An eye accessing cues exercise is a calibration drill, not a chart you memorise. You ask one person a series of questions, watch where their eyes move in the half second before they answer, write down what you saw, and then ask them what they actually did inside their head. After twenty or thirty questions you have a map of that one person. You do not have a rule about everybody, and anyone who tells you otherwise is selling something.
| Question | The short version |
|---|---|
| What is the exercise? | You question a partner, log the first eye movement after each question, then check your reading against what they report. |
| How long does it take? | 15 to 20 minutes in pairs, three rounds of roughly ten questions. |
| What exactly do you watch? | The first movement after the question lands, before the words start. |
| What do you write down? | Question type, direction of the movement, and what the person says they did internally. |
| Does the standard chart hold? | Sometimes for one person across a sitting. Not reliably across people, and the research does not support it as a general rule. |
| What is it not? | Lie detection, and not a read on anyone who has not agreed to be watched. |
What are eye accessing cues?
Eye accessing cues are the small, fast eye movements people make while they search for information. The claim behind them is that the direction corresponds to the kind of search: pictures in one direction, sounds in another, body feelings in a third. Richard Bandler and John Grinder described the pattern in Frogs into Princes in 1979, after watching people answer questions and noticing that the eyes moved before the mouth did.
The part that survives scrutiny is the pause itself. People do look away to think, and the movement is involuntary and early, which makes it a usable signal. The part that does not survive is the fixed map: the assumption that up and to your left means a remembered picture for every human being you will ever meet. That is the assumption this exercise is built to test rather than to confirm.
Does the standard eye accessing cues chart hold up?
As a universal chart, no. The most cited test is a 2012 paper in PLoS ONE by Richard Wiseman and colleagues, “The Eyes Don’t Have It: Lie Detection and Neuro-Linguistic Programming”, which found no relationship between the predicted eye directions and whether people were lying or telling the truth. Several smaller studies on the visual, auditory and kinaesthetic version of the map came out equally flat.
That result deserves to be stated plainly rather than argued away. What it rules out is the generic chart and the lie-detection use. What it does not rule out is calibration: checking, on this person, in this conversation, whether their own movements repeat in a stable way. That is an individual pattern, and it is the same logic behind any other calibration exercise in NLP, where you build a baseline before you interpret anything.
So treat the chart the way a doctor treats a reference range. It tells you where to look first. It does not tell you what you are looking at.
How do you run the eye accessing cues exercise?
Work in pairs, sitting at a slight angle rather than face on, with a notebook you can write in without looking down for long. One person asks, one person answers, and you swap after each round.
- Agree the frame. Say out loud that you are going to watch eye movements and that your partner can stop at any point. Being observed without consent is where this skill goes bad.
- Round one, ten questions, no interpretation. Ask, watch, write the direction down. Do not guess what it means yet.
- Debrief round one. For each question, ask your partner what they did internally: saw a picture, heard a voice, felt something, talked it through silently. Write their answer next to your observation.
- Round two, ten new questions, now predicting. Before they speak, say your reading quietly to yourself, then check it in the debrief. Count hits and misses.
- Round three, mixed and faster. Shuffle question types so you cannot anticipate, and keep counting.
- Score it. Compare your hit rate in round two with round three, and compare both against the chart’s prediction for the same questions.
Ten questions per round is enough. Thirty in a row and your partner starts performing for you, which quietly contaminates everything after question fifteen.
Which questions should you ask?
Use questions that force a specific kind of internal search and that have no emotional weight. Your partner should have to look something up, not confess something.
| Search you are provoking | Example question | What the chart predicts |
|---|---|---|
| Remembered picture | What colour is your front door? | Up, to the subject’s left |
| Constructed picture | What would your front door look like painted bright green? | Up, to the subject’s right |
| Remembered sound | How does the ringtone on your phone start? | Sideways, to the subject’s left |
| Constructed sound | How would your name sound sung very slowly? | Sideways, to the subject’s right |
| Internal dialogue | What do you say to yourself before a long drive? | Down, to the subject’s left |
| Body feeling | What is it like to step into cold water? | Down, to the subject’s right |
Left and right in that table are from your partner’s point of view, which is the opposite of what you see across the table. Half the mistakes in this exercise are nothing but that mirror problem, so write “their left” in your notes rather than an arrow.
What should you log, and how do you read it?
Four columns are enough: question type, the direction you saw, what your partner reported, and whether that matched. Keep it on paper so the scoring is visible at the end instead of remembered generously.
- Repeating within the person is the only finding that means anything. If three remembered pictures all sent their eyes the same way, you have something to work with for the rest of that conversation.
- No pattern at all is a legitimate result. Some people barely move their eyes, some close them, some stare straight ahead and search without any visible cue.
- A pattern that contradicts the chart is common and is not a mistake in your observation. Write it down as it is and use that instead of the chart.
- Movements that follow the words rather than preceding them are usually gesture or emphasis, not search. Those belong in a general sensory acuity log, not in this one.
What goes wrong in this exercise?
Mostly five things, and all five are fixable inside one sitting.
- Asking a question your partner can answer without searching, so there is nothing to watch.
- Watching too late and catching the gesture instead of the first movement.
- Flipping their left and your left.
- Letting your partner explain while you are still writing, which collapses observation and interpretation into one blur.
- Sliding into lie detection, which is the one use the evidence clearly rejects.
The hardest part to learn from a page is the timing, because the window is roughly half a second and you only notice you missed it after it has gone. That is why we run it live with an observer on each pair in the trainings I teach in Lisbon, in groups of no more than ten, so somebody else can say “you looked down at your notes right then”. If you want that kind of feedback rather than a chart to memorise, the seven-day practitioner training in Lisbon is where this gets drilled properly.
Carolin Mallmann, Licensed Trainer of The Society of NLP, Lisbon
FAQ
What is an eye accessing cues exercise?
It is a paired drill in which one person asks questions, the other answers, and the observer logs the first eye movement after each question before checking it against what the answerer reports doing internally. It takes about 15 to 20 minutes and produces a map of one person, not a general rule.
Do eye accessing cues actually work?
Not as a universal chart. A 2012 PLoS ONE study by Richard Wiseman and colleagues found no link between the predicted eye directions and lying. What can hold is a pattern inside one person during one conversation, which is why the exercise is framed as calibration rather than as reading a code.
Can you use eye accessing cues to tell if someone is lying?
No. That use has been tested directly and failed. Treating an eye movement as evidence of deception is both unreliable and unfair to the person in front of you, and no responsible NLP training teaches it that way.
How many questions do you need per round?
About ten. Fewer gives you too little to compare, more and your partner starts anticipating the question types and performing the answers. Three rounds of ten, with a debrief after each, is the version that works in a training room.
What if my partner’s eyes do not move at all?
That is a valid result and it happens regularly. Some people search with their eyes closed, some hold a fixed gaze, and some move so slightly that the cue is not readable. Record it as no visible cue and calibrate that person on other signals such as breathing, pauses or voice tempo.
Is it ethical to watch eye movements in everyday conversation?
Observing is normal, drawing private conclusions and acting on them is not. In practice this means you run the exercise with consent, you keep your readings as hypotheses, and you check them by asking rather than by assuming you already know what someone is thinking.
Keep exploring
- Calibration Exercises: 6 Ways to Train Your Eye for People, for the broader drills that make eye movements only one signal among several.
- NLP Coaching vs Therapy: Where the Line Is, for where observation skills belong in coaching work and where they do not.
Carolin Mallmann, Licensed Trainer of The Society of NLP, trained by Richard Bandler. She has trained 300+ people and runs English-language NLP trainings in Lisbon, in groups of max. 10.
