The Guilty Knowledge Test: Lykken’s Better Question

David Lykken thought the polygraph was asking the wrong question, and in 1959 he proposed a better one.

Do not ask a suspect whether they committed the crime, because the answer to that question is contaminated by everything in the previous two chapters. Ask instead whether they recognize a detail that only the perpetrator would recognize.

The Guilty Knowledge Test — now generally called the Concealed Information Test — is administered as a series of multiple-choice items. The victim was found in a kitchen. A bedroom. A garage. A bathroom. A cellar. The murder weapon was a knife. A rope. A hammer. A firearm. A brick. Each item includes one correct alternative and several plausible incorrect ones, matched for familiarity, and the examinee is asked to respond identically to all of them, usually by saying no.

What is measured is not the answer. It is the physiological response to each alternative: skin conductance, cardiac deceleration, respiratory suppression, and in laboratory versions the P300 event-related potential.

The theory is Sokolov's orienting response. A stimulus that is personally significant produces an involuntary orienting reaction distinguishable from the response to neutral stimuli. A person who was in the kitchen will react differently to kitchen, and cannot easily prevent it, because recognition is automatic and precedes any decision about what to do.

The logic is fundamentally different from the polygraph's, and better founded in two respects.

First, it does not require any theory about how a lying examinee compares question categories. It requires only that recognition produces a measurable response, which is well established.

Second, and decisively, an innocent examinee cannot fail through arousal. Someone who does not know where the body was found has no basis for responding differently to kitchen than to garage. A frightened, angry, guilt-ridden innocent person will be uniformly aroused across all items, which produces no signal. False positive rates in laboratory work are low and defensible, and the probability of a chance pattern across multiple items can be computed.

Japanese police have used the CIT operationally for decades, running thousands of examinations a year, which makes Japan the only jurisdiction where this has been tested at national scale.

The limitation is operational rather than theoretical, and it is severe. The test requires crime details that are known to the investigators and reliably unknown to the innocent — not published in the press, not disclosed during earlier questioning, not leaked, not visible to bystanders. In a modern investigation with media coverage and multiple interviews, protected details are scarce, and one contaminating disclosure destroys an item.

Which means the better instrument is used least, and the worse one is used most. The difference is not scientific quality. It is that the polygraph will run on any case, and the CIT requires an investigation disciplined enough to keep something back.