9.   Is the NLP "Eye Accessing Cues" Model Really Valid?

Note:   John Grinder and Carmen Bostic St Clair discuss the question of eye accessing cues experiments in their book Whispering in the Wind, pages 80-81.



Some of the most regularly asked questions about NLP, especially by newcomers, are concerned with the "eye accessing cues" model: Some of the most regularly asked questions about NLP, especially

  • How are they used?
  • What do they tell you?
  • Do they show what someone is thinking about?
  • Do they show whether someone is lying?
  • Is the model really valid?

The brief answers are:

  • To determine someone's "train of thought" on some matter, referred to in the FoNLP as a "strategy"?
  • If you correctly understand the relationship between the two, a person's eye movements will tell you which sensory (or "representational") systems a person is accessing.
  • No, they don't show you what someone is thinking about, only which sensory systems they are accessing.  In NLP-related terms, they show the process, but not the content.
  • Eye accessing cues show whether a person is remembering or constructing a visual, auditory or kinaesthetic "image".  A constructed image is not necessarily false, so the answer is "No, you cannot tell whether someone is lying just by watching their eye movements".
  • The validity of the "eye accessing cues model" depends on how appropriately it is used.  Some people assume that ALL right-handed people's eye movements fit the diagram below, and all left-handed people's movements fit a "mirror" version of that model.  What Bandler and Grinder actually said was:
    You will find people who are organized in odd ways.  But even somebody who is organized in a totally different way
    will be systematic; their eye movements will be systematic for them.
    (Frogs into Princes, 1978/1979. page 27)
  • What looks like relevant research (though I haven't yet had an opportunity to read the article) is summarised online here: Fred H. Previc, Carolyn Declerck and Bert de Brabander (2005), Why your ďhead is in the clouds during thinking: The relationship between cognition and upper space.  Original Research Article in Acta Psychologica, Volume 118, Issues 1-2, January-February, pages 7-24.

Sharpley and Heap - the "Troublesome Twos"

As described below, the "eye accessing cues", as used in NLP-related techniques, are fairly straightforward. What seems to confuse people is the question of how to use them.
Many of the experiments reviewed by Sharpley (first review), Sharpley (second review) and Heap, for example, were effectively useless simply because most if not all of the experimenters were thoroughly confused about when to track eye movements and when to "track and match" a person's use of sensory predicates.  The experimenters (and Sharpley and Heap), seemed to assume that if both techniques were related to the representational systems then they must be, in effect, more or less the same thing.

For this reason I want to start with a short explanation which I hope will clarify the point.

Eye Movements and Predicates

Observing sets of eye movements and "tracking and matching" spoken sensory predicates are separate techniques, used to collect different information, for different purposes.

From an FoNLP* point of view we are using all of our representational systems - visual (what we can see), auditory (what we can hear), kinaesthetic (what we feel, both physically and emotionally), olfactory (what we can smell) and gustatory (what we can taste), all of the time.  However, and this is crucial, whilst using all of our rep' systems (senses) we usually pay more conscious attention to the signals from just one or two of the five systems, depending on the context.

For example, if I am talking to someone, face-to-face, my attention will probably be mainly focused on what I can see and what I can hear.  If whatever I can smell seems entirely natural for the location I'm in I will probably pay little or no conscious attention to what I can smell.  If, however, I suddenly get the impression that I can smell something burning then I am likely to bring my olfactory sense into conscious awareness, whilst still paying conscious attention to my visual and auditory senses, as I try to locate the source of the smell.  If it turns out to be a bonfire in a safe location, then whilst I may go on noticing the smell for a while, if I'm busy with other things it will soon fade out of my conscious attention.

If we relate this to the eye accessing cues and sensory predicates we can get an accurate description of how they work together without duplicating each other:

  • Throughout the event described above, a person's eye accessing cues will show a series of representational system accesses that might change quite rapidly.  This tells us all of the rep' systems the person is accessing - though the person making the eye movements may be consciously aware of only a small number of those accesses.
    In NLP-related terms, what we are watching is a sequence of thoughts which go together to make up a strategy.
  • At the same time that those eye accessing cues are taking place, the person making the movements will be paying more conscious attention to some of his/her thoughts than others.  The sensory system that a person is currently consciously focused on is their "preferred representational system" of the moment.  But because a person's focus of attention can change a number of times within a single "event", in the late 1970s this term was discarded as implying a degree of fixity about the use of the rep' systems that made the term more misleading than useful.

So, eye accessing cues indicate a series of sensory accesses, each one of which may be in or outside of conscious awareness.  The use of sensory predicates - "That's looking good", "That sounds good", etc., indicate which sensory system the speaker is actually aware of at the time of speaking.
And by the way, eye accessing cues should be observed in clusters, not one at a time, whilst sensory predicates should be noted, and responded to, individually.

Beware - Student Researcher at Work!

Anyone who has belonged to an online NLP chat group for any length of time will probably have witnessed at least one flying visit from someone who claims to have evidence that "NLP" or the "NLP eye accessing cues" has been disproved.  And it is certainly true that there appears to be plenty of experimental evidence to that effect.  UNLESS you understand the FoNLP in general and the eye accessing cues in particular - at which point the "evidence" suddenly loses it's value.
To show what I mean, I would like to use the following web page as an illustration: http://www.lieseeker.com/would.htm.

The article you will find there is entitled Would Pinocchio's Eyes Have Revealed His Lies?  A Study Of Eye Movements As Indicators Of Deception, and it describes a research project by someone called Christopher Robert Dillingham II which he wrote up as: "A thesis submitted in partial fulfillment of the requirements for the degree of Master of the Arts in the School of Communication in the College of Arts and Sciences at the University of Central Florida, Orlando, Florida," dated 1998.
In fact it is a classic example of the sort of evidence which is used by critics of NLP.  That is to say, it appears to be useful, relevant information, but in practice it is seriously misleading.
Let me explain.

In the Abstract, Mr. Dillingham claims that:

"This experiment investigated whether or not NLP eye movements are reliable indicators of deception."

And later on, in the section headed Results, Mr Dillingham adds a little more detail when he writes:

"This study investigated whether or not there are significant differences in the direction of eye movements when telling the truth and lying."

And the outcome?  As described in the Abstract:

"This experiment did not obtain significant support for using eye movement as a reliable detector of deception.  Differences in eye movement between both the lie and truth conditions were at levels of chance."

Does that seem pretty conclusive?  Well it might seem to be.  But in practice Mr Dillingham's findings were irrelevant to the FoNLP, because he made two crucial errors:

Firstly, Mr Dillingham apparently never bothered to find out what claims had really been made for the eye accessing model.  Secondly, to give him the benefit of the doubt, he apparently misread one of his key pieces of supporting evidence.  Thus he wrote:

"There have been criticisms of the NLP model.  For one, NLP practitioners have never officially claimed NLP eye accessing cues can detect deception

Yet in the last paragraph of their report Vrij and Lochun (whose paper for the Dillingham is referring to), clearly state that:

"... Detecting lies on the basis of eye movements is completely reprehensible: NLP-theorists have never suggested that this is possible, and research has convincingly shown that it is impossible." (Vrij & Lochun, 1997)

Maybe I'm being naive, but where is the criticism of the NLP model in that sentence?

In my personal opinion it seems that Mr Dillingham has at least a suspicion that his approach is inadequate, but equally clearly sufficiently ill-disposed towards what he thought of as "NLP" to put his suspicions to one side.  Instead he merely notes, in the section headed Limitations:

"During this study, it was observed that there was often a second or even a third eye movement after the interviewer asked the subjects questions.  These eye movement not [sic] analyzed within this study, although an analysis of those second eye movements currently is in progress (Dillingham & Cullars, 1998)."

As we have already seen, the omission Mr Dillingham describes in this last quote actually invalidates his whole experiment, along with any claim that he had sufficient understanding of the eye accessing cues model to design a valid experiment in the first place.
These very basic and crucial flaws notwithstanding, experiments like this - all equally worthless for one reason or another - are regularly quoted as though they are useful evidence regarding alleged shortcomings of NLP.

In this multipart FAQ I will deal with some of the key issues, as follows:

  1. What are the NLP eye accessing cues?
  2. What is the importance of the eye accessing cues to NLP as a whole?
  3. What are the crucial considerations when trying to evaluate the eye accessing cues in laboratory experiments?
  4. The Bottom Line

What are the NLP-related "Eye Accessing Cues"?

In the early days of their experiments with the techniques they had modeled from Perls, Satir and Erickson, Bandler and Grinder noticed that people tended to move their eyes in a fairly consistent fashion (though the consistency tended to be on a person-by-person basis.
From those observations the following "eye accessing cues" chart was drawn up:

Before going deeper into the subject of the eye accessing cues model and it's validity, we need to be very clear about what precisely constitutes NLP.

In a nutshell, what we usually refer to as NLP is actually two things rather than one - and only one of them is "NLP".

According to NLP's co-developer, John Grinder, NLP itself is purely and simply a specific process of modeling various elements of someone's behaviour - usually, but not necessarily, for the purpose of showing other people how to replicate those elements so that they can produce similar results.  That is the whole of NLP.  All the rest is a collection of ideas and models, exercises, methods and techniques designed, adapted and adopted to support the modeling process.

For many people, one of the most attractive features of the genuine NLP-related techniques, is their essentially pragmatic nature.  That is to say, the various techniques are only there because they have worked for someone at some time in the past, and not because someone has theorised that any particular technique ought to work.
By the same token, it is recognised that no technique will work every time, or under all circumstances, or for everyone.

On that basis, firstly it has to be said that this just one of many models in the NLP toolkit.  Even if it turned out that the eye accessing cues model is entirely wrong (and there is no evidence to support that assertion), it really wouldn't affect the validity of any other model or technique, or of NLP itself because they are separate entities.
Secondly, being realistic we should actually expect that a model - including the eye accessing cues - won't always be absolutely accurate.  That's not a fault or a failure.  That's life!

An NLP Eye Accessing Cues Experiment: Uncovering the Myth

This section relates to the article at: http://www.kevinhogan.com/EyeAccess2.htm

It is, perhaps, the supreme irony that, in claiming to have carried out experiments which disprove the main claims for "primary eye accessing systems", as described in Bandler and Grinder's book Frogs into Princes, Mr Hogan has himself produced nothing but a myth.

For example, in the first four sentences Hogan portentously delivers a crushing body blow to his own credibility, thus:

"For years I was completely certain that if people looked up and to the left they were remembering a visual image.  If they were looking up and to their right they were constructing a future memory.  Itís what I was taught and itís what I believed to be true.  The more carefully I watched people in various settings, the more I began to realize that it just didnít seem right!"
(see http://www.kevinhogan.com/NLPeyeaccess.htm, opening lines.)

Just one question - why did it take him so long?  All he had to do was read one of Bandler and Grinder's best known books - Frogs into Princes:

"You ask somebody a question.  They say 'Hm, let's see,' and they look up and to their left, and they tilt their head in the same direction.  When people look up they are making pictures internally.
"Do you believe that?  It's a lie, you know.  Everything we're going to tell you here is a lie.  All generalizations are lies."
(page 18)

And the rest of Hogan's report fares no better.

In his Introduction, Hogan writes:

"This research aims to study what, if any, correlation there is in eye accessing cues to construct and remembered events and primary representational systems (visual, auditory, kinaesthetic).  Though thought patterns are not intended to be consistent over time because of the very nature of behavior, the correlation between the eye movements and primary representational systems when asked questions of sensory nature should be consistent."

In other words, even if the eye movements in the test subjects didn't match up with the "standard" eye accessing cues chart, they should at least be consistent.  Note this last condition.  It points us to one of two fatal flaws in the experimental design.

The first flaw lies hidden behind this seemingly innocuous comment on the first page of the report:

"We deleted appendixes A and B to spare you all the questions we asked subjects but you can go to www.kevinhogan.com for the entire list of questions used in this research.)"

For good measure, the offer is repeated at the end of the article, in the Conclusions section:

"... we [the research team] are willing to provide the questions and structural information for individuals to replicate this study in its entirety ..."

Strangely enough, this last offer doesn't seem to have been implemented very wholeheartedly.

'Strangely', because whilst the report of his "experiment," and the offer of further information, are still posted on his site, Mr Hogan now goes so far as to say, "I don't have access to the original sets of questions any more" (reply to e-mail enquiry in 2002).

'Strangely', because, far from being particularly lengthy, the full set of experimental questions, plus subject's details, take up little more than one side of a standard sheet of photocopier paper.

'Strangely', because although Monica Piechowski, (one of Mr Hogan's colleagues in the design and implementation of the experiment), can readily supply a copy of the questions, she seems unable to lay her hands on the list which shows what rep system each question is testing for.

'Strangely', because although the experiments were carried out in the late 90s, it seems that all of the relevant information is held only in hard copy format - "I've got about six 2" binders stuffed for this project alone" (Monica Piechowski in reply to e-mail enquiry in 2002)

Or maybe not so strange, because both the questions and the structure of the experiment are seriously flawed, rendering the whole experiment useless.

In order to be meaningful, the questions used in an experiment like this (36 questions in the Hogan study), should be constructed so that they unequivocally call for one response and one response only (Visual Recall, Auditory Construct, or whatever).
It is pointless to ask ambiguous questions and then say "Well the subject should have done whatever they do for an auditory construct" when the question might just as readily prompt a kinaesthetic response.
In the case of the Hogan questions the experimenters seem to have believed that they were testing for:

  • Visual Construct
  • Visual Recall
  • Auditory Construct
  • Auditory Recall
  • Kinesthetic Construct [not a standard NLP category*]
  • Kinesthetic Recall [not a standard NLP category*]

(*   It seems to be generally agreed that all kinesthetic experiences are recalled, on the grounds that whilst you can imagine a purple elephant without ever having seen one, you cannot construct a feeling except by actually experiencing it - that is, by recalling it.  On that basis the eyes down left position (as seen by the observer) is labelled simply "kinaesthetic", since there is allegedly no need to specify whether it is "recalled" or "constructed".)

I've run the questions by several people and confirmed my own impression that well over two-thirds of the items could easily be taken as calling for some form of Visual response.
For example, question 3 asks, "What would it feel like to swim in a bathtub of warm noodles?"  I suspect that this is intended to trigger what Hogan would call a Kinesthetic Construct answer, but a very possible first response might be: Visualise a bathtub full of noodles.
Question 7, (another item intended to elicit a Kinesthetic Construct answer?) asks, "What would it feel like to walk through Jell-O barefoot?".  Again some subjects might be inclined to visualise a carpet of Jell-O before they get round to considering the "kinaesthetic" aspect of the question.

On December 2, 2003, Mr Hogan wrote to the "Whispering in the Wind" forum to comment on his experiments.  His post included two sentences which directly relate to what I've just been saying:

"The statistical analysis only showed that looking forward tended to access visual memory.  That was it.".

Just so, Mr Hogan, and given the questions you were asking, that's exactly the result we would have expected you to get!
Far from undermining the eye accessing cues model your experiments seem to have confirmed one small part of it.

Another problem with the Hogan questions is that so many of them require very little accessing activity on the part of the subject in order to return an adequate answer.
As Joseph Riggio explained in an online discussion regarding the testing of eye accessing cues:

The key is to design questions that can only be answered by accessing singular sensory-specific information, and which require a transderivational search in the corresponding sensory system ... The question must be complex enough that the participant must search for the information requested and not have it available in immediate response memory (IRM), e.g. "What is your name" = IRM ...

This qualification clearly undermines the value of number of the questions in the Hogan experiment, including:

10. What color is the carpeting in your car/truck?
11. What is your favorite music group?
14. What color were your mother's eyes?
15. What does cat fur feel like?
18. What noise does a pig make?
20. What is your favorite sound?
28. What color is the door of your refrigerator?
32. What color is the sky today?

According to Monica Piechowski, (in reply to an e-mail query in 2002), the questions for the Hogan study were created by "Kevin [Hogan], Ron [Stubbs], Book Resources on Eye accessing [unspecified], previous research [unspecified], and myself [Ms Piechowski]."  (For "Book Resources" and "previous research" see the References section at the end of the Hogan report?)
After reading the list of questions drawn up for the Hogan experiment, the main impression I'm left with is that none of the inputs to the process of designing the questions including previous researchers knew how to construct a question suitable for testing this model.

In response to another question - about ambiguous questions - Ms Piechowski says:

I think we did throw a few questions out, as to which ones, I can't even remember.  The funny part is, it didn't change the results, I remember that much.

At a guess, it made no difference because the questions that got thrown out were no more 'fit for purpose' than the questions that stayed in.  The same underlying flaw in the design of the questions, (with their strongly visual bias), would also explain the consistency of the results across four separate runnings of the experiment.  The questions were no less ambiguous/biased just because they were asked of different people at different geographical locations.

In Perspective

One reason why Mr Hogan's experiment has received so much attention may be its apparently impeccable academic background.  For example, in his opening comments Mr Hogan writes:

"The specific project .... you are about to read about was completely sponsored by the University of Wisconsin ..."

Which could easily be taken to mean something rather grand and official.  So perhaps we should remember that Mr Hogan's main interests in life are hypnosis and other persuasion skills.
When I put the question to Mr Hogan and to Ms Piechowski (by e-mail) the picture they painted was rather more mundane:

I asked:
"Kevin says/You say "The specific project ... you are about to read about was completely sponsored by the University of Wisconsin". What does that actually mean? Did they make money available or what?"

Monica Piechowski:
"They offered us their classrooms, students, and supplies to use for the research.  Had there been actual expenses, they would have covered a portion of the cost if it was reasonable.  UW-RF[University of Wisconsin, River Falls] was the only source used for this as there were no other expenses than photocopying. ... UW-RF was the overseeing body in charge of making sure that all ethical guidelines were followed and that the research met required criterion."

Kevin Hogan:
"They ran our stat analysis, let us have access to computers necessary for experiments/research and made sure our controls met scientific standards."

The University also provided a "Faculty Sponsor" - Dr. Brad Caskey, now head of the Psychology department at the UW-River Falls campus.  According to Monica Piechowski, however, this simply meant that Dr. Caskey:

"... was the one in charge of making sure that we conducted the research within the APA Research Guidelines and the Institutional Research Review Board.

As far as Mr Hogan and Ms Piechowski are aware (in replies to an e-mail enquiry in January 2003), Dr. Caskey had no training in NLP, and was therefore in no position to evaluate anything other than whether the format of the experiment conformed to APA guidelines and the requirements of the IRRB.

The involvement of the University of Wisconsin in the Hogan experiment, such as it was, seems rather less dramatic, or extensive than a layman might expect from the phrase "completely sponsored by the University of Wisconsin."

One further point worth making here concerns the statistical analysis of the test results.  Mr Hogan writes:

" ... therefore we would like to present to you this report in almost-APA [sic] style for your edification.".

Yet later on, in the Discussion section, Hogan writes:

"With the similarity in results in our study and the overwhelming center results, we decided not to run correlational or other statistical measures simply because the numbers speak for themselves as they currently are."

I really would like to hear the APA's response to such an excuse for not carrying out a proper analysis of the experimental results, especially in light of Mr Hogan's later claim that:

"There is a significant showing that the centered center movement has a relationship with the remembered events ...."

In an experimental situation the word "significant" has a precise, technical meaning which is arrived at and measured by statistical analysis.  That analysis tells the researcher not simply whether the relationship is significant, but also how significant the results are.
In this writer's opinion, to make a claim that the results were "significant" without providing the relevant analysis and its results is amateurish in the extreme.  It is also rather mystifying, is it not, given Mr Hogan's statement that the University's sponsorship of the experiment meant that:

"They ran our stat analysis..."

even though the "correlational or other statistical measures" are something "we decided not to run"?

Never Take Your Subjects For Granted

Perhaps the biggest flaw in many experiments on the eye accessing cues has been the absence of any debriefing of the subjects.  A flaw indeed, since information regarding a similar experiment carried out by Stever Robbins whilst he was an undergraduate at MIT (see http://www.nlp.org/random/police-interrogation.htm) shows how the debriefing sessions threw an entirely different light on the apparently negative test results:

As part of the experiment I asked, "How many chairs do you have in your living room?" expecting [the subject] to access visual information.  [The subject] showed a KINESTHETIC eye accessing cue.  That counted as a non-correlation.  During the debrief, the subject said, "Remember when you asked me about the chairs?  I suddenly remembered how wonderful it felt when my mother rocked me to sleep in those chairs."

As Stever says, "People don't necessarily access the information you think they're accessing."

But hey, this is just an excuse, isn't it?  Just because some student can't get it right doesn't mean that fully qualified, experienced researchers couldn't get it right.  Does it?

Well, it turns out that Stever's problem was a very real one, and nothing to do with his experience or lack thereof, as shown by this comment from a paper that appeared in the Journal of Psycholinguistic Research, Vol. 29, No. 6 (2000):

Studies of syntactic ambiguity resolution using reading paradigms have provided, and continue to provide, valuable information ....  However, they also have some intrinsic limitations.  One limitation is that reading-time measures provide only a general measure of processing difficulty.  That is, they do not provide information about what is being processed, or how it is being processed, but merely indicate whether the processing requires additional time compared to some baseline."
Eye Movements and Lexical Access in Spoken-Language Comprehension, Tanenhaus, M.K., Magnuson, J.S., Dahan, D. and Chambers, C.
Accessible, at the time of writing, at: http://ipsapp008.lwwonline.com/content/getfile/4932/6/2/fulltext.pdf
(Italics as in the original text)

Strategy - What Strategy?

What we've just seen, and frequently find in other apparently unsuccessful attempts to study the eye accessing cues model in the laboratory, is the tendency of ill-informed experimenters to overlook the implications of the attendant "strategy".  For example, one critic of the eye accessing cues model writes:

"Consider the following; we ask a subject to 'Think about the front door to their home and contemplate what colour it is'.  We might think that this would induce a recalled visual image, but how do we know this is the case?  It might be that the candidate actually remembers an accident when they fell over the door step, so while the subject may be telling us that their door is yellow they are actually accessing a kinaesthetic state."

In the first place the writer appears not to understand some very basic NLP terminology.  That is to say, "kineasthetic" is not a "state" at all, it is one of the 5 representational systems.  A "state" is a far more complex entity (see Glossary) which includes elements of all five rep systems.  The idea of an exclusively kinaesthetic state is a complete non-starter.
Not surprisingly, having taken this first wrong turning, the writer seems to be assuming that rep systems are mutually exclusive - that accessing kinaesthetic material (the memory of falling over the door step) occurs in place of an access to visual material (the memory of a yellow front door).  He then goes on to make the erroneous assumption that:

"... we are left with the not unreasonable conclusion that any results arising from this and other studies with similar flaws would be useless as no reliable analysis was achieved."

Well, it's certainly true that a poorly designed experiment is likely to produce poor results.
But that does not mean that we cannot carry out "reliable analysis".
Take the case of a man involved in a demonstration of eye accessing cues on a training course.  He was asked to imagine he was listening to his favourite piece of music - and promptly looked up and to his right - the "visual recall" (Vr) position for many people.

The question seems to unequivocally call for an auditory access, so why the visual eye accessing response?  At a theoretical level, without debriefing, and if we took the first eye movement as the only important one, we would have to conclude that the association between the Ar position and recalling an auditory experience does not hold true for this person.  In real life, the NLP trainer who was carrying out this demonstration pointed out both eye movements - including the subsequent transition to the Ar position - and also asked the subject to describe their strategy on this occasion.
It turned out that the subject had a very carefully selected music collection and a high quality sound system.  For him, "my favourite piece of music" didn't just mean "tune X", it meant "tune X played by the Y symphony orchestra, with Z conducting the performance, played on my sound system."  His accessing strategy to hear the music had involved seeing himself selecting a specific CD from his own collection and putting it on his own CD player in order to be able to hear it at its best.

It seems that the crucial self-limitation in the Hogan experiment, and probably many others, is this assumption that if the model is correct then asking a certain type of sensory-related question "should" get you a single, instant response in the corresponding sensory system.

The truth of the matter is that even a seemingly simple question may trigger a complex strategy.  The subject described by Stever didn't remember ONLY the kinaesthetic experience but rather both the kinaesthetic and the visual experiences, and possibly several more.
In order to make sense of the subject's response we must track the entire strategy, not just a single eye movement.

As Bandler and Grinder pointed out very early on, accessing cues don't function as 'one off' events.  Just as body language signals have to be evaluated in groups, or clusters, eye accessing cues must be evaluated in groups, or strategies.  Some of these accesses might last for several seconds, or even longer, and some might pass as quickly as the blink of an eye (!).
If you ask me what was said at a meeting last week, I may need to start by imagining myself going into the meeting and seeing the various people seated round the table.  I might then remember that the upholstery on my chair clashed with the colour scheme in the rest of the room (and had obviously been borrowed from another office), and that there was what looked like a small tea or coffee stain on one arm.  In fact I could, quite legitimately, go through a whole series of visual accesses in order to answer a question that appears to call for a straightforward auditory access.  Moreover, if my chair was particularly comfortable (or uncomfortable), or the meeting was a bit tense, I might also access some kinaesthetic memories before I get to the information I've been asked for.

And even that is only part of the story.  As the late Dr Steven Heller (see below) pointed out, we can be accessing three rep systems at the same time, at different levels of consciousness (conscious, unconscious and what he called "out of conscious").
Which is why the idea that you can carry out a meaningful experiment by noting just one eye movement per question is simply not on.  Unfortunately, without that all important debriefing, and without a far better understanding of the eye accessing cues model, there was little or no chance that Mr Hogan's team would even realise that there was this basic flaw in their procedure.

The problem is that only way of really telling what the subject is thinking - in the Q & A format most commonly used in laboratory experiments - is to ask them to give a continuous commentary about what they are thinking, as compared to what the experimenter imagines they ought to be thinking.  This is never going to happen, however, as that approach would introduce a constant thread of auditory digital processing into the situation, thus disrupting the whole accessing process.
Not only that, but a key factor here, as in many psychological experiments, is the need to keep the subject from knowing what the experiment is about or, experience shows, they are very likely to start doing what they guess is expected of them.  So, one way you definitely can't find out the strategy behind a given set of cues is by asking about it whilst the experiment is going on!  But guess what - someone actually tried to do exactly that.

More Bad Design

This next example of truly awful experimental design comes from the pages of the Spring 1994 issue of Rapport, the house magazine of the ANLP (UK Association for Neuro-Linguistic Programming).  Unfortunately we are given relatively few details of the experiment, but from what the author/experimenter does tell us it is crystal clear why the experiment produced the results it did produce - and why those results are totally useless.

Firstly we are told that the author:

"... had been concerned for some time that there seemed little in the way of what I considered to be 'good science' which had been conducted on access cues, and my academic training in the scientific method made me 'bound to' test the basis for the assertions about eye movements and predicates."

From what you've read so far you may well have spotted the key errors in this statement:

  1. What does the author mean by "good science", and is it relevant in this context?  NLP is intimately related to psychology, and psychology is not a science.
  2. Note the phrase "my academic training".  The author/experimenter has a PhD in chemistry, not in psychology.  Does he assume that any old academic training is relevant here?
  3. And "the scientific method", is this really appropriate?  In a word, "no".  The so-called "scientific method" (which scientists themselves regularly ignore) is fine where what is under investigation has relatively few variables and can be expected to behave in a consistent manner - as is usually the case in chemistry experiments.  None of this applies in the case of human psychology, however.

Anyway, this author continues:

"The premise I wanted to test was a simple one: if a person's eye movements indicated a particular preferred representation system (V, A or K), then their written and spoken predicates should confirm the same preference for that mode.
(italics as in the original)

To which the simple answer is "No they 'shouldn't'".
Because, as anyone who understands the NLP sensory predicates will know, preferred rep. systems are entirely context specific.  This means, in practice, that far from using a single rep system at all times, most people regularly shift from one rep system to another, and may change rep systems several times in the space of a short conversation.  Which is the whole point of learning and developing calibration skills.
Clearly this experiment was doomed even before it began.  Not through any shortcoming in the model itself, but because even though the experimenter was nominally an "NLPer", he apparently did not understand the model he wanted to test.

The experiment seems to have consisted of two phases, though the article doesn't give a very clear description.  If I've understood correctly, in the first phase each of 100 students was the subject of a video-taped interview which was then reviewed to check the eye movements against the spoken predicates (and we already know why that's unlikely to produce useful results - because it ignores the "strategy" element).
In the second phase of the experiment the subjects were instructed "to 'go inside', and get various experiences, and to note which modalities they were using" (italics as in the original) an approach virtually guaranteed to confuse the whole accessing process.
On the basis of the results gained, the experimenter described what he claimed were two "bombshells":

  1. "... just because a person's eye movements indicated they were 'visuals', for example, it did not mean they tended to speak or write more 'visual' predicates." (italics as in the original)
  2. "We found there was no correlation between the subjects' observable external accessing and the way they actually got the experiences internally!" (italics as in the original)

Just one point needs to be made about the first "bombshell": There is no such thing as "a visual" (or "an auditory" or "a kinaesthetic") in the sense that the label is being used here, and any statement that depends on the assumption that people are limited to unvarying use of a single rep. system is invalid by definition.
The second "result" may sound more impressive, until we understand that the experimenter was not studying the eye movements at the time of the access, : On the contrary:

"The modalities were recorded, and compared with [the students'] accessing cues from the first experiment."

In other words, the second experiment also depends on the assumption that the subjects should, to confirm "NLP theory", only ever use a single rep. system!  Since this assumption is entirely false, the experimental design is fatally flawed, and the results, far from being "bombshells", neither proved nor disproved anything.  They are, in fact, simply worthless and irrelevant.

The Bottom Line

Quite apart from the fact that the eye accessing cues are just one element of NLP, it is arguably almost impossible to devise a foolproof laboratory test of the eye accessing cues model in the way many experimenters approach the task.
But is this really any wonder, given that the very idea of such an experiment turns the whole business of eye accessing cues on its head?

It is important to understand just how eye accessing cues are used, in the context of NLP.
They can be used proactively:

  • To help someone to access a certain type of information, or
  • To help someone to break out of a "stuck" situation caused by the their use of only one rep system in a variety of situations.

Or reactively (see below).

Critics such as the one who mistakes "rep systems" for "states", also assumes that there is some value and importance in trying to (I quote) "induce a particular cognitive state in the subject" by means of asking questions.  In practise, nothing could be further from the truth.  There are a variety of techniques which can be, and are, used to achieve the required effect - pointing in the relevant direction, looking in the relevant direction, or talking the client through from one rep system to another.  You could even use the pedestrian but effective step of asking the person to use the rep system you want them to use - as in asking someone to make pictures when helping them to learning the NLP spelling strategy.  (For an excellent discussion of the practicalities of the eye accessing cues, and rep systems in general, see Monsters and Magical Sticks, Steven Heller, particularly Chapters 6-8 inclusive).

The eye accessing cues can also be used reactively - watching for naturally occurring cues in order to identify which rep system the person is currently working in.  This is of use if we want to track the person's strategy in a given situation, or in order to match that rep. system and thereby create greater rapport.  In this case we would look for naturally occurring evidence first, and then direct the person to access a particular rep system in order to check whether we had calibrated their eye movements accurately.  Which is how the model was arrived at in the first place.

Bandler and Grinder did not propose the model then try to prove its validity, rather they noticed that clients' eye movements tended to follow coherent patterns, and that the patterns correlated to whatever rep system(s) was/were indicated by their use of language.  Thus, a person who referred to not being able to see their way out of some situation was likely to look upwards as they said it.
Taking the hypothesis that there might be an overall relationship between eye movements and verbal cues Bandler and Grinder looked for further evidence, and tested the accuracy of their observations.  Which is how they developed the "standard" eye accessing cues model, including "strategies", based on the most basic trial and error level of experimentation.

In other words, Bandler and Grinder were indeed, in effect, constantly "asking" their clients what they were thinking - not with blunt/disruptive questions, but by the way they interacted with their clients at frequent intervals.  And "frequent" does mean frequent.  Grinder has been quoted as saying that recalibration needs to take place at least every 30 seconds or so during an interaction to ensure that you are accurately tuned in to the other person's current rep system(s).

The commentator who gave us the example of the yellow front door later poses the rhetorical question:

"... what can be done to effectively interpret and analyse eye movements and their associated cognitive state [sic], if they have any relationship at all?  And the answer of course is nothing, no such systems currently exist and certainly no reliable and dependable approach is available to achieve this exquisite level of processing."

As we have seen, this argument has no foundation in fact.  Joseph Riggio has explained what is necessary to construct valid questions, though we have also seen that the experimental design used to elicit specific eye accessing cues is itself next to worthless since it takes no account of how the eye accessing cues are used in real life.
We have also seen that, far from being a pipe dream, there has been a methodology for testing the eye accessing cues for as long as the model has existed.  Which is hardly surprising since it is the methodology was used to build the model in the first place.

So what genuine objections to the model are left?  In short, none.
Of course it may be that when critics demand details of a "reliable and dependable approach" they mean an approach which "works every time" - something which produces unvarying, bog standard results with no exceptions.  In which case they have misunderstood human psychology.  The only kind of psychological experiments which produce more or less standard results are those which are so trite that the results have little or no practical use.
If, on the other hand, he means "works often enough to be useful" then the fact that the model is so widely used, with such frequent success, would seem to suggest that it is as "reliable and dependable" as we can reasonably expect any "model" to be.

It's a Model

I suspect that the misplaced analytical approach experimenters have tried to apply to the eye accessing cues model is due to the fact that, all too often, the experiments are concocted by people with, at best, no more than a passing familiarity with NLP.  Apart from anything else, it is not unusual to find that critics of the eye accessing cues model ignore the fact that it is just that - a "model" - and not, as one particularly vociferous critic has claimed, something that all NLP practitioners regard as an inviolable set of rules carved in stone.

And having started with back-to-front approach, experimenters then build error upon error by ignoring everything except the initial reaction or response - totally overlooking the "strategy" aspect of the thought process going on (literally) behind the eye accessing cues.  Given their lack of understanding of the basic thinking behind NLP, this is hardly surprising, perhaps, but it is not the only reason why the experimental results are of no genuine significance.

What, then, is the bottom line on the experiments which allegedly undermine the value of the eye accessing cues model?

In short, they are all too often based on ignorance, misunderstandings, and flawed experimental design.

It seems that we are only just beginning to realise that human behaviour is far too complex to be tested by something as relatively unsophisticated as the "scientific method". Laboratory experiments are great for tracking the underlying patterns in series of consistent events involving a limited number of controllable variables.  But human psychology is anything but consistent, or "controllable" in that sense, because we so often act according to our feelings rather than our reason.
Contrary to many people's cherished beliefs, we do not, by and large, think then act - we act according to 'gut instinct', and only afterwards do we concoct an ostensibly rational explanation about why we acted as we did.
In other words, I believe that these experiments have an in-built guarantee of failure because their designers ignore the fact that they are dealing with human beings, not machines.

A Better Mousetrap DIY

Finally, you can actually run a meaningful experiment on the eye accessing cues yourself - and from the comfort of your own living room!

An important factor when dealing with the eye accessing cues, is to calibrate a person's eye movements as they naturally occur in order to track which rep system(s) a person is currently using.
This approach is obviously a long, long way from bluntly asking a series of 30 or more disjointed questions in the hope of pushing someone into exhibiting relevant eye movements - even if the questions were "fit for purpose."

In practise, it would make far more sense to film or video the subjects in conversation or being interviewed about some interest they have, and then reviewing the recording to see if what they say ties up with their eye movements.
This should be irrespective of what sensory predicates the person uses.  Not because they aren't important, but because we think far faster than we speak (5 times faster?), so there will quite possibly be several eye movements for any particular spoken predicate.

As a brief example, I was watching a programme a few months ago in which a young woman was describing an assault she had suffered.  At one point in the recording the young woman's eyes went quite quickly through the following positions (from the view's point of view):

Vr   ⇒   Ar   ⇒   K   ⇒   Ad

(That's up and right (visual recall), horizontal right (auditory recall), then down and left (kinaesthetic recall) - this was followed by a brief shake of the head - and finally down right (auditory digital - "self talk").)

What the young woman then said was along the lines of:

"It was very dark and I could hardly make him out, and then he told me to do what he told me and everything would be all right. : I was just too frightened to do anything else."

Where did the Ad, come in?  That was where the young woman decided what she would actually say about whatever it was she had just recalled.

What you see may not be that clear cut every time, but it's a very easy form of experiment/practice, and one you can use yourself whilst watching an interview or chat show on television.

(A viewing tip:  Notice how many politicians now stare straight into the TV camera, no matter what they are saying.  Not so long ago - 5 years or thereabouts - this was far less common.  Has someone been spreading the word about the eye accessing cues at Westminster, on Capitol Hill, etc.?)