Neurolinguistic Programming

Abstract

Neurolinguistic programming (NLP), which MeSH classifies under learning, is a set of communication and psychotherapy techniques developed in the 1970s by Richard Bandler and John Grinder. Its central proposal is that people have a preferred representational system, visual, auditory, or kinesthetic, readable from their verbal predicates and eye movements, so that matching a person's system builds rapport and produces change. These claims are testable, and they have been tested repeatedly since the early 1980s. The empirical record is consistently negative, and NLP is therefore widely classified as a pseudoscience: reviews and controlled studies find no reliable evidence for a detectable preferred system, no benefit from matching it, and no support for its broader efficacy claims. This article describes what NLP claims, how those claims were tested, and why the scientific consensus rejects them.

Keywords: neurolinguistic programming, preferred representational system, pseudoscience, predicate matching, eye-accessing cues

Neurolinguistic programming is a model of interpersonal communication and a collection of therapeutic techniques built on the premise that subjective experience is coded in sensory-specific representational systems and that skilled attention to a person's language can detect, and then influence, how they think (Bandler & Grinder, 1975). It is included here not as an established part of cognitive psychology but as an instructive case: a framework that borrowed the vocabulary of the field, made empirically checkable predictions, and was found wanting when those predictions were examined. Its history is a study in how a scientific-sounding theory is evaluated, and in what separates a testable claim that fails from one that was never testable at all.

What Neurolinguistic Programming Is

Neurolinguistic programming originated in the mid-1970s at the University of California, Santa Cruz, where Richard Bandler, then a student, and John Grinder, a linguist, set out to identify what made certain psychotherapists unusually effective. They studied the gestalt therapist Fritz Perls, the family therapist Virginia Satir, and the hypnotherapist Milton Erickson, and proposed that the therapists' skill lay in identifiable, transferable patterns of language, which they set out to formalize (Bandler & Grinder, 1975). The name encodes the theory's ambition: neuro for the nervous system through which experience is processed, linguistic for the language that both reflects and shapes it, and programming for the idea that habitual patterns can be deliberately altered.

The framework rests on a foundational assumption that experience is represented in sensory modalities, and that individuals come to rely preferentially on one of them, chiefly the visual, auditory, or kinesthetic. From this assumption NLP derives its signature techniques: reading a person's preferred modality from the sensory verbs they use and the direction of their gaze, then matching one's own language to it to establish rapport and, ultimately, to guide change. It is a large and loosely bounded body of practice rather than a single theory, marketed for psychotherapy, management, sales, and self-improvement, and it has never been organized into the kind of falsifiable propositions a mature science requires.

Core Claims and Techniques

The most clearly stated and most testable of NLP's proposals is the preferred representational system (PRS). A person is said to favour one sensory modality when thinking, and to reveal it in the predicates they choose: a visual person is said to speak of seeing what is meant, an auditory one of something sounding right, a kinesthetic one of grasping an idea. A practitioner who identifies the PRS and answers in the same modality is claimed to build rapport more effectively. This is a genuine empirical claim, with two separable parts: that a stable, detectable preference exists, and that matching it changes an outcome.

A second and widely reproduced claim concerns eye-accessing cues: the direction a person's eyes move while thinking is said to indicate which representational system they are using, with upward movements signalling visual processing, lateral movements auditory, and downward movements kinesthetic or internal dialogue. The pattern is presented in NLP training as a near-universal map, allowing an observer to infer another person's mode of thought from gaze alone. This too has been tested directly. Across three studies coding the eye movements of people telling truths and lies, including footage from real high-stakes press conferences, the movements showed no correspondence to the NLP pattern (Wiseman et al., 2012). A later eye-tracking study found that gaze is not random and does shift with the type of question a person is answering, but that the directions observed are not the ones NLP predicts (Marconi et al., 2023).

Beyond these, NLP includes the Meta-Model, a set of questions intended to recover the fuller deep structure of meaning behind a speaker's abbreviated surface utterance by challenging its deletions, distortions, and generalizations, a framing borrowed loosely from the transformational grammar of the period; and anchoring, the deliberate pairing of an internal state with a cue, such as a touch or a word, so that the cue can later re-evoke the state, a straightforward relabelling of associative conditioning. These techniques are harder to test than the PRS because they are described flexibly, but it was the PRS, the framework's most concrete prediction, that drew the first and most decisive empirical scrutiny.

The Evidence Base

The preferred representational system was tested directly and early, because it is exactly the kind of claim experimental psychology is built to examine. If people have a stable PRS, independent judges should agree on a person's system, different indicators should converge, and the system should predict something. Christopher Sharpley reviewed the research on predicate matching and the preferred representational system and found the construct unsupported: the evidence did not show that a reliable preferred system could be identified, nor that matching predicates improved counselling outcomes (Sharpley, 1984).

Proponents replied that the studies had tested NLP incompetently, using untrained raters and artificial settings, so that the negative results reflected flawed method rather than a false theory (Einspruch & Forman, 1985). This is a legitimate objection in principle, but it cuts both ways: a theory that can only be tested by its own certified practitioners, and that attributes every failure to insufficient skill, has begun to insulate itself from disconfirmation. Reviewing the expanded literature, Sharpley concluded that the additional data remained non-supportive, and raised the deeper worry that NLP might be framed so as to be untestable rather than merely unproven (Sharpley, 1987). That distinction, between a hypothesis that fails a fair test and one built to evade any test, is the hinge on which the scientific assessment of NLP turns.

Scientific Assessment

The verdict of the wider scientific community has been uniformly unfavourable. A committee of the US National Research Council, convened to evaluate techniques proposed to enhance human performance for the US Army, examined NLP among them and found no scientific evidence supporting its claims or its influence-strategy model (Druckman & Swets, 1988). Two decades later, surveying the whole research base, Witkowski catalogued the empirical studies in an NLP research database and found the large majority non-supportive, concluding that NLP functioned as pseudoscientific decoration rather than an evidenced method (Witkowski, 2010).

The critique has also been conceptual. Roderique-Davies characterized NLP as cargo cult psychology, borrowing Feynman's image: it adopts the outward forms and vocabulary of science, neurology, linguistics, programming, without the substance that would justify them, and the neuro in its name is not warranted by any demonstrated link to neural processes (Roderique-Davies, 2009). Systematic reviews of NLP applied to concrete outcomes reached the same place from the clinical side: a review of NLP interventions for health-related outcomes found the existing trials few, small, and methodologically weak, providing little or no evidence of effectiveness (Sturt et al., 2012). The convergence of a government science panel, a comprehensive database audit, a conceptual critique, and outcome-focused systematic reviews is what underwrites the pseudoscience label: not a single failed study, but the absence of support across every level at which the framework has been examined.

Worked Example

Consider the core empirical question behind the preferred representational system: do two independent ways of assessing a person's PRS agree more often than chance? Suppose an assessor classifies each person into one of the three systems, visual, auditory, or kinesthetic, once from the sensory predicates in a speech sample and once from eye-accessing cues. If the PRS is a real, stable trait, the two methods should converge well above chance. If it is noise, they should agree only as often as two independent guesses.

Take the chance baseline first. If each method assigns the three categories with equal probability, the probability that both independently land on the same category is the sum over categories of the squared category probability: (1/3)² + (1/3)² + (1/3)² = 3 × (1/9) = 1/3, about 33 percent. So even with no real trait at all, the two assessments will coincide roughly a third of the time. A demonstration that the PRS is real therefore has to clear that 33 percent floor by a wide, reliable margin. What the reviews found is that observed agreement between indicators, and between raters, hovered near this chance level and failed to predict outcomes, which is exactly the pattern expected if no stable representational preference is being measured (Sharpley, 1987). The arithmetic makes the point sharp: a construct whose two flagship measures agree no better than coin-of-three guessing is not measuring anything, and no amount of practitioner skill in applying it can rescue a signal that is absent at the level of detecting it.

Current Directions

NLP is no longer an active research programme in mainstream cognitive psychology; the contemporary literature is largely evaluative, asking whether specific applications survive controlled test rather than developing the theory. A systematic review of NLP in organizational settings found the evidence base for its psychological outcomes weak and methodologically limited, with too few rigorous studies to support the confident claims made in training and consultancy markets (Kotera, Sheffield, & Van Gordon, 2019). The same review noted how much NLP practice outruns its evidence, a gap that has persisted for four decades.

Isolated clinical trials continue to appear, typically testing a packaged NLP intervention against usual care for a specific complaint. A recent randomized controlled trial reported that an NLP-based intervention improved self-rated comfort after cesarean birth relative to routine care (Doğan, Alp Dal, & Beydağ, 2025). Such studies are worth noting, but they illustrate the standing interpretive problem rather than resolving it: a brief, attention-rich, expectancy-laden intervention can improve a subjective outcome through well-understood placebo and attention effects without validating any specifically NLP mechanism, and single small trials do not overturn the consistent negative weight of the systematic reviews. The honest current summary is that NLP's specific claims remain unsupported, and that the burden of proof, unmet since the 1980s, still rests with its proponents.

Key Researchers

Richard Bandler (b. 1950). Independent; with John Grinder he co-created NLP in the 1970s, modelling the language patterns of the psychotherapists Fritz Perls, Virginia Satir, and Milton Erickson. Wikipedia

John Grinder (b. 1940). Independent, formerly University of California, Santa Cruz; the linguist who co-created NLP with Bandler and supplied the transformational-grammar framing behind the Meta-Model of language. Wikipedia

Yasuhiro Kotera (b. 1980s). University of Nottingham; he led a systematic review of NLP in organizational settings, finding the psychological-outcome evidence base weak and methodologically limited. ORCID - Wikidata

Gareth Roderique-Davies (b. 1971). University of South Wales; he characterized NLP as cargo cult psychology, arguing that its scientific-sounding vocabulary is not matched by evidence. ORCID - Wikidata

Christopher F. Sharpley (b. 1949). University of New England, Australia; his 1984 and 1987 reviews found no reliable empirical support for NLP's preferred-representational-system and predicate-matching claims. ORCID - Wikidata

Discussion

Neurolinguistic programming is valuable to cognitive psychology chiefly as a worked example of demarcation, the problem of telling science from its imitations. It is not that NLP made no checkable claims; the preferred representational system was concrete enough to test, which is more than can be said for much pseudoscience. It is that when the claim was tested it failed, and the response was not to abandon or revise it but to reframe it in ways that made further testing pointless, attributing every null result to the tester's lack of certification (Einspruch & Forman, 1985; Sharpley, 1987). A theory maintained this way is no longer answerable to evidence, and that unanswerability, as much as any single negative finding, is what places NLP outside science.

The case also illustrates why the appearance of scientific grounding deserves scrutiny in its own right. NLP's name and terminology promise a basis in neuroscience and linguistics that its content does not deliver, and the promise itself does persuasive work, lending authority to techniques sold for therapy, management, and education (Roderique-Davies, 2009; Witkowski, 2010). That some NLP-branded interventions occasionally produce measurable benefit does not rescue the theory, because those benefits are explicable by the general, well-established effects of attention, expectancy, and a structured therapeutic relationship, mechanisms that require none of NLP's specific apparatus. The lasting lesson is that a framework must be judged by its evidence and its openness to refutation, not by the plausibility of its vocabulary.

Glossary

Anchoring (NLP).

The deliberate pairing of an internal state with a cue so that the cue can later re-evoke the state; a relabelling of associative conditioning.

Cargo cult psychology.

A field that reproduces the outward forms and vocabulary of science without the substance that would justify them; applied to NLP by Roderique-Davies.

Deep structure.

In the transformational-grammar framing NLP borrowed, the fuller underlying meaning of an utterance that the Meta-Model's questions aim to recover.

Eye-accessing cues.

The NLP claim that the direction of a person's eye movements while thinking indicates which sensory representational system they are using.

Falsifiability.

The requirement that a scientific claim be stated so that some possible observation could refute it; NLP's central critique is that its claims are often framed to evade this.

Meta-Model.

A set of NLP questions meant to recover the fuller meaning behind a speaker's utterance by challenging its deletions, distortions, and generalizations.

Modelling (NLP).

The NLP practice of observing an expert performer and abstracting transferable patterns of language and behaviour, the method Bandler and Grinder claimed to apply to expert therapists.

Neurolinguistic programming.

A 1970s set of communication and psychotherapy techniques premised on sensory representational systems; widely classified as a pseudoscience.

Predicate matching.

Answering a person in the sensory modality of the verbs they use, claimed by NLP to build rapport; not supported by controlled research.

Preferred representational system.

The NLP claim that a person favours one sensory modality (visual, auditory, or kinesthetic) in thought, detectable from language and gaze.

Pseudoscience.

A body of claims presented as scientific but lacking supporting evidence and, typically, resistant to disconfirmation.

Rapport.

A sense of harmony and responsiveness between people; NLP claims to build it by matching a person's representational system, a claim controlled research does not support.

Representational system.

In NLP, a sensory modality, visual, auditory, or kinesthetic, in which subjective experience is held to be coded.

Surface structure.

In the grammar framing NLP borrowed, the abbreviated spoken form of an utterance, contrasted with its deep structure; the Meta-Model claims to bridge the two.

Testability.

The property of a claim that allows it to be confirmed or refuted by evidence; its absence is central to the critique of NLP.

Frequently Asked Questions

What is neurolinguistic programming?
Neurolinguistic programming is a set of communication and psychotherapy techniques developed in the 1970s by Richard Bandler and John Grinder, based on the idea that experience is coded in sensory representational systems that can be read from language and gaze and then influenced (Bandler & Grinder, 1975).

Is NLP scientifically valid?
No. Its central claims have been tested since the early 1980s and are not supported, and reviews across clinical, organizational, and performance domains find no reliable evidence of effectiveness (Sharpley, 1987; Sturt et al., 2012).

Why is NLP called a pseudoscience?
Because it uses the vocabulary of neuroscience and linguistics without the supporting evidence, and its claims are often framed to resist disconfirmation, so that failures are attributed to the tester rather than the theory (Roderique-Davies, 2009; Witkowski, 2010).

What is a preferred representational system?
It is NLP's claim that a person favours one sensory modality, visual, auditory, or kinesthetic, when thinking, and reveals it in their word choice and eye movements; controlled research has not found such a stable, detectable preference (Sharpley, 1984).

Do eye movements reveal how someone is thinking?
NLP claims the direction of a person's gaze indicates their representational system, but direct tests do not support this map: eye movements during lying did not match the NLP pattern, and an eye-tracking study found gaze shifts with question type but not in the directions NLP predicts (Wiseman et al., 2012; Marconi et al., 2023).

Did an official body ever evaluate NLP?
Yes. A US National Research Council committee examined NLP among techniques proposed to enhance performance and found no scientific evidence supporting its claims or its influence model (Druckman & Swets, 1988).

Does any study show NLP working?
Occasional small trials report benefits on subjective outcomes, such as a recent trial on comfort after cesarean birth, but these are explicable by attention and expectancy effects and do not validate NLP's specific mechanisms against the weight of negative reviews (Doğan, Alp Dal, & Beydağ, 2025; Kotera, Sheffield, & Van Gordon, 2019).

How does NLP relate to hypnosis and conditioning?
NLP drew on Milton Erickson's hypnotherapy and rebranded associative conditioning as anchoring, but adopting the surface of those practices did not confer their evidential standing, and NLP's distinctive additions remain unsupported (Bandler & Grinder, 1975; Einspruch & Forman, 1985).

References