The answer, as it currently exists
Here is an answer to the question about a time you had to make a difficult call, given by a quality manager at a small food manufacturer who is invented for this article, since the argument needs one story shown in two states and no real interviewee can supply both.
> In my previous role I identified a labelling discrepancy on a high risk allergen line, escalated it through the quality management system, and made the decision to hold production. It cost us close to a day of output, but it protected the customer and it protected the brand. Afterwards I led the root cause review that changed our changeover procedure. What it taught me is that in food safety you never trade certainty for schedule.
That is a competent answer. It has a situation, an action, a result and a lesson, in that order, and it comes in under forty seconds. It would pass in most rooms.
It is also the fourth time she has delivered it, and almost none of it is what happened.
The day itself
What happened was this. She was covering the afternoon shift for somebody on leave, which is why she was on the floor at all. An operator on the bagging station mentioned, in passing and without much alarm, that the date code on the film did not look right against the paperwork. She went and looked. The code was wrong, and she could not tell from looking whether the error was in the label roll or in the batch record, which are two entirely different problems with two entirely different consequences.
She had about forty minutes before the shift changed and the line would be handed to people who had not seen any of this. Her own manager was not answering. She was not certain she had the authority to stop a line on her own signature, and she was fairly sure that if she was wrong about it she would hear about it for a year.
She stopped the line. She then spent the rest of the afternoon expecting to be told she had overreacted, and she was not told that, because the batch record turned out to be right and the label roll had come off a pallet that had been split with another product. The root cause review was two people and a whiteboard on a Thursday. The procedure change was one extra check by the operator at changeover, which she had to argue for twice because it added ninety seconds.
Every part of that which made the decision difficult is absent from the answer she now gives. The uncertainty about which of the two errors it was, the forty minutes, the unreachable manager, the question about her own authority, the fear of being wrong, the operator who noticed it and mentioned it casually. What is left is a clean line from problem to resolution to lesson, and the clean line is the part an interviewer can do nothing with.
What a retelling is, as distinct from a memory
The loss is not carelessness and it is not embellishment. It is what retelling does.
Marsh's review of this literature draws the distinction the whole article rests on. Laboratory recall asks for detailed and accurate remembering. Conversational retelling depends on the speaker's goals, the audience, and the social context, and because most remembering happens in conversation, retellings of events are frequently incomplete or distorted, with consequences for the speaker's later memory. Two mechanisms carry the effect: selective rehearsal, meaning that what gets said gets strengthened, and the schema that the retelling activates (Marsh, 2007).
The experiment underneath that is worth holding in detail. Participants retold a story three times or not at all, and by instruction the retellings were either entertaining or accurate. Entertaining retellings carried more affect and fewer sensory references. On a later memory test, the people who had retold with an accuracy goal recalled the greatest number of story events, and their accounts were the most accurate, the most detailed and the least exaggerated (Dudukovic et al., 2004).
The parallel to an interview is close but not exact, and it is worth being precise about the gap. Nobody is telling a hiring story to entertain. They are telling it to impress, which is a third goal the study did not test. What the study establishes is narrower and still decisive: a goal other than accuracy, applied to three retellings, measurably degraded what the teller could later recall about the event. An interview answer is told under a goal other than accuracy, most of the time, for more than three tellings.
Then there is the audience. After tuning a message to a particular listener, communicators' own memories for the topic tend to shift toward the view expressed in the message, and three studies traced when that happens. The bias appeared when feedback signalled that the audience had successfully identified the target, and not after failed identification. It appeared for communicators tuning to an in-group audience and not to an out-group audience, and the difference was mediated by how far the communicator trusted the audience's judgment about other people (Echterhoff et al., 2005).
Read that against a job interview and it is close to perverse. The more you regard the person across the table as someone whose judgment is worth something, and the more they nod as though they have understood you, the more your own memory of the event moves toward the version you just gave them.
The fourth telling
That gives the object this article is about, and it needs a name so that you can recognise one of your own.
The fourth telling is the version of a story that has been optimised for delivery and can no longer be used as a memory. The number is not a measurement and it is not meant as one. It is a label for a state, and the state has four marks.
It has a moral, arriving at the end as a general principle rather than as something you learned about a specific week. It has no timestamps, no day of the week, no weather, no detail about what else was going on. Its uncertainty has been removed, so that what was an open question at the time now reads as a problem correctly identified. And the other people in it have become functions rather than people, so that the operator who noticed the date code disappears into the passive voice of a discrepancy being identified.
The state is close to invisible from the inside, which is why nobody catches it in their own material. A story you have told four times comes out smoothly, and smoothness is the only signal you get while you are telling it. It reads from the inside as the mark of a story you know well, when what it indicates is a story you have stopped consulting, and the difference between those two is the whole subject here.
The cost you do not see being charged
Rehearsal feels free. It is not.
Anderson and the Bjorks ran three studies in which participants studied categories, then repeatedly practised retrieving half the members of half of those categories. Recall of the unpractised members of the practised categories was impaired on a delayed test. The impairment survived controls for output interference, which points at a retrieval-based suppression lasting twenty minutes or more, and it fell on the high-frequency members rather than the low-frequency ones. Their conclusion was that the retrieval process itself is implicated in everyday forgetting (Anderson et al., 1994).
Later work carried it into conversation. In a modification of the same paradigm, pairs studied material and one member of each pair recalled part of it aloud while the other listened. The final memory tests showed the expected forgetting in the speaker and also in the listener, across paired associates, across stories, and in the third experiment across free-flowing conversation with no controlled practice at all (Cuc et al., 2007).
Put those beside the quality manager. Each time she gives the polished version to a friend, an old colleague or an interviewer, she is practising retrieval of a subset of that afternoon. The unpractised parts of the same afternoon, which is to say the forty minutes and the unreachable manager and the split pallet, become measurably harder for her to reach. Rehearsal is not additive. It is a selection, and selections cost.
Preparation is a test, not a read-through
This is where the standard advice gets it backwards, and there is a clean experimental reason why.
Roediger and Karpicke had students study prose passages and then either take immediate free-recall tests without feedback or restudy the material the same number of times. When the final test came five minutes later, restudying won. On the delayed tests at two days and one week, prior testing produced substantially greater retention, even though restudying had increased students' confidence in their ability to remember (Roediger & Karpicke, 2006).
Reading your notes on the train, running the answer in the car, saying it out loud to your partner the night before: all of that is restudying. It raises confidence and it is the version of preparation that feels like it is working, which is exactly the trap the study describes. Being asked something about the event you have never been asked before, and having to go back to the afternoon to answer it, is testing.
Putting the day back
So the instrument is not a better answer. It is a set of questions about the event that you have never been asked, answered in writing, the night before.
Take one story you expect to use, and answer these without consulting anything you have written before. What month was it, and what else was happening that month. Where were you standing when you first heard about it, and what were you holding. Who else was there, and what did each of them want out of the situation. What did you not know at the moment you moved, and how long would it have taken to find out. What did you do in the hour afterwards, and who was the first person you told.
Almost none of that belongs in the answer. That is the part people get wrong about this exercise, so it is worth saying plainly: the reload is not material to add to the telling. It is there to make the afternoon reachable again, so that when the interviewer asks a question you did not prepare for, you are retrieving an event rather than searching a script for a line that fits.
Twenty minutes per story, and three or four stories, is the whole of it.
The opposite case is worth a line, because it comes up. Where a story is one you have never told anybody, none of this applies and you should leave it alone. An untold story is already in the condition the reload is trying to restore, and the risk with it runs the other way, which is that you will rehearse it into a fourth telling in the week before the interview and arrive having done the damage in advance.
What this gets wrong, if it does
Interviewers want a structured answer, and there are frameworks for producing one. They do, and the structure is not the problem. A container is fine. The failure is when the container is the only thing that survived, and you can deliver a perfectly structured answer out of reloaded material. What you cannot do is reload from a structure.
Preparation is surely a good thing. It is, and the argument is about which kind. The finding to sit with is that the form of preparation which raises confidence most is not the form that holds up at a delay, and confidence is the only feedback you get while you are doing it (Roediger & Karpicke, 2006).
If I go back to the original I will ramble. The reload happens the night before, not in the room. Being able to reach a detail and choosing to say it are separate acts, and the second one is a matter of judgment you already have.
My best story had a clean moral at the time. A few do. The test is whether you can answer one question about it that you have not been asked before. Where you cannot, the moral is not the point of the story any more. It is all that is left of it.
What she said to the second question
The follow up, when it came, was short. The interviewer wanted to know what made her think the label was wrong rather than the batch record.
The fourth telling has no answer to that, because the fourth telling has already decided that a discrepancy was identified. What she said was that she did not think it, that she had no way to tell which of the two it was, and that what she did in the forty minutes was go and find out whether the pallet had been split, because a split pallet was the only explanation that put the error in the label rather than in the product.
That is the answer she already had, on the afternoon it happened, and had not been able to reach for three years. An interviewer pressing on a story is not testing your honesty and is not admiring your polish. They are leaning on it to find out whether it holds anything, and a story told four times has usually had the load taken out of it. Hers held on the second question, because the night before she had spent twenty minutes putting the day back rather than the speech.