A lie involves trying to mislead someone about something the speaker believes is false. An inaccurate statement can also come from misunderstanding, uncertain information or memory error. A technically true answer may still be misleading, and withholding information can raise another question about honesty. Begin by identifying what the account actually claims. Keep three questions separate. Is the information wrong? Did the person know it was wrong? Were they trying to mislead? Evidence that answers the first question may leave the others unresolved. Our cue is: check the account, not the performance. Charles Bond and Bella DePaulo combined studies of unaided judgments about truth and lies. Average accuracy was about fifty-four percent, with truths recognised more accurately than lies. That is an average across research tasks, not a probability you can assign to the person in front of you. Aldert Vrij, Maria Hartwig and Pär Granhag reviewed research on nonverbal deception cues. Such cues were faint and unreliable for practical lie detection. Eye contact, pauses, fidgeting, accent or confidence cannot settle accuracy. Someone telling the truth can be nervous, and someone misleading you can appear calm. Changing an account may be a correction. An unchanging account may be accurate, rehearsed or shared with other people. Neither pattern answers the question by itself. Possible reasons for dishonesty include avoiding consequences, protecting privacy or gaining an advantage, but inventing a convincing motive adds no independent evidence that it happened. Elizabeth Loftus reviewed experiments on misinformation after an event. Later suggestions can affect what people report remembering. That finding does not make a particular memory false, and it does not justify dismissing an account of harm. Ask open questions before suggesting an answer, and record a revision as a revision. Here is a fictional purchase. A seller says a parcel was posted on Tuesday. Tracking says the carrier received it on Thursday. That discrepancy matters. Clarifying what the seller meant by posted, and checking the delivery agreement, can establish more than studying how confidently they repeat the claim. Use independent, relevant information you have legitimate access to. Two people repeating the same source are not two independent confirmations. Do not manufacture traps, provoke reactions or treat an artificial intelligence system's confident answer as evidence. Suspicion does not give you permission to snoop. Sometimes the practical decision does not need a complete verdict about honesty. You could require written approval before spending money or confirm delivery before paying the balance. Those arrangements address a risk without relying entirely on your ability to infer another person's mind. Uncertainty does not oblige you to accept an arrangement you cannot reasonably trust. If questioning someone could expose you to retaliation, seek a suitable source of support instead. You can make decisions about access, money or participation without provoking the person to collect more evidence. Think of one disputed claim in an ordinary situation. What exactly could you verify, and what would remain unknown about the speaker's knowledge or intention? Choose one legitimate check or safeguard. Pause here, and pause the player for longer if helpful. [Pause: 9 seconds. Pause the player for longer if helpful.] After checking, record only what the result establishes. A corrected date might resolve a planning problem without settling why the original date was wrong. The written entry has research links and further practice. Let evidence support the conclusion it can carry, rather than asking demeanour to supply the rest.