«BY GEORGIOS DIAMANTOPOULOS A THESIS SUBMITTED TO THE UNIVERSITY OF BIRMINGHAM FOR THE DEGREE OF DOCTOR OF PHILOSOPHY DEPARTMENT OF ELECTRONIC, ...»
So, are eye-movements random? From the nine studies that performed statistical analysis on eye-movements and question/task modality, only three report no statistical significance (Cheney et al., 1982; Elich, 1984; Sandhu, 1991) whereas the remaining six reported some statistical significance even though they were unsupportive of the EAC model (Thomason et al., 1980;
28 Farmer et al., 1985; Wertheim et al., 1986; Dooley and Farmer, 1988; Baddeley and Predebon, 1991; Burke et al., 2003); this would suggest that eye-movements are indeed not random.
Further, if the relationship was simple (e.g. baseline movements) the results would be more coherent. Later in this review, research from other fields that suggest that eye-movements are linked to internal processing will be considered.
From the aforementioned data, a trend is visible in some studies where a large sum of the elicited eye-movements has been stares or baseline eye-movements. It is surprising that this trend has not raised any suspicion about the validity of the questioning methodology in the past since the interpretation of both stares and baseline eye-movements is ambiguous. The term stares refers to the central position of the eye within the eye-socket when the eyes are not focused. According to the EAC model this position is associated with visual access. However, stares can also be regarded as failure to activate the transderivational search and it is unclear how to differentiate between that and visual access (also reported by Ehrlichman et al., 1974; Ehrlichman and Weinberger, 1978). This ambiguity is also true of baseline eye-movements; as per the EAC model, baseline eye-movements are associated with auditory eidetic access (auditory constructed sounds or words) but they can also be connected to an internal rehearsal or speech preparation process (Bandler and Grinder, 1979, p. 18). Once again, since there is no available distinction between the two cases, it is theoretically consistent with the EAC model that consistent elicitation of baseline eye-movements can also be regarded as a failure of the elicitation process to activate the “transderivational search”.
Given the statistical variance of the recorded data of one to seven eye-movements, it is surprising that no past research has performed a comprehensive frequency analysis on the number of eyemovements that occur in response to questions or introspection and their temporal location. It is suggested here that if a predictable relationship exists between eye-movements and internal representations and if this relationship is to be discovered, it is necessary to record and analyse all eye-movements that subjects make (also pointed out by Ehrlichman and Weinberger 1978).
Recent developments in eye-tracking technology may allow this to be done reliably and without the immense effort involved in manual rating; depending on the intrusiveness of particular system used, there can be minimal interference with subject-experimenter rapport.
29EYE-MOVEMENT RESEARCH RELEVANT TO THE EAC MODEL
A very important finding for the EAC model by Christman et al. (2003) showed that the retrieval of episodic memories is selectively enhanced when it is preceded by saccadic eye-movements (fast movement of the eyes towards or away from an object, or without a visual stimulus) and not when preceded by pursuit eye-movements (eye-movements used to smoothly follow a moving object). Non-visual eye-movements such as those referred to by the EAC model fall under the category of saccadic eye-movements and thus an important link between eye-movements and memory retrieval is hereby established.
In a different area of eye-movement research and after a series of experiments (Brandt and Stark, 1997; Demarais and Cohen, 1998; Spivey and Geng, 2001), Richardson and Spivey (2000) adopted a “Hollywood Squares” paradigm where subjects were presented with a two-by-two grid of squares, each filled with an object and associated with an auditorily-presented semantic property. Consistent with earlier accounts, they found that when the objects were removed from the grid and the subjects were questioned about one of the properties, subjects tended to look at the blank region of space where the property had been previously presented. This spatial indexing effect is related to the eye fixation and not attentional focus, is independent to fixations on separate locations in absolute space and even though spatial location is irrelevant to the task, it is consistently and automatically encoded.
The spatial indexing phenomenon agrees with neurological research in imagery and perception where it has been shown that there is a large overlap of brain area (re-)activation between perception and imagery (Kosslyn, 2005; Buckner and Wheeler, 2001; Handy et al., 2004). Not only is there similar brain activation but the eye-movements are re-enacted and they play a functional role (Laeng and Teodorescu, 2002).
The NLP creators developed the EAC model based purely on their own observations of people’s behaviour and it may be tempting to attribute these observations on spatial indexing but on closer inspection, such a conclusion would be erroneous. It would mean that every time a representation is accessed, our eyes move in the same direction as they did when this representation was encoded. In all probability, an illogical conclusion if we consider that most information in our lives is presented to us within a central attention window of a limited viewing range while the eye-movements associated with the retrieval of internal representations are 30 relatively spatially extreme and most probably outside the limits of this window. Also, the study conducted was concerned with very short-term recall – what about longer-term accesses? Of course, neither argument could make a strong stance without collecting further evidence.
EYE-MOVEMENTS IN DYADIC INTERACTIONSThe contexts where the EAC model is supposed to hold true are unclear; no explicit claims have been made by Bandler and Grinder (1979) and the original context is that of dyadic “natural” human interaction. It is possible that this is the only context where it holds true and only when a certain condition is met: rapport. The working assumption so far has been that the relationship between the researcher and the subject is not important and several studies have not reproduced this context (Thomason et al., 1980; Gumm et al., 1982; Burke et al., 2003). Further, some studies (Cheney et al., 1982; Elich et al., 1985; Baddeley and Predebon, 1991) required the subjects to interact with a light switch to enable blind rating of the eye-movements, which may have influenced the results by creating an artificial environment, thus jeopardising the rapport condition.
There are several important findings regarding eye-movements during dyadic interactions that
are relevant to the EAC model:
If a person is looking upwards and sideways and there is no apparent object to which their gaze is directed, 4-year-old children can infer the person is thinking (Baron-Cohen and Cross, 1992).
There is some evidence that the relationship between the experimenter and subject affects the rate of the eye-movements (increased rate of eye-movements with highanxiety questions, MacDonald and Hiscock, 1985). No change in direction was observed but the experiment controlled only for change in lateral direction.
No conclusion may be made about whether the position of the experimenter (face-to-face versus behind subject) affects eye-movements (Kinsbourne, 1972; Ehrlichman and Weinberger, 1978) but it is certain that they occur even when no other person is present (Ehrlichman and Barrett, 1983; Kocel et al., 1972).
According to McCarthy et al. (2006) eye-movements in dyadic interactions are also culturally-biased. In a complex-question task, all three groups of Trinidadians, Canadians
It was also recently explicitly shown that verbal questions will elicit eye-movements when there is nothing to look at (Ehrlichman et al., 2007); an interference theory that was originally put forward has been refuted. The remaining theory is that people naturally shift eyes rather than focus them and they suppress those eye-movements when there are useful visual cues in the environment such as the face of another person (Ehrlichman, 1981). In this view, eyemovements could be regarded as an integral part of thought and brain activity.
Another question that several studies have attempted to answer before is whether eyemovements are reliably consistent over time but the results are mixed and hence inconclusive.
Templer et al. (1972) found them not to be reliable, in contrast to Bakan and Strayer (1973).
Dorn et al. (1983) tested the same subjects with the same questions after a week and found that the eye-movements were different. However, during debriefing some subjects reported that they had recalled their previous response to each question instead of generating a new one.
Returning to the issue of question complexity examined earlier, several studies have made a distinction between questions that elicit eye-movements and questions that do not; reflective versus factual questions (Day, 1964; Duke, 1968), reflective versus over-learned (Ehrlichman et al., 1974) and complex processing versus over-learned, immediately available and syntactically simple (Ehrlichman and Weinberger, 1978).
DISCUSSION In this review, the deficiencies of past EAC research have been identified and requirements that should inform future EAC research have been established.
The pattern that has emerged from reviewing past EAC model research is that working assumptions have not been identified fully. Granted, this is a difficult task for such a complex investigation and it is a privilege to be able to learn from past research. A fine example of this is 32 the use of the term representational system. In the original NLP texts (Grinder and Bandler, 1976) this term was defined very loosely and it was not until recently that definitions more suitable for academic research have appeared (Dilts and DeLozier, 2000). Linguistically speaking, people’s ability to answer questions about the world such as “what colour is the sky?” presupposes that they are able to access those representations. Even though this presupposition may be sufficient for empirical and experimental research, it is customary to clearly state what definition of the term is assumed; this has been lacking and the definition has been taken for granted.
A new set of working assumptions is proposed (see Table 2) that with further research should lead to a more rigorous experimental methodology.
To begin, it is important to establish criteria for the eye-movement elicitation methodology. An attempt was made by Baddeley and Predebon (1991) to import the criteria from lateral eyemovement research but no discussion of its applicability or relevance was made. Certainly, a formalised classification of question complexity is required (e.g. reflective versus over-learned) or at least the responses have to be controlled; either by adopting a question model that continually refines the requested detail or with a consistency test of the eye-movement responses. Similarly, Ehrlichman and Weinberger (1978) suggested a test of which questions consistently elicit left or right eye-movements, and which do not. Alternatively, an established means of exploring the phenomenology of the subject’s experience can be used (e.g. Varela and Shear, 1999).
It is also vital to record all eye-movements present because there is no way to predict the number of eye-movements without extensive question analysis. In order for the measurements to be reliable, human rating of the eye-movements should be avoided and a recording method whose accuracy and reliability is known ought to be used. Further, given the possibility that nonvisual eye-movements are coupled during conversation (like visual eye-movements are, see Richardson and Dale, 2005), ideally both the eye-movements of the subject and the experimenter need to be recorded.
Last, but not least, subjects with a similar cultural background and same native tongue need to be selected to eliminate cultural bias and because internal translation processes would interfere with the results, respectively.
Thirty years after its introduction, NLP remains a largely unexplored field within academia and the EAC model was the main target of evaluation in the past. Even though the results of these evaluation efforts have not been consistent and thus conclusions can only be tentative, they have been used as evidence to discredit the model itself and NLP as a whole.
Perhaps the EAC model has been seen as a simplistic part of NLP but the inherent complexity of the EAC model and its study should be evident from the critique of past EAC model research that has taken place in this chapter. Past research has implicitly adopted incomplete and erroneous assumptions and therefore the EAC model requires further research attention.
More importantly, if the EAC model is considered a simplistic part of NLP and yet no definite conclusions can be drawn from this relatively large set of research, this is a clear indicator of how much more complex it would be to investigate larger and more complex NLP techniques or models. If academic value is to be extracted from NLP, more weight would need to be given by future research.
Though research in the EAC model has stopped for several years, it may now be a good time to continue these efforts, as the link between eye-movements and neurology is clearer and eyetracking technology suitable for this purpose is going to presented in this thesis. As discussed, saccadic eye-movements were recently shown to aid the retrieval of episodic memories and even though (at least without further investigation) spatial indexing cannot account for the EAC model it is a positive indicator that eye-movements are related to the process of encoding information.
Along with research in dyadic interactions that is now available, these advances justify the investigation of an ad-hoc model such as the EAC model which is still “current” in NLP circles.