Saul Kassin and Katherine Kiechel needed a laboratory paradigm for false confession, and they found one in a task that could not be more trivial.
Students were asked to type letters read aloud by a partner, on a computer they were warned about: do not press the ALT key, it will crash the program and the data will be lost. After sixty seconds the computer crashed — automatically, on its own, nothing to do with the participant — and the experimenter appeared, distressed, and accused the student of hitting the forbidden key.
Every participant was innocent. Nobody had touched it.
Two variables were manipulated. Pace: the letters were dictated either slowly or fast enough that the participant could not be certain what their fingers had done. And false evidence: in half the sessions the confederate partner, when asked, said she had seen the participant hit the key.
The measures were three. Compliance: did the participant sign a written confession. Internalization: did they, when asked afterwards by another student in the corridor, say they had done it. Confabulation: did they produce details of doing it.
Overall, sixty-nine percent signed. Twenty-eight percent internalized. Nine percent confabulated details.
In the worst condition — fast pace plus a false witness — every single participant signed, sixty-five percent internalized the belief, and thirty-five percent generated specific false memories of an event that had never occurred. One reported, in detail, brushing the key with the side of her hand while reaching.
This is a sixty-second interaction with no stakes, no isolation, no fatigue, no threat, and no authority beyond an annoyed graduate student.
Two conditions produced it. Uncertainty about one's own actions, and a confident external account.
The mechanism is source monitoring, the process Marcia Johnson identified by which we tag a mental representation as having come from perception, from imagination, or from someone's report. When there is a strong memory trace, the tagging is easy and the false account is rejected. When the trace is weak or absent — because you were typing fast, or because it was three weeks ago, or because nothing memorable happened — the system substitutes the most authoritative account available. This is Loftus's misinformation effect from chapter 65, running on autobiographical action memory.
Which yields the conclusion this chapter exists to state, and it is the most counterintuitive proposition in Part 4.
Innocence is not protective. It is an aggravating factor. Kassin's later work documents the phenomenology of innocence — innocent suspects waive their rights at far higher rates than guilty ones, precisely because they believe they have nothing to hide and that the truth will protect them. And when an interrogator asserts evidence that cannot exist, the innocent suspect has no explanation for it except that something happened they do not remember.
The guilty suspect knows the evidence is a bluff. Only the innocent one has to take it seriously.
Cross-ref: Central Park Five chapter