Context makes Clean clean

Contextually clean questions in change work
Print Friendly, PDF & Email

Table of Contents

All questions are either contextually clean questions or they are not clean. 

1. Introduction

This article examines the idea of ‘contextually clean questions’ in coaching, therapy and other applications where a person has a desire to achieve something or make changes in their life. It addresses:

What makes a question contextually clean?

What is the value of these questions?

Under what conditions does it makes sense to ask such a question?

The use of contextually clean questions has been extensively documented in relation to Clean Language interviewing (see Section 13 for references). However this is the first article to explore their use where change is the desired outcome. Although the idea and use of these questions has been around from the early days of David Grove’s invention of Clean Language, it wasn’t until 2009 that Wendy Sullivan coined the phrase ‘contextually clean’ in relation to market research interviewing using Clean Language.

This article is aimed at practitioners who have some training and practical experience in facilitating cleanly in coaching or therapy. It’s a long read which goes into the topic in-depth and explores some very subtle distinctions. My expectation is that by the end of the article I will have shined a little more light on the fascinating question: What is meant by ‘clean’?

Because contextually clean questions need to be seen in context, you can download a complete Symbolic Modelling transcript which includes twelve specialised questions. I have annotated the transcript explaining why each of these questions is contextually clean.

2. What do we mean by context?

All questions are either contextually clean questions or they are not clean. 

Another way to say this is: No question is inherently clean since ‘cleanness’ is a function of the context in which the question is asked (even though conventionally we talk about ‘clean questions’ as if they are independent of a context).

To understand this maxim I need to explain what I mean by ‘context’. Context is a notoriously slippery concept. In terms of asking Clean Language questions in a Symbolic Modelling coaching or therapy setting, we are referring to four kinds of context:

    1. The purpose of the session (i.e. the client would like to change or gain something that enhances their life)
    2. The inherent logic of the client’s inner world (a.k.a. their metaphor landscape)
    3. The physical environment in which the session takes place
    4. The wider situation (e.g. social, organisational, cultural context).

All four kinds of context will exert an influence on the client and need to be taken into account by the facilitator. However they are not necessarily equally relevant. In general, in Symbolic Modelling the first two are always highly relevant. The latter two become more relevant under certain circumstances, for example when working with a client from a different culture.

In this article I will concentrate on the first two types of context.

Next, I explore what makes a question ‘contextually clean’, the value of these questions, and under what conditions/context does it make sense to ask such a question.

3. Classically clean basic questions

Penny Tompkins and I first categorised the most common questions David Grove asked by analysing transcripts and recordings of sessions he conducted between the mid-1980s and the mid-1990s. We allocated David’s questions into three categories: basic, specialist and other. These categories are not fixed and over the years we made several small reallocations depending on: (1) The context – what they are being used for; and (2) Our deepening understanding of what makes Clean Language ‘clean’.

The main distinction that separates the basic and specialised categories is the universality of the question. In other words, what is the range of contexts in which the question is likely to be clean? The basic questions are near-universally clean because they only rely on fundamental ways people use space, time and form to organise their experience. In a therapy or coaching context the following eight questions (with slight variations) can be considered the most universal of all Clean Language questions:

    • And what would you like to have happen?
    • And is there anything else about […]?
    • And what kind of […] is that […]?
    • And where/whereabouts is […]?
    • And that’s […] like what?
    • And when/as [x], what happens to [y]?
    • And what happens just before [event]?
    • And what happens next? / And then what happens?

There are three things to note:

Each application of Clean Language needs its own set of basic questions. While creating a ‘cleanness rating’ for evaluating research interviews, I called the set of basic interview questions classically clean to distinguish them from the basic coaching and therapy questions. For example, the question ‘And what would you like to have happen?’ is rarely applicable in Clean Language interviewing and is therefore not included in that ‘classic’ set.

Secondly, ‘near’ universal does not mean ‘completely’ universal. There can be contexts where even asking basic Clean Language questions is not clean. For instance, if the question points in a direction that is antithetical to the client’s desired outcome, or when a question does not ‘fit’ within the logic of the client’s inner world (see Calibrating whether what you are doing is working or not for an example).

Thirdly, whole therapy and coaching sessions can be conducted with just the eight questions listed above. My recent research has shown that, on average, these questions form about 80% of the questions Penny and I ask.

The eight basic Clean Language questions are ‘basic’ and ‘universal’ for another reason. First, in total they only make use of 21 short simple words. Secondly, almost all of these words relate to concepts which are fundamental to all languages. These concepts are called ’semantic primes’. Originally introduced by Anna Wierzbicka, research into semantic primes has given rise to a small list of concepts that are said to be innately understood because they appear in every language so far studied and they cannot be expressed in simpler terms.

It follows that since the basic questions are mostly constructed from semantic primes (plus a few direct derivates) they have an inherent universality. 

4. Contextually clean specialised questions

Given that four out of every five questions Penny and I ask come from the basic set, it means the remaining 20% are clean because of other criteria. Most of these questions come under the heading of ‘specialised clean questions’. They are not as universally clean as the basic or classic set because they have to fulfil certain conditions before they can be considered clean.

In Metaphors in Mind we defined the ‘rule’ for using a specialist question (as they were previously called):

When using any specialist question there is an important rule: your question must be congruent with the logic of the client’s metaphors. In other words, the client must indicate that the appropriate conditions exist before the question is asked. They can do this directly, through presupposition and other forms of inherent logic, or by nonverbal behaviour. (p. 138)

We described the ‘appropriate conditions’ (context) under which 21 ‘specialist’ questions could be asked and remain clean (listed on p. 283 of Metaphors in Mind). Just because David asked a question didn’t make it clean. Like all of us, he was prone to asking the occasional unclean question! We needed to model out his clients’ inner contexts which qualified each of the 21 questions as clean. 

Recent self-modelling of Penny and my change-work transcripts has revealed that variations of just five specialised questions make up 70% of all the contextually clean questions we ask:

    • And what would [perceiver symbol] like to have happen?
    • And where does/could […] come from?
    • And does [thing] have a size or a shape?
    • And is there anything (else) that needs to happen for/to [desired Outcome/necessary condition]? 
    • And how do you know […]?

Because we use them so often I call them ‘the big-5’ specialised questions. 

In addition to the big-5, there are other specialised questions which get asked occasionally. You can download a list of them with the appropriate conditions for their use. The analysis of our transcripts showed that a question from this list was used approximately once per session.

In Section 8 below (Modelling contexts and conditions) I explain why and when Penny and I have used the big-5 questions.

There is a third category of contextually clean questions which I will address next.

5. Novel clean questions of a third kind

There is a third kind of contextually clean question that is used even more rarely. These are questions that only make sense within a particular client’s inner and outer circumstances. These questions have to be ‘tailored’ to the specific logic of the client’s metaphor landscape. They also have to be created in-the-moment because the nature of a client’s logic cannot be predicted in advance. Here is an example of a novel contextually clean question (asked in two ways on rows 15 and 17):



Ah, [Touches forehead] it’s like points but very, very tiny, very small points, which feels like some kind of activity there.



Tiny, tiny, small points there in your head that starts the movement of life. And so what kind of points are those tiny, tiny points? 



I see like a trajectory to my desired outcome to what I would like to have happen.



And when you see a trajectory to what you’d like, what’s the first point that you see?



[Big exhale] The points are so small that I don’t see them, but I feel them.



And you don’t see them, you feel them. And where do you feel that first point?



In the middle of my forehead. 

[Notes: All examples in tabular format are taken from the published transcript of a Symbolic Modelling session with a Ukrainian psychologist called Starting to Feel the Movement of Life. C=Client, P=Penny, J=James.]

Why are these two questions contextually clean when they introduce an idea, “first”, into the client’s landscape? A “trajectory to” presupposes the client is experiencing of a series of “points” heading in a direction with one of them being a ‘first’ in the series. Although we introduce the idea of “the first” point, it is congruent with the client’s metaphor which makes it contextually clean. We asked these questions because, in general, the first point (or event) in a series is usually significant, since without the first point, the later points are unlikely to happen.

It takes great skill to create these questions and remain clean. It involves more than refraining from introducing concepts; it includes only introducing sequential, causal or structural words that are inherent in a client’s logic (see Section 7). 

The rarity of these questions is clear from the analysis of Penny and my sessions. Novel clean questions and the other-than-the-big-5 specialised questions together account for only 5%, that is one in 20, of all the questions we asked.

On occasion it is contextually clean for a coach or therapist to introduce known circumstances in the client’s life into the session. In Clean Language interviewing these are called ‘topically clean questions’ since they enable interviewers to direct attention to features of the topic under investigation which the interviewee may not raise. The equivalent in therapy/coaching is the introduction of highly relevant factors known to both parties. For example, asking about the effect on children in counselling couples who are responsible for a child. However, this use of contextually clean questions is outside of this article’s scope.

6. Degrees of inference

Another way to explain ‘contextually clean’ is to understand the notion, degrees of inference, introduced by Caitlin Walker in 2002.

Facilitators have their own way of perceiving the world which influences how they model the client and every question they ask. David Grove used to say that every question was ‘up to something’. In other words, all questions include inferences the questioner is making about the way the world works or the way another person makes sense of the way the world works. 

Inferences about the exterior world are likely to be shared among people who have had similar experiences of the physical world. Interior worlds are much more idiosyncratic. In fact, often disarmingly so.

The inferences we make about a client’s inner world – the way it is structured, the logic that maintains coherence over time, and the kinds of perceptions that are possible – can be ‘closer to’ or ‘further from’ what a client presents verbally and nonverbally, as depicted in the top half of Figure 1.

Figure 1: Degrees of Inference (adapted from a model originated by Caitlin Walker)
Figure 1: Degrees of Inference (adapted from a model originated by Caitlin Walker)

The lower half of Figure 1 lists ways in which the variety of inferences are embedded in language. These are further explained in the article Clean Conversations.

Figure 2 maps the cleanness of questions on to the degree of inference model. The basic questions form the first ring and the specialised questions are in the second adjacent ring. The outer rings include questions in which the facilitator introduces more and more concepts or implicit structures not previously presented by the client. They are therefore increasingly leading questions.

Figure 2: The Clean-Leading continuum – relating cleanness of questions and degrees of inference. (Modified from an unpublished draft of Cairns-Lee, Lawley & Tosey, 2021)
7. Inherent logic

It is impossible to create a model without inference. However, some inferences are cleaner than others.

David Grove introduced us to the idea of ‘inherent logic’. That is, we can infer certain essential features are intrinsic to a concept or metaphor. For example, if a client says “I want to build a bridge” to another person, we know very little about the metaphorical bridge they have in mind but we do know that ‘a bridge’ has to have certain characteristics in order for it to called ‘a bridge’: A bridge is something that spans one or more features (such as a river, road or gap that is often regarded as an obstacle). Thus a bridge will likely have two ‘ends’, a ‘between’, and an ‘over/under’ aspect. I say ‘likely’ because fantastical things are always possible in the metaphoric world.

Making use of inherent logic requires the ability to recognise these intrinsic features when modelling a client’s metaphor landscape. While it would be appropriate to start by asking a basic question such as “And what kind of bridge is that bridge?”, most of the big-5 specialised questions would also fit with the inherent logic of the bridge metaphor. For example, things such as bridges tend to have an origin and have a particular size and shape, thus we could legitimately ask:

    • And where could that bridge come from?
    • And does that bridge have a size or a shape?

Furthermore, how the “build” of the bridge will happen may be unspecified (probably even to the client at this stage), but it does presuppose some kind of constructing process is needed, and therefore it would be contextually clean to ask:

    • And is there anything that needs to happen to build that bridge?

And we could even point attention towards the client’s motivation, the “want to”, with this question:

    • And how do you know you want to build that bridge?

The fifth of the big-5 specialised questions, ‘And what would bridge like to have happen?’ does not fit within the inherent logic of the metaphor since there is no indication the bridge has an intention.

Inherent logic is what enables other-than-basic clean questions to be contextually clean.

8. Modelling contexts and conditions

Under what conditions is it appropriate – i.e. clean – to ask an other-than-basic question? What specific client indicators suggest that asking a contextually clean question is an option? 

This is an important point. Just because the conditions exist doesn’t mean a contextually clean question should be asked. Every question has to earn it’s place. The facilitator has to take into account the logic of the client’s metaphor landscape and their desired Outcome when deciding what seems most salient for the client to attend to next.

The Big-5

Let’s start with the ‘big 5’, the specialised questions Penny and I most commonly use. Table 1 summarises generic conditions which prompt a facilitator to consider using one of these questions. Each question invites the client to attend to a particular feature of their experience (which we have called: form, intention, necessary conditions, source and knowing). 

Table 1: Conditions under which the five most common specialised questions are asked

Noticing when these conditions exist requires the facilitator to do more than simply listen or repeat back the client’s words. The facilitator needs to track the relevant indicators that imply how the client is making sense of their experience. These usually reside in the background of a client’s description.

The following paragraphs describe the common indicators which legitimise asking each of the ‘big 5’ contextually clean questions in coaching and therapy.

Remember that the two primary purposes of a Clean Language question are:

    • what it asks the client to do with their mind-body to process or make sense of the question; and 
    • what and where they are invited to attend to within their inner world – that is, the location and timeframe within their metaphor landscape.


People often refer to their experience as if it is thing-like. Clouds and feelings are not ‘things’, but when we say, “The cloud moved across the sky” or “I let out my anger”, we’re talking about them as if they are. Since a defining characteristic of things is that they occupy an area of space, it follows that they have a size and/or a shape. (We do not ask questions which reference other ‘sub-modalities’ such as colour, texture etc. because they are not an inherent property of things and we do not want to introduce features into the client’s inner landscape).

The ‘size or shape’ question is mainly used to invite the client to further attend to an unformed thing-like symbol, to notice the perceptual area the symbol occupies and whether it has an outline. As a byproduct, the client will often resort to metaphor to describe their experience, as in this example (from the same session shown above):



I have some impulse [Touches chest]. I have some impulse here.



And does that impulse have a size or a shape?



Oh, it’s like one of the sparks of Bengali fires. [Laughs]


Symbols in a metaphor landscape can play an active or a passive role. A passive symbol might be a rock that is not trying to do anything or make anything happen and is not having a response to what’s happening. It’s just a rock doing what rocks do. However, if the client says “It’s an obstructive rock” they are implying the rock’s function is to be obstructive; or “It’s a lonely rock”, implying the rock has experiences and can relate to other symbols in the landscape. These indications give the facilitator permission to ask “And what would that rock like to have happen?” in the expectation that the client will discover something about this symbol’s role in the landscape.

Continuing the session from which the previous examples were drawn, the client says that before a “boot” had “smashed” the “sparkle” of the “Bengali fire”, the sparkle:



[Big exhale] It was joyful, and alive.



And when it was joyful and alive, it was joyful and alive like what?



Like a child’s love. Like a child’s play. [Big exhale] Like love. Like dance. Like a holiday. Like celebration. [Several big exhales during a long pause]



And that sparkle was like a child’s love. Like a dance. Like a holiday. And what would that sparkle like to have happen now?



[Big exhale] She wants to be alive. She wants to live. [Teary eyes]

Necessary conditions

Once a client has stated a desired outcome, it can be useful at some point in the session, to ask ‘And is there anything that needs to happen for/to [desired outcome]?’. This question does not suggest the client needs to do anything. Rather, given the desired outcome has not yet become an actual outcome, this question enquires whether anything needs to happen before said outcome happens. Similarly, if a client has stated that a certain condition or action is necessary for their desired outcome to happen, then the same question can be asked of the condition.

In this next example, taken from close to the end of a different session, the client says:

I don’t need to be aware of the need. I just need to be aware. Full stop.

Two questions later, I ask:

And so is there anything that needs to happen for you to be aware like that?

Client: I’m already aware like that. Yes, I’m already aware like that.

Hence, I could reasonably assume that nothing else needs to happen and the session could continue to completion.

The above format of this question is cleaner than the ‘And what needs to happen?’ shown in previous specialised-question lists because the ‘What?’ version implies something needs to happen – when it might not.

However, the ‘What needs to happen?’ format can still be contextually clean under a stricter condition, that is, the client’s needs to have stated, or their logic presuppose, that something does need to happen.

The following example continues from where the ‘Bengali sparkle’ extract left off. The client’s last statement “She [sparkle] wants to be alive. She wants to live” strongly implies that something will need to happen to bring “sparkle to life” again. It is therefore unnecessary to ask ‘is there anything that needs to happen for sparkle to live?’ as the next question demonstrates:



She wants to live. And that sparkle wants to live. And what needs to happen for that sparkle to live?



Pause] Maybe acknowledgement that she exists [big exhale].

Given that often more than one thing needs to happen for a desired outcome to happen, it is usually contextually clean to follow up with, ‘And is there anything else that needs to happen for/to [desired outcome/necessary condition]?’. 

I acknowledge that the format I am now promoting may mean an additional question will need to be asked. However, my aim is for specialised questions to contain the least degree of facilitator inference, thereby maintaining the highest standard of clean. 

Until recently I mostly used the ‘What needs to happen?’ version, hopefully when the stricter condition had arisen – but I now realise not always. And I have seen many facilitators ask this question before the client has stated, or their logic presupposes, that something needs to happen.

In a similar vein, it is vital that the necessary condition questions are only asked of outcomes that the client has stated they want, would like, or need. If asked of anything else it is liable to be ‘pushy’ and leading and therefore not contextually clean.


Things and experiences usually do not appear out of nowhere, rather they are perceived as coming from or coming out of somewhere else (usually, but not necessarily, a prior source). Therefore, enquiring ‘And where could/did […] come from?’ invites the client to identify the source of a particular symbol, action or relationship. That source might be: a location (e.g. the start of a river), a cause (e.g. excess rain leading to a flood), a previous owner (e.g. a benefactor), or the origin (e.g. conception) to name but four of the common kinds of answers clients give to this question.

The following example continues immediately after line 36 above:



And where could an acknowledgment like that come from?



Maybe from my opened eyes. Now I begin to understand that I was afraid of this boot. I was scared, I was scared to live. I was scared to carry inside of me this sparkle. Because if I have the sparkle inside of me the boot can come and smash it.

This new realisation likely indicates that something is changing for the client.

The source question can prompt a client to make the most unexpected of associations and their attention can end up in a multitude of times and places. A client’s attention may leave the current landscape and arrive at a prior metaphorical event, or a childhood memory, or on the planet Zog. 

This presents a potential downside. The facilitator cannot know in advance where the client’s attention will end up until they have responded to the question. It may be at a space-time of major significance, and it can be a distraction from attending to a pivotal moment in the current metaphor landscape. Therefore, the facilitator needs to have assessed that the client’s logic means that sending the client on the quest to find a source is likely to be beneficial. 


Language creates the ability to describe experiences with abstract concepts and unspecified words. This can be a great advantage and it can separate us from our lived experience. 

For instance, a client used the expression “… as if I exist in this world”, leaving their lived experience unspecified. By asking the “And how do you know …?” specialised question I invited the client to consider the experience or criteria they use to be aware of existing in the world:

And so when you exist in this world with calm and balance, how do you know you exist?

I have an impression that there is some consequences of actions. And when I see these consequences and actions I have a feeling that my body functions. But functions not like now, because now it’s like my body is broken and chaotic, and in a lot of pain.

The session continued with the client further developing the “feeling my body functions”. As demonstrated in this extract, people usually answer this question with a more sensory-based description or an example.

While the basic format of this question orientates the client to the present, the question can also be asked of past and future experiences, e.g. ‘And how did you know [experience]?’; ‘And how will you know when [experience]?’.

9. The value of contextually clean questions

There are three kinds of knowledge: known, tacit and emergent. Contextually clean questions can play a role in facilitating the client to access all three kinds of knowledge. By the way, I am using ‘knowledge’ in a very wide sense of the word.

Known knowledge

Clients come with all sorts of prior understanding, although sometimes it is not until they are asked a question that they become aware of the information or experience in that moment. Specialised questions can invite a client to attend to certain features directly. 

For instance, a client may be fully conscious of a certain image they are seeing with their minds-eye, but until they are asked, they may not have been attending to its size or shape. Once they do, they may discover something unusual or surprising (“Wow, it’s lopsided”), or put the image in relative proportion (“it’s smaller than I had first thought”), or trigger an association (“it’s the same size as …”).

Often, clients have simply not put enough attention on what is important. For example, a client may fret or be paralysed by how they are going to achieve some long-term goal and as a result spend little or no time considering the very next step in the process.

Tacit knowledge

Contextually clean questions are useful for inviting the client to consider features of their experience they might not usually attend to. This is particularly relevant for tacit knowledge: Knowledge the client has not articulated to themselves before, and may not even realise they know until they are asked the question. David Grove used to say “I’m interested in what’s not there, that needs be there, for what is there to make sense.” This is a perfect description of tacit or implicit knowledge.

Specialised questions can facilitate a client to access a deeper knowing beyond their everyday story. They can invite a client to examine aspects of themselves they haven’t done before. These may be characteristics they don’t like, are ignoring, discounting or deceiving themselves about. For example, a client may discover that an apparently restrictive bubble surrounding them has an intention to protect a fragile heart from words that wound.

Equally, tacit aspects may be resources which are undervalued or unrecognised. For instance, the source of a resource is usually even more resourceful than the resource itself. Clients are more likely to access such a state with the skilful use of the ‘Come from?’ question. For example, another Ukrainian client had a desire, “a want to support” a family where two days previously the father had been killed in the war but the client couldn’t find the words. I asked:

And so where does your want to support the family come from?

“From love” she replied. The client had become aware that the source of her desire was another resource. After a few more clean questions “love” became a resource metaphor that the client needed to be able to find the words to support the family:

It’s like planets – like a galaxy where the planets – the words go [by] one-by-one [her gestures depicting the metaphor].

Not something anyone, including the client, could have predicted in advance.

Emergent knowledge

In conjunction with the basic questions, contextually clean questions can help create the conditions for emergent knowledge: knowledge, perspective and experience that is created in the moment. This knowledge doesn’t exist in a client’s system until the confluence of conditions brings forth a creative realisation. One of the aims of Symbolic Modelling in a coaching or therapeutic context is to facilitate the client’s system to create the conditions which encourage spontaneous creativity and the emergence of new knowledge, perspectives and experience.

Symbolic Modelling is an entirely ‘additive’ process. It never seeks to take away any features of the client’s experience. It maintains this additive stance because, according to Complexity theory: (1) it is impossible to know in advance the precise conditions needed for emergent knowledge to be generated from a client’s creative unconscious (or wherever you believe it comes from); and (2) even small changes can cascade into something more systemic. 

It’s difficult to give a short example of the appearance of emergent knowledge because these moments happen as a result of the accumulative effect of the client processing all the prior questions, not just one contextually clean question. David Grove said “Change takes place in a context”. This is why we emphasise the importance of facilitating a client to establish (and maintain awareness of) an embodied metaphor landscape. This is the context in which spontaneous creativity flourishes. Contextually clean questions play a supporting role by helping to fill in some of the gaps, extend the boundaries of the landscape and bring tacit knowledge to the fore. 

The transcript of the session we have been following illustrates this point. Towards the end of the session, the client is “surprised” by her inner world (usually an excellent sign of pending changes) and she sets a new desired Outcome, “I would like the sparkle to become bigger”. To check whether what is wanted is possible, I ask the specialised question ‘Can [a want/need] happen?’.



I am a bit surprised by everything. I didn’t see my fear. And I didn’t see the sparkle. And when I started to see and acknowledge this fear, it is an important point for me. I didn’t see that this is so deep.



And now that you acknowledge the fear and you acknowledge the sparkle. And when you acknowledge the fear and you acknowledge the sparkle, what would you like to have happen?



[Clears throat and smiles] I would like the sparkle to become bigger.



And can that sparkle become bigger?



Now the sparkle is like a salute on the day of independence of Israel [Big smile] and now I begin to understand that this story is about Holocaust. [Tears]

To my surprise, the question gives rise to the emergence of a realisation and a spontaneous change of the “sparkle” into “a salute” that transforms the client’s state and perspective.

10. Learning to use contextually clean questions

To become proficient at using and creating contextually clean questions during the back-and-forth of a session requires developing a range of skills:

  • Exquisite listening to what and how what is being said is being said
  • Maintaining a soft focus on the client’s gestures and other body movements
  • Understanding what constitutes clean and leading questions
  • The ability to symbolically model-in-the-moment
  • Appreciation of the: 
    • Logic of form
    • Logic of time (and sequence) 
    • Logic of space
    • Logic of causality 
    • How presupposition works.

The ability to create novel contextually clean questions in real time requires practice offline. Penny and I spent many hours with David Grove attempting to craft novel questions that felt right to him. It seemed to us like he tested variations of a question – changing a word, the syntax or the rhythm – against his intuitive ‘cleanometer’, and as he put it, where and how it might “land” in the client’s perceptual space. 

I remember being a demonstration client on a therapist retreat sitting in silence when David paused the session. He conversed with the group about trying out various question formats. When after an hour, David finally turned back to me and asked one of the questions I was ready to explode – I had been running every question through my system and was full to overflowing with answers!

As these skills are acquired and honed offline, you will learn to remain clean in real time even when enquiring about the background, tacit, and as yet unspoken aspects of a client’s inner world.

11. What to watch out for

Specialised and novel contextually clean questions can be seductive because they make it easier for a facilitator to unintentionally slide out of maintaining a clean perceptual environment for the client. Then inferences and assumptions derived from the facilitator’s logic can intrude into the question in subtle and often unnoticed ways.

Staying ‘clean’ is more challenging than it seems. Even highly experienced practitioners can find it difficult to stay 100% clean once they start using questions outside the basic set. The clean-to-leading continuum represented by the arrow in Figure 2 can be considered a slippery slope since the opportunity for leading is increased as the facilitator slides down the incline.

In other words, once a facilitator allows themself to use other-than-basic questions it increases the difficulty to remain within the client’s idiosyncratic logic, especially when novel questions are created in the moment. 

My analysis of 17 transcripts of Penny and my sessions showed that about 20% of the questions we asked were clean because of their contextuality. This is an average; in one session no specialised questions were asked, while the session with the most had 12 (see the downloadable transcript from Section 1). It is important to note that these figures are relevant to Penny and my sessions. They are certainly not rules nor even guidelines. Other clean facilitators may ask more or less contextually clean questions than we do.

Over the years we have noticed patterns in the ‘misuse’ of specialised questions. Primarily it happens when a client has not presented sufficient context or logic for the question to be contextually clean. Table 2 shows some typical inappropriate examples of the big-5 specialised questions and suggests a cleaner alternative question. 

Table 2: Typical conditions in which asking the big-5 specialised questions is inappropriate (unclean)
Table 2: Typical conditions in which asking the big-5 specialised questions is inappropriate (unclean)
12. Concluding thoughts

Cognitive Linguistics is based on the idea that there is a correspondence between our language and the organisation of our cognition. Thus, we can infer things about the way another person is making sense of their experience from their language and behaviour.

It is easy to make these inferences. We do it all the time. The challenge is to make inferences based on a client’s descriptions and logic, rather than our own. To do this we need, to a large degree, to set aside our own way of making sense of the world. This ability doesn’t come naturally. It needs training.

Despite Carl Rogers’ theory of counselling, there is no such thing as a “non-directive question”. All questions are directive – even clean questions – in that every question invites the client to consider one thing rather than another. The difference is that Clean Language questions do not introduce any content and they only make use of near-universal cognitive structures or idiosyncratic logic.

Clean Language questions direct (although I prefer to use the metaphor ‘invite’) the client to attend to a particular space-time in their inner world. And Clean language does this by only incorporating information that has already been supplied by the client (using their exact words as necessary).

Thanks to Gregory Bateson I now understand that words require a context to have meaning. Calling the questions we have been examining contextually clean questions emphasises the need for a particular context to exist to legitimise the question. In the early stages of a session, the basic clean questions are usually all that are needed to facilitate the client to identify a desired outcome and begin to establish a metaphor landscape. Once a landscape exists, exploring the landscape follows naturally, and this is where specialised questions can come in handy.

We find it fascinating that, given the thousands of questions people can ask, there are so few that leave the client with the freedom to answer in their own way. If we examine the pattern that connects all these non-leading questions we find they are modelling questions. And on top of that, they are questions which invite the client to self-model. One of the features of only using Clean Language questions is how they support the facilitator to stay in modelling mode.

Clean Language questions are a way to facilitate a client to go beyond noticing the what, where and when of their inner world and to self-model their own logic – the glue that holds their metaphor landscape together. 

My recent analysis of the questions that Penny and I asked in 17 sessions drawn from the last 10 years revealed the following average percentage of questions asked:

80% – The 8 basic clean questions

15% – The big-5 specialised questions 

5% – Other specialised and novel contextually clean questions 

That is 95%, 19 out of every 20 questions asked came from (variations of) of just 13 questions. 

Contextually clean questions used in interviewing are well-documented. In this article I have presented the first in-depth exploration of their use in change work since Metaphors in Mind was published in 2000. I have defined ‘contextual clean’, explained its application and given examples from actual sessions. 

I have also examined why contextually clean questions are important in therapy and coaching:

  • They allow for a greater range of questions while remaining clean. 
  • They invite the client to attend to features not directly referenced but that can be presupposed to exist in the background 
  • They enable a greater degree of precision of the placement of a question in the perceptual time and space of a client’s inner landscape.
  • They can build deep rapport since the client knows the facilitator is not only listening to their words, but to the logic of their experience by ‘modelling in the moment’.

And hopefully I have demonstrated why it is that all Clean Language questions are either contextually clean questions or they are not clean.

13. Other resources

Specialised clean questions

James Lawley & Penny Tompkins (2000). Metaphors in Mind: Transformation through Symbolic Modelling. See pages 103–119, 138–144. 159–160, summary sheet p.283.

Gina Campbell (2013). Mining Your Client’s Metaphors: A How-To Workbook on Clean Language and Symbolic Modeling Basics Part II: Facilitating Change.

InsideClean Series 1 workshop 4: Using specialised clean questions.

Clean-Leading continuum

Phil Swallow & Wendy Sullivan, (2004). Whose map is it anyway?, Proceedings of the Integral NLP Conference, 12-13 June.

Inherent Logic

Mark Johnson (1987). The Body in the Mind,

Seth Lindstromberg (1998) English Prepositions Explained.

The inherent logic of containers and balance.

Clean Language Interviewing

Cairns-Lee H., Lawley J. & Tosey P. (Editors). (2022). Clean Language Interviewing: Principles and Applications for Researchers and Practitioners. Emerald Publishing.

Cairns-Lee, H., Lawley, J., & Tosey, P. (2021). Enhancing researcher reflexivity about the influence of leading questions in interviews. The Journal of Applied Behavioral Science.  [Download preprint version]


Much appreciation goes to Jacqueline Surin and Penny Tompkins for their thought-provoking questions and editing of my drafts.

Print Friendly, PDF & Email
body * { color: inherit !important; }