The eye chart is one of the most recognisable images in NLP: a head with arrows showing that looking up and left means remembering an image, up and right means constructing one, and so on around the compass. Somewhere along the way it picked up a darker promise: that you can tell when someone is lying by where their eyes go. That promise has been tested, and it failed. This post explains the model, the failure, and what remains genuinely useful.
The model
The claim, in its standard form: eye movements track internal processing. For a right-handed person, looking up and to the left is said to indicate a remembered image, up and to the right a constructed (imagined) image, level left a remembered sound, level right a constructed sound, down left an internal dialogue, and down right a feeling. Trainers taught these positions as "accessing cues": read the eyes, read the thought. The model came from Bandler and Grinder's early observations of clients, coded the way they coded everything: watch what people actually do.
What failed testing
The model had problems from early on. Through the 1980s, researchers such as Sharpley and colleagues examined the alleged link between eye movements and the modality of thought. The results consistently failed to support it: ask people to recall visual versus auditory memories and their eyes do not reliably go where the chart says they should.
The lie-detection version received a decisive blow in 2012. A research team led by Richard Wiseman published three studies in the journal PLOS ONE under a title that says it all: "The Eyes Don't Have It: Lie Detection and Neuro-Linguistic Programming". In the first study, they filmed people lying and telling the truth and coded their eye movements; the gaze patterns did not match the NLP predictions. In the second, participants who were taught the NLP eye-movement theory did no better at detecting lies than a control group. In the third, they coded the eye movements of people in high-profile press conferences; again, no difference between liars and truth tellers. Three studies, three failures.
The honest summary is simple: eye movements are not a reliable lie detector. Do not use them to judge a suspect, an employee, a spouse or a politician. People who tell you otherwise are either repeating an old claim uncritically or selling something.
Why the claim was so sticky
Because it feels like access: a visible, external sign of a hidden internal process. And because trainers and audiences wanted it to be true, the anecdote machine kept it alive long after the data stopped supporting it. This is the field's overclaiming problem in miniature, and it is the main reason critics treat NLP claims with justified suspicion. The field as a whole is judged by its worst surviving claims, and this is one of them. Our honest limits post walks through the rest.
What eye and attention observation is still good for
Throw out the lie detector, keep the attention. Watching where someone looks, and how, is still useful in ways that have nothing to do with decoding truth:
State and engagement. Gaze direction, blink rate and where attention goes tell you about engagement and state, not about honesty. A person staring at the ceiling during your presentation is not lying; they may be thinking, distracted or computing. Reading a room means noticing these shifts, not decoding them.
Calibration in rapport. Noticing when someone's attention leaves the conversation lets you adjust: change pace, ask a question, reconnect. That is sensory acuity, the skill of observing without interpreting prematurely. It is one of the genuinely trainable NLP abilities, and it survives the death of the eye chart.
Context matters. Gaze is affected by handedness, culture, habit, the layout of the room, and whether a question is about the past, the future or a hypothetical. Even in the best case, eye movement is a weak, noisy signal with many confounds. Observing it as one channel among several, without conclusions attached, is honest practice.
The discipline of naming failures
Here is the thing we actually want you to take from this post: NLP is full of useful tools, and it is also full of claims that failed testing. An approach that names both, in the same syllabus, is one you can trust. An approach that only ever tells you success stories is a sales pitch.
We teach eye accessing cues exactly that way: what the model claimed, what the research showed, and what remains useful after you subtract the fiction. If you want that kind of training, see the Spirit of NLP programme. The free webinar is where you can check us out first.