Which was harder, and why.
  2. What single piece of evidence would most change your mind, whichever way you lean.

If either case came out obviously weaker, go back to it. The test isn't which side you end on; it is whether somebody holding the other one would recognise what you wrote for them.

Connections

Back. Lesson 4 was what a question can do to an account, and this lesson is the same subject with real stakes and a much larger evidence base. Lesson 1's reconstructive account is why condition five exists at all: a confidence statement taken later is a different measurement. Focus and Deep Work lesson 1's instrument question is what makes "initial" readable as a scope condition rather than a detail.

Forward. Lesson 6 is the techniques, which is a lighter subject deliberately placed after this one. Lesson 7 comes back to conditional claims and what happens to them in transmission.

Go deeper

Sources

  1. John T. Wixted and Gary L. Wells, "The Relationship Between Eyewitness Confidence and Identification Accuracy: A New Synthesis", Psychological Science in the Public Interest 18(1), 2017, pages 10 to 65. Read in part: the Summary, Box 3, the section introducing the five recommendations, and one field-study passage. The evidence review was not opened, and the body says so. Supports: the quoted opening about the legal system and the reasonableness of the traditional position; the quoted statement of the question with its parenthetical list of pristine conditions; the quoted two-part finding; the quoted qualification about non-pristine conditions; Box 3 verbatim and complete; the quoted statement that low confidence implies low accuracy and high confidence implies high accuracy; the field study's 72 high-confidence identifications at 96 percent correct with model estimates of about 97 and about 50 percent; and the quoted qualification about jurisdictions that haven't reformed.
  2. The one-sentence explanations of what each of the five conditions protects against are this course's own, said as such at the head of the list. None of them comes from the paper. The read level is the Summary, Box 3, the section introducing the five recommendations, and one field-study passage; the paper's own discussion of the conditions is in the evidence review this course did not open, and each explanation is reasoning from what the condition says. The statement that this course cannot say what the paper concludes about courtroom confidence is the same limit stated from the other side.

Check your understanding

This lesson has a 6-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.

Open this lesson in the app