Article Categories
[ Show ] All [ Hide ]
Clean Language
Article Selections
[ Show ] All [ Hide ]
6. Comments on our methodology

Up to now we have presented the product of our modelling of Robert. Now we take a step back and look at the process by which we arrived at that model. We conclude with a review of Symbolic Modelling as a product modelling methodology.


You can now turn your attention from modelling Robert to modelling our methodology. We haven't produced a formal model of our approach although our notes of How to do a Modelling Project go some way towards this. You can get a sense of what we did in two ways: reviewing the format of this report since it's structure replicates our process; and using our description below to  imagine yourself doing what we did.

First iteration: Modelling from observation
We have studied with Robert and read all of his books so at some level we will have previously created some kind of mental model of him. Therefore while observing him model Martin our aim was, as much as possible, to set aside our preconceived ideas and to discover something new. We did this by alternating between a not-knowing unconscious-uptake state and a more conscious musing-about-what-is-happening state. Both of these states share an intention to not jump to a conclusion. In the first, the intention is just to receive and take on; in the second it is to muse cleanly (see Judith DeLozier's article, Mastery, New Coding and Systemic NLP; and our article, A Model of Musing).

In the breaks and overnight we downloaded our initial impressions and compared our notes. However at this stage most of our model construction was happening out of awareness. By the end of the day we were in a position to describe some of what we saw and to identify a few general patterns in the way Robert does 'Robert modelling'. We consider the list at the end of Section 3 to be a set of general capacities inherent in his approach. Although we detected some of these patterns during the first iteration we also added to and reorganised the list throughout the second and third iterations.

We may have had some intuitions and had identified some patterns, but we didn't have a model yet.

Second iteration: Modelling in the moment
We chose to model the topic of 'selecting what is essential' because it is fundamental to all modelling methodologies. Every modeller has to handle a mass of information, much of which is not directly relevant to their outcome. And while it is possible to construct an effective model out of non-essential bits, it will be neither efficient nor elegant.

Since we had observed Robert's exterior behaviour the previous day, our aim when interviewing him was to invite him to attend more and more to his interior perceptual mind-body space. We wanted to find out what he did on the inside while he was doing what we had observed on the outside. Having heard him answer many questions from the group we also wanted to get to those aspects of his modelling which he had not mentioned.

When we sat down with Robert we didn't have a plan of the questions we wanted to ask. Rather we started with a general aim to identify and locate some of his key symbols, and to find out how they worked together to produce the ability known as 'selecting what is essential'. After that the direction of our questions was guided by the emerging organisation of the information – and our desired outcome.

An important point to note is that we did not 'take on' Robert's process, i.e. we did not put it in our body and in our perceptual space like Robert does when he is modelling. Instead we constructed our model in Robert's perceptual space, keeping the elements (symbols) where they were located from his perspective. That's why we gestured to his body/space, and not ours.

Also we did not do as much backtracking and meta-commenting as Robert did when he was interviewing Martin. In part this is because Robert-as-exemplar was doing a lot of backtracking himself, and in part because we are doing a form of 'internal backtracking'. We visit the symbols in the exemplar's inner metaphor landscape and muse on them so we can ask enough intelligent questions to keep them revealing more of their process. From our point of view this has a similar effect to external backtracking: it keeps the modeller connected and close to the totality of the information being presented; and it has a different effect on the exemplar and on the flow of the interview.

By the end of the interview we still didn't have a complete model, but our notes were good enough to present a first-pass model to the group. This consisted of a single locational diagram covering the key symbols we had jotted down in our notes and a verbal description of the main bits of Robert's process we had identified so far:

Penny Tompkins and James Lawley's first-pass model of: 'Robert Dilts: How I know what's Essential'

Penny Tompkins and James Lawley's first-pass model of:
'Robert Dilts: How I know what's Essential'

Although we had a sense of the iterative nature of Robert's approach, we didn't have a conscious representation of how it all worked together. That only came in the third iteration.

Having taught modelling for decades, Robert is obviously not your average exemplar. Many of his answers contained a huge amount of information neatly packaged. This created somewhat of a dilemma. Often what an exemplar does is so out of their awareness that facilitating them to describe what they do is a painstaking process. Not so with Robert. The challenge for us as modellers was to cope with the concentration of information – only some of which was getting into the model we were constructing around him – and to continue to ask questions which would shine light into the few unlit corners of his mind. When Robert said "the most honest answer is I don't know" and "You just do it", he is saying what all exemplars say when they reach the edge of their meta-cognitive map. The purpose of our questions was to tease out his 'tacit knowledge'.

It is also interesting to note that while interviewing Robert we were faced with what Gregory Bateson might have called a code-congruency dilemma. Not only were we trying to acquire a model of Robert modelling, we were attempting to do it by demonstrating our model of modelling. Fran Burgess has noted that expert modellers like to be modelled with their own methodology. To be congruent with Robert’s approach we would have done lots more backtracking, more preparatory explaining and more overtly taking on his model in our bodies – just as he did. However, our outcome was to provide the group with a different description of modelling. So most of the time we refrained from doing that in order to more clearly demonstrate our approach. One potential downside was that at times it seemed Robert wasn't getting the cues from us that would have let him know we 'got' what he was describing.

Third iteration: Modelling from recordings
The detailed analysis and model construction in the third iteration involved going through the video and transcripts over and over. We distilled what was essential and arranged it into a model. Getting to this stage involved both a quantitative and qualitative analysis.

Our quantitative analysis was very simple and yet gave us a sense of scale. (See our article, Big Fish in A Small Pond.) For example we looked at:
  • The proportion of words spoken by Robert while interviewing Martin (in Clip 01 it is approximately 60:40. This ratio changed at other times during the the interview).
  • The number of questions we inferred Robert asked himself (40 during our interview with him).
  • The number of times Robert used the metaphor "fit" (20 during our interview with him).
You can see that we picked out a few 'extreme' behaviours to quantify. We did this to find out if our intuitions were based on solid evidence. We were surprised to find out that in all three examples our intuitive count underestimated the frequency of occurrence.

A key element in constructing a model involves detecting patterns. Without access to sophisticated computer software this will need to be some form of qualitative analysis. We used a number of mental filters to repeatedly comb through the text of our interview with Robert. For example we highlighted all the metaphors Robert used. We also distinguished between different kinds of information: external and internal behaviour; outcomes and activities; actual examples and abstractions; perceptual perspectives; organisational levels; the relationship between verbal and non-verbal metaphors; etc. In the Appendix, Section 12 we show a summary of our qualitative analysis which provides a stepping stone from the transcript to our final model.

Another question or frame we were always considering was: How is what is being said and done in the moment a fractal of a more general process? More specifically, how could Robert's self-modelling be an example of how he models others?

If you review the words we selected to go in our diagrams you will see we favoured Robert's embodied metaphors (e.g. activating energy, connect to center, marked inside, guided by a feeling, capture the fit, fill in, explore a direction, install in myself, etc.). We did this for a number of reasons: because Robert used so many; because his body was so obviously involved in his modelling; because he said his "somatic mind" was important; but mostly because embodied metaphors are an excellent way for an acquirer to get a sense of how to do someone else's internal process (see our article Embodied Schema: The basis of embodied cognition).

Once we had identified the key metaphors and noted the sequence of mental processes, we enacted them in the same places in our body and perceptual space, and in the same order as Robert – and noticed our responses. When something didn't work – we couldn't go from one behaviour to the next, or there was a clash/inconsistency, or something didn't fit, or there was an unnecessary piece, or we couldn't get out of a loop, or whatever – we would adjust the combination and sequence of metaphors. Our aim was to stay close to Robert's description while searching for the minimal number of elements in the minimal number of steps that would produce the required result – selecting what is essential.

Our method involved lots of trial and feedback. Initially the feedback came from our own reactions as we tried on Robert's processes and from our joint system as we discussed what we were discovering about our prototype model. Once we had settled on a reasonably robust model we asked individual colleagues to try it out and eventually we invited a group to be our guinea pigs. Their feedback was incorporated into our model and thus we passed through another iteration.

What's still to be done?
In reviewing our model we realise the are a number of gaps. There are three ways we could fill in, firm up and refine the model: a second interview with Robert; a detailed analysis of the videos of Robert modelling Martin; or observe Robert modelling another exemplar.

It wasn't until after the interview that we realised fitting parts together is itself a selection process, but at a higher logical level than the original selection of parts. Because we were focussed on 'selecting what is essential' we didn't pay as much attention to the processes of 'fitting' as we could have. Hence our model of Phase II is a little thin. Although we have some ideas about how Robert noticed parts that fit together and how these are subsequently fitted together into an imagined picture, then into an imagined movie, and finally into physical representation of the model, we would like to get more examples of the different kind of fits, and then look for patterns of similarity and difference at each level of fit.

Also, we have little idea what Robert does with anomalies: those parts that have been selected for significance but that don't seem to fit anywhere (assuming this happens). Robert gave a clue when he described the testing of things that activated his radar: "Are they still significant? Do they feel resonant? Sometimes they feel more resonant." From this we guess that things that don't fit lose their (relative) significance. Somehow they are 'un-marked' and disappear from the bag of selected parts, or perhaps they drop out of the picture.

There are plenty of other areas we could investigate, for example how Robert "explores a direction" when he has got a fit. Also we noted that Robert said "If I go out of the center, then there’s all kinds of feelings you can have and you can get lost in feelings." A useful side area to model would be what first lets him know he has gone, or is going out of his center; and having gone out, how does he get back?

The least complete part of our model is Phase IV. There is another whole modelling project to be done to find out how Robert takes his internally constructed movie and turns it into a physical representation, a model, that can be acquired by someone else. How does he do that?

These are interesting questions that will have to wait for another day. As will a forth iteration. We sense it is possible to refine our model into an even more compact form, but for the time being we like it the way it is.

Symbolic Modelling as a product modelling methodology

Symbolic Modelling emerged out of our modelling of one of the most innovative therapists of our time, David Grove. Our original aim was to generalise David's approach so it could be used in contexts in addition to individual therapy. It wasn't until we were well into the project that we had two light-bulb moments: David was continuously modelling his clients – but in a way we had never seen before; and that his process could be coded as a modelling methodology in its own right.

Below we highlight four features of Symbolic Modelling – metaphor, Clean Language, modelling systemically and outcome orientation – which make it suitable as a product modelling methodology, particularly if your outcome is to capture the internal experience of your exemplar.

The role of metaphor
Metaphor is central to Symbolic Modelling. The expanding fields of Embodied Cognition and Cognitive Linguistics are demonstrating that metaphor is fundamental to how we think, feel and act (see George Lakoff on the research into metaphor and embodiment).

Throughout this report we have shown how noticing an exemplar's metaphors helps us to 'get' – both cognitively and somatically – the way they do things. This applies not only to explicit metaphors (like: tool, radar, movie) and analogies (Mozart), but also to the many more implicit metaphors such as those highlighted in the Appendix, Section 12. Metaphor enables us to acquire an embodied sense of the interior perspective and internal activities undertaken by an exemplar.

Noticing metaphors is only the first step. Next we consider the logic inherent in these symbolic expressions. Then we wonder how they work together to automatically produce the exemplar’s behaviour. One way we do this is to muse on the presumed entailments of the metaphors and what that tells us about the nature of the exemplar's way of doing things. For example, we pointed out some of the entailments of Robert's radar metaphor in our comments on Clip 03. When we considered the metaphors of radar, guided, direction and explore together a theme emerged which suggested a systemic process: detection results in a direction which is explored, leading to further detection and so on.

Noticing metaphors, considering their inherent logic, conceiving a model based on these metaphors, and checking it’s relevance happens throughout the modelling process.

Clean Language
The beating heart of Symbolic Modelling is Clean Language. We believe questions based on David Grove’s clean approach are modelling questions par excellence because they:
  • Are short, simple and use the exemplar’s exact words
  • Ask for information about ‘what is’; they don't disagree, deny or negate an exemplar's experience in any way
  • Keep the modeller close to the exemplar’s information – their words, their nonverbals and their perspective
  • Keep the modeller's intrusions into what is being modelled to a minimum
  • Direct the exemplar's attention where it needs to be – their interior world.
In short they get the modeller out of the way and encourage an exemplar to self-model.

'Pure' Clean Language questions were designed as part of David Grove’s therapeutic process and they do not contain any reference to the therapist. However, since product modelling is not a therapeutic process you will notice that while we keep to the spirit of the principle, at times we relax this criteria somewhat.

We have found that being constrained by the discipline of Clean Language is a way to develop your capacity to model. In becoming proficient at Clean Language you learn to: pay exquisite attention to what the exemplar says and does; utilise their exact descriptions in your questions; let the logic of the exemplar’s information guide your exploration; hold more and more of their process in mind; and become attuned to the idiosyncrasies of their experience.

The video clips and transcript show how we used Clean Language to interview Robert Dilts. In the Appendix, Section 11 we provide a list of the questions we asked. We have removed our repetition of Robert’s words and our few side comments to make it easier for you to see the structure of the questions and how they demonstrate the features listed above.

Modelling systemically
We think humans can be considered to be complex adaptive systems. Therefore it seems congruent to use a systemic, bottom-up perspective to model them.

From a systemic perspective, when we consider how a person consistently performs a complex behaviour we are seeking to identify ‘circular chains of causation’ rather than linear strategies. These will involve escalating and dampening feedback loops which form ‘operational closure’ (see our article, Feedback Loops). That is, these processes have enough internal autonomy to run themselves despite changing circumstances. One example is an ability that is flexible enough to achieve consistent results in a variety of new situations. In Robert’s case, to be able to select what is essential from a wide range of exemplars he has never met before.

Rather than attempting to fit what the exemplar says into predetermined categories or existing models, bottom-up modelling means we let a structure emerge out of the information itself (see our article What is Emergence?). This requires a great deal of trust that the deep structure will become evident, and the result can be something surprisingly fresh.

Outcome orientation
Maintaining an outcome orientation is vital to a modeller who does not want to get lost in the mass of information presented by an exemplar — much of which will not be relevant. The conundrum is how to work systemically with an emergent process (bottom-up) and have a predetermined desired outcome (top-down). We do this by making our desired outcome for modelling ‘a dynamic reference point’ for everything we do and for every question we ask. In this way a desired outcome is more of a signpost than a destination, and the actual outcome — what is happening moment by moment — provides the pathway. It grounds the process in sensory evidence.

Within our overall outcome for modelling we aim to identify the fundamental pieces of the exemplar's process and then figure out how they work together as a whole. We call these outcomes within outcomes 'vectors' since they determine the direction of our questions over time periods of a few minutes (see our article Vectoring and Systemic Outcome Orientation).

In essence, a systemic outcome orientation enables us to direct our clean questions to areas in the exemplar’s metaphor landscape where something new is likely to emerge. Then we hold their attention in those places long enough for them to find out more about what they do so excellently.

Concluding remarks
We hope you have enjoyed and learned from our wheels-within-wheels presentation of this material. Our aim has been to present the results of our modelling of Robert Dilts and to demonstrate that Symbolic Modelling is a valuable addition to the modelling methodologies already in use.

We would like to emphasise that a model’s usefulness is independent of how it was produced. You can make use of our model of Robert and ignore Symbolic Modelling; or you can adopt a Symbolic Modelling approach without taking on board Robert’s method for selecting – or you can use both.

Having acquired the product of our modelling – selecting what is essential – we can see how it can be applied in many areas. We recommend you try it on and notice whether the feedback you get from the world suggests you now have more choice and greater flexibility to pick out and utilise the ‘signal from the noise’.
Over the years the Symbolic Modelling process has proved fruitful in teasing out ways in which people do things excellently. It is beginning to be applied in academic research as a method to investigate phenomenological information. It has also been used to model the written word in the form of  transcripts of client sessions, business meetings, questionnaires, organisational announcements and market research. And these applications haven’t even scratched the surface of what is possible.

There is much to be learned from comparing, contrasting and combining the different modelling methodologies. And as we said at the beginning, they all have their place depending on the context, your outcome and who will be the acquirers.

Once again, our thanks go to Robert Dilts, Martin Snoddon, Fran Burgess and Derek Jackson whose commitment to making modelling more prominent made it possible for us to offer you this report.

 »  Home  »  Symbolic Modelling  »  SyM and NLP  »  Modelling Robert Dilts Modelling
 »  Home  »  Modelling  »  How to Model  »  Modelling Robert Dilts Modelling
Article Options


Clean Language
Symbolic Modelling

James Lawley
Penny Tompkins

more info

view all featured events