What a question does to a memory
75 min
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.
- State what the 1974 experiments did and found, and at what read level this course holds them
- Distinguish a response-bias account from a memory-change account, and say what would separate them
- Trace a famous study to its source, and say what the trace could and could not establish
This is the most famous experiment in the subject, and this lesson is about two things: what it found, and how hard it turned out to be to check.
The claim
A question isn't a neutral instrument. Ask somebody about an event and the words you use supply material, which the assembly in lesson 1 can take up. The 1974 experiments are the demonstration everybody knows.
This course could not open the original paper.1 What it read is a published teaching summary, from an examination board's own resource pages, and every figure below is held at that level and not higher. The second half of this lesson is about what that means.
What the summary reports
Experiment 1. Participants watched films of traffic accidents and were asked "about how fast were the cars going when they ______ each other?", with the blank filled by one of five verbs.1
| The verb | Mean speed estimate |
|---|---|
| Smashed | 40.8 mph |
| Collided | 39.3 mph |
| Bumped | 38.1 mph |
| Hit | 34.0 mph |
| Contacted | 31.8 mph |
Nine miles an hour between the highest and the lowest, for the same films, differing only in one word.
Experiment 2 is the one that matters, and it is where the idea comes from.1 A hundred and fifty participants saw one clip of a multi-car crash. Fifty were asked how fast the cars were going when they smashed into each other. Fifty were asked the same with hit. Fifty were not asked about speed at all.
A week later, all of them were asked a set of questions about the accident, including: "Did you see any broken glass?"
There was no broken glass in the clip.
| Condition | Said yes | Said no, by subtraction |
|---|---|---|
| Smashed | 16 | 34 |
| Hit | 7 | 43 |
| Control | 6 | 44 |
The second column is the reported one. The third is fifty minus it, which assumes nobody refused or said they did not know, and the summary this course read does not say.3
Before the interpretation. A week after a single word was changed, sixteen people in one group said they saw glass that was not there, against seven and six. Write down two different explanations for that, before reading the authors'.
Show the answer
Most people find the first and the second is the one this lesson is for.
The first is that the question changed what they said. People are not certain either way, the word "smashed" implies a violent collision, and when somebody is asked a week later whether there was glass, the frame they were given nudges the guess. Nothing about the memory changed; the answer did.
The second is that the question changed what they had. The word entered the memory, the assembly now includes a more violent collision, and a violent collision is one with broken glass in it. On this reading the representation itself was altered by a word.
Those are the authors' own two interpretations, which the summary reports.1 And notice that they predict the same result in this experiment, which is why the experiment doesn't settle it.
What would separate them is the interesting question, and this course can't answer it from what it read. Something that pulled the two apart would be needed: a measure sensitive to what somebody has rather than what they say, or a design where the two accounts predict different outcomes. This is the shape of a distinguishing test and not a report of one.3
What the study can and can't carry
Take the broken-glass result seriously and look at what it is made of. Sixteen against seven against six. The difference between the smashed condition and the control is ten people, out of a hundred that were asked; between smashed and hit it is nine. Ten is the comparison the lesson rests on, because the control group is the one that was asked no speed question at all.
That is a real difference and it is a very small study, by any modern standard, and it is the foundation of an idea now repeated in every discussion of eyewitness memory.
Three things follow, and the third is the one this course insists on.
One. The phenomenon is much better established than this study. Fifty years of work followed it, and this course has read almost none of it, which means the lesson can teach the phenomenon and must not claim the evidence.
Two. A lesson that leaned on the size of the effect would be overreaching. Ten people is enough to notice something and not enough to say how big it is.
Three. A famous study isn't automatically a strong study. Fame is a fact about citation, and this is as clear a case of the two coming apart as this course has met.
The trace, which is this lesson's real subject
This course tried to read the original and could not. Here is the trail, step by step, because walking it is more useful than the result.
- The paper is Loftus and Palmer, 1974, in the Journal of Verbal Learning and Verbal Behavior, pages 585 to 589. Every citation this course saw gives those details and they agree with each other.
- No copy this course could reach turned out to be the paper. Whether it is open access somewhere this course did not look is not something the trace establishes.
- What was reachable was a teaching summary, published by an examination board as a resource for students, giving both experiments, both tables and the authors' two interpretations.
- So the figures in this lesson are at one remove. They agree with the other summaries this course could reach, and agreement between summaries is not verification: it is what you would see either way, because summaries copy each other.
That is the trace, and it ended at step four. It is one of the five endings Habits and Self-Discipline lesson 7 and Focus and Deep Work lesson 7 both name: the trail leads somewhere you can't get to.
And it is the honest state of this lesson. A reader who wants the figures verified has to open the paper, and if they do, they will know something this course doesn't.
Somebody says this is over-scrupulous. The figures are in twenty textbooks, they all agree, and nobody seriously doubts them. Why does the lesson make a fuss?
Show the answer
It is a fair challenge and there are two answers, one weaker and one stronger.
The weaker answer is that agreement between secondary sources is not evidence. They copy each other, which is exactly how the twenty-one days in Habits and Self-Discipline lesson 4 became a fact, and how the figure in Time Management lesson 5 came to have no printed source at all. Twenty textbooks agreeing is one source, twenty times.
The stronger answer is that the specific things that go missing are predictable, and this course has watched them go missing twice in the two lessons named above. Not the headline figure, which survives. The range, the criterion, the sample and the instrument, which are what let you tell what the figure means. In this case the read level is exactly the thing at risk: it would be very easy to write "Loftus and Palmer found" as though this course had read them, and nobody would ever know.
And there is a cost to the fuss, which the lesson should own. This is a real, well-replicated phenomenon, and a reader who takes away "the broken glass study is dodgy" has learned the wrong thing. The study is fine as far as this course can tell. What is uncertain is this course's access to it, which is a statement about the course.
So the useful version: the effect is real and broadly supported, the figures here come at one remove and are stated as such, and if that distinction seems fussy, it is the same distinction that caught the twenty-one days and the twenty-three minutes in the two courses named above.
Three things people get wrong about this study
"Leading questions implant memories." Too strong for what this experiment shows, which is consistent with two accounts, one of which involves no change to the memory at all. The authors give both.
"It proves eyewitness memory is unreliable." It shows that a question can change an account. What that means for a real identification, under real procedures, is lesson 5, and the answer there is more interesting than this sentence.
"Everyone in the smashed group reported glass." Sixteen of fifty. The commonest misrepresentation of the finding, and it inflates a third into everybody.
Practice
Take 20 minutes, and stop at twenty whether or not you have finished.
Pick a study you have heard about but never read. The marshmallow test, the Stanford prison experiment, the one about posture and confidence, anything with a name.
Follow the trail, writing each step.
- Who says it, and what do they cite? If nothing, write "nothing", and you are finished.
- Find the citation. Can you reach the paper? Not an abstract, not a summary. The paper.
- If you reach it, does it say what you were told? Check one specific figure.
- If you can't reach it, what is the best source you can reach, and what level is that?
Five endings and all of them count: you reach the paper and it agrees; you reach it and it doesn't; the trail ends in nothing; the trail ends somewhere you can't get to; or twenty minutes went.
The fourth is the commonest and it is what happened to this lesson.
Take 15 minutes.
Think of something you witnessed this week that somebody else also saw. A meeting, an incident in the street, a conversation.
Write two questions about it. One as neutrally as you can manage. One with a word in it that supplies a frame, the way "smashed" does.
Then three lines.
- The two questions.
- What each one supplies that the other doesn't: not just tone, but the specific thing a respondent could take up.
- Which one you would have asked without thinking, honestly.
Most people find the leading version is the natural one, which this course expects rather than has measured.3 Questions that supply a frame are easier to ask because they are more specific, and specificity is what makes them leading.
Connections
Back. Lesson 1's reconstructive account is what makes the second interpretation possible at all: a recording can't take up material from a question. Lesson 3 was what retrieval does to a memory; this is what a question does. Habits and Self-Discipline lesson 4 and Time Management lesson 5 are both traces that ended in the same place this one did.
Forward. Lesson 5 is the same subject with the stakes turned up and a much better literature behind it,2 and it is where the pessimistic reading of this lesson gets qualified. Lesson 7 makes the trace a habit.
Go deeper
- Reconstruction of automobile destruction: An example of the interaction between language and memory (Journal of Verbal Learning and Verbal Behavior, 1974). This course did not open it, and the link is to the publisher's record. If you can reach it, you will know something this course doesn't, which is the exercise above with a guaranteed result.
- The Relationship Between Eyewitness Confidence and Identification Accuracy: A New Synthesis (Psychological Science in the Public Interest, 2017), read in part by this course and the subject of lesson 5. Read its Summary after this lesson: it is the same field fifty years later, with a much larger evidence base and a more careful conclusion.
Sources
- Elizabeth F. Loftus and John C. Palmer, "Reconstruction of automobile destruction: An example of the interaction between language and memory", Journal of Verbal Learning and Verbal Behavior 13, 1974, pages 585 to 589. The original was not opened. What this course read is a published teaching summary of it, from an examination board's own resource pages, and the body says so where the figures appear and again in the trace section. At that level, supports: the five verbs and their mean speed estimates; the design of the second experiment with its three groups of fifty; the week's delay; the broken-glass question and the absence of glass in the clip; the sixteen, seven and six; and the authors' two interpretations.
- This lesson introduces no other evidence. The eyewitness literature that followed belongs to lesson 5 and is cited there.
- The third column of the second table is derived rather than reported: the summary gives the number who said yes in each group, and fifty minus that assumes nobody refused or said they did not know. The table and the body both say so. The statement that the two interpretations predict the same result in this experiment, and the description of what would be needed to separate them, are this course's own reasoning from the two accounts as the summary states them, marked inline in the predict block. The expectation that most readers will find the leading question the natural one is an expectation rather than a measurement, and the exercise says so. The four-step trace is this course's own account of what it did, and the conclusion that consistency between secondary sources is not verification is its own reasoning rather than a finding.
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.