ChapterPDF Available

Nehyba, J., & Svojanovský, P. (2017). Clean language as a data collection tool. In V. Švec, J. Nehyba & P. Svojanovský, et al., Becoming a teacher: The dance between tacit and explicit knowledge (p. 130-147). Brno: Masaryk University.

Fih Chapter
Jan Nehyba, Petr Svojanovský
is chapter explains and evaluates how the data collection method entitled Clean
Language (described in detail in Chapter3) was implemented in our research. In the
context of pedagogical sciences, it is anew way of interviewing, which helps to gather
data as closely as possible from rst person (Searle, 1992; Varela, 1999). In the context
of our research, we assume that the more we obtain data from the position of the rst
person, the better the quality of data. We understand high-quality data as information
collected rst hand from the world of the research participants, i.e. information that
is aected as little as possible by the researcher’s perspective during the interview.
In line with the denition, we believe that tacit knowledge is less conscious than
other knowledge, and dicult to articulate. It follows that in research interviews, it is
important to ensure as far as possible that the structuring of such knowledge comes
directly from the informants. In other words, the less the researcher intervenes in the
interview, the greater potential for the elicitation of tacit knowledge. Clean language
enables informants to explore what Petitmengin (2014) refers to as the microstructure
of their experience, thus helping them to grasp what is less conscious and dicult to
articulate. Intervention by the researcher in the content of the interview (paraphrasing
or interpreting what has been said or introducing completely new topics) can distract
the informant from accessing essential details that may contribute to awareness and
the ‘naming’ of less obvious aspects of their own experience.
In this chapter, we explain how we understand the term Clean Language and how
we interpret it, and then move to the actual analysis of how we used this method in
practice, i.e. in conducting the research interviews.
5.1 Conducting interviews using Clean Language
Although the method of Clean Language interviewing is based on several clear ideas
(for example, repetition of the participant’s verbal and non-verbal expressions, use of
cleanquestions), dierent aspects of this method can be highlighted; for example,
whether emphasis is placed on the natural formulation of questions asked in an inter-
view, or whether more emphasis is placed on the use of strictly ‘cleanquestions. ese
dierences may appear as subtle nuances but they greatly inuence how an interview
is conducted. ere is adierence between strictly adhering to alist of cleanquestions
and asking questions guided by the informants previous answer(s) (although these po-
sitions are not necessarily mutually exclusive). erefore, we consider it important to
present how we think our research team understood and applied the Clean Language
interviewing method. In the following paragraphs, we present those aspects of Clean
Language interviewing we consider important.
e value of cleanquestions in an interview relies on an objective concept of clean-ness.
‘Clean’ questions have no deeper’ meaning and donot demonstrate empathy for the
interviewee. e assumption is that these very specic questions help to systematically
eliminate the interviewer’s own assumptions, so that they donot unduly inuence the
interview. Our use of Clean Language interviewing involved using questions taken
from aclean questions list93. e belief is that these specic questions help us systemati-
cally eliminate our own assumptions, so that they donot inuence us when conducting
an interview. Our view is that Clean Language interviewing helps us, through the use
of clean questions, to “minimize” any assumptions within the questions themselves.
Every question has certain assumptions but clean questions are designed to contain
as few of these assumptions as possible. For example, the question: “What kind of
X is that X?” assumes only some form of existence of X, as opposed to the question:
“What doyou think about X?”, which assumes that the informant has to think, and
not, for example, feel, something about X, etc. (X represent aword or anon-verbal
gesture of interviewee). “What kind of X is that X?” assumes that since the interviewee
has mentioned X, then X will have some qualities which enable the interviewee to
distinguish X from not-X.
Although, through their construction, even clean questions inuence how the inform-
ant approaches their experience, they dothis much less than traditional open interview
questions (see Chapter3). As aresult, the Clean Language interviewer becomes more
self conscious about their own language when conducting aresearch interview. We
see this ‘sensitization’ as the most important benet of Clean Language interviewing.
Clean Language also helps researchers to recognize and minimize, rather than
eliminate, assumptions in relation to individual interviewees that aect the rapport
or the relational level of communication (cf. Hulburt, 2011, Chapter20). is more
contextual and relational concept of clean-ness in an interview represents areturn to the
original idea of the creators of Clean Language, Grove and Panzar “We cannot dene
in advance the grammar, syntax or vocabulary of aclean question. A‘clean’ question is
unique to each client. We can give general rules dening clean questions. Nonetheless,
we have to discover which questions will t which client.” (1989, p.23)
93 ese lists of questions dier from each other (to agreater or lesser extent), both in terms
of quantity and quality (inclusion of adierent type of questions on the list), depending
on the context and practice of each author. For lists of ‘clean‘ questions, see, for example,
Lawley and Tompkins (2000); Harland (2012a); Way (2013); Tosey, Lawley, & Meese (2014);
McCracken (2016).
In this contextual model of clean-ness, the role of rapport (the relational level of the
interview) is crucial for obtaining high-quality content, which for us, is data that is
as close as possible to the rst-person perspective. In our research, rapport was sup-
ported in particular by specic verbal comments. In itself, successful use of exclusively
clean’ questions creates asafe environment in which the informant can concentrate
on their own inner world and in this sense supports acertain rapport between the
informant and the researcher. Clean Language interviewing aims to maintain the rap-
port between the informant and their inner world of experience, however, to achieve
this, it is necessary also to maintain researcher-informant rapport94 .
Clean Language interviewing primarily inuences the process of conducting an interview
and aims to minimize inuencing the informants experience in terms of content. We can-
not conclude, however, that Clean Language does not inuence the interviewee. On the
contrary from the perspective of social constructionism (Gergen, 1999), the extended
mind (Rowlands, 2010) and in reference to neo-pragmatism (Rorty, 1999), this is not
even possible. e dierence is in how Clean Language interviewing inuences the
informant. It deliberately inuences the interviewee so as to keep their attention in
their own eld of experience95 to be able to see the phenomenon in question from the
closest possible position to their own96. We donot inuence the content of their atten-
tion by adding new topics but inuence what part of their eld of experience they talk
about. erefore, we can refer to this method as asecond person interview, which helps
the interviewee come as close to themselves as possible (however, from our perspec-
tive, we can never cross this border). us, it is about the degree to which we come
closer”, in our interview, to where the interviewee’s attention is. We also perceive it as
clean when we come close to where the interviewees attention is using clean language
syntax97, and then directing their attention to the “edge” of their perception of personal
experience (using aclean question). For example, the informant makes the statement:
“Isee myself connected, how all that pupils’ energy is owing to me.We keep the
interviewee’s attention on the entire description of their experience by repeating their
words, and subsequently, we direct it to the kind of “connection it is, although the
94 We are aware from personal experience that experts in Clean Language are able to establish
rapport by using ‘clean’ questions and by non-verbal expression (mirroring, etc.).
95 Cf. Urban (2015, p.44): “Husserl introduced the term eld ... with aconscious reference to an
analogy to common experiential elds such as visual eld, tactile eld, etc.
96 is assumes adivision between one’s own experience and the experience of another, which
is our personal construct. Our own experience refers to how Iexperience writing these lines
while the experience of another refers to how someone else experiences writing another chap-
ter of this book. erefore, Itry to access this experience of another, to see what this experi-
ence of another looks like. If Iwanted to return to my own experience, it would mean that
Ihave to return to how Iexperience focusing on the experience of another.
97 Syntax is how aresearcher composes aquestion for the informant. In formalized form, it con-
sists of three parts: (1) … and [client’s words]. (2) … and as/when [client’s words], (3) [clean
question]? However, the researchers did not always exactly observe the three parts; sometimes,
they would only use 3. aclean question.
interviewee’s attention would originally be directed to, for example, elaboration on the
topic of pupils. If our question “hits” where the informants attention is or where it is
directed, we can immerse” them even more in the “stream” of their own experience. As
aresult, it can help the informant access even that content that is not obvious to them98.
However, the aim of the research interview is not only “immersion” but a“balance” of
this immersion and nding information in relation to the research question.
In relation to the topic of clean-ness in an interview, it should be noted that the very
concept of ‘clean-ness’ is ametaphor, and some authors even consider it, in the context
of experimental research, an embodied metaphor that inuences our moral evaluation
(Zhong & Liljenquist, 2006; Schnall, Benton, & Harvey, 2008). is is then represented
with conceptual metaphors: CLEAN IS GOOD and DIRTY IS BAD. We are aware
of this tendency, however, we understand the term ‘clean’ dierently. To conduct an
interview in Clean Language does not mean to conduct agood interview but to come
as close as possible to the informant’s rst-person perspective. We emphasize that an
interview conducted in clean language is dierent, not better. It produces adierent
type of data than interviews conducted in aconventional way (cf. the hermeneutic
conception of understanding as ‘other’, not ‘better’, Grondin, 2007, p.174).
5.2 Analysis of conducting research interviews
In this section we rst emphasize the selection of interviews analyzed and describe
each phase of the analysis. We then move to the ndings and, nally, to the discussion
of the results.
In total, the research team conducted 44 in-depth, unstructured interviews between
September2013 and January2016. e interviews were conducted by seven trained
researchers. All the researchers had been trained in how to conduct interviews using
Clean Language and they consulted with Clean Language experts James Lawley, Penny
Tompkins and Caitlin Walker. ree researchers had additional practical experience
conducting interviews using ‘clean’ questions (researchers 1, 2 and 7) because they had
attended an ocial workshop on Clean Language interviewing that included practical
training. Four researchers had only received several hours of training (researchers 3,
4, 5 and 6). Researchers 1, 2 and 3 each conducted approximately one-quarter of the
total number of interviews, providing approximately three-quarters of the data col-
lected. To evaluate the manner of implementing Clean Language when conducting
research interviews, the researchers randomly (by drawing lots) chose one interview
each, which was subsequently analyzed (i.e. 3interviews in total). e last (fourth)
interview for analysis was randomly selected from the remaining batch of interviews
conducted by one of the remaining researchers (Table4).
98 Cf. the use of trance as elicitation of experience (Lifshitz et al., 2013).
Table 4
Basic information about research interviews and their analysis.
of research
Number of
Topic of the analyzed
Informant in
the analyzed
Researcher 1 10 23% 1 subjective conception
of teaching Karel
Researcher 2 12 27% 1 subjective conception
of teaching Ema
Researcher 3 10 23% 1 didactic transformation
of content Františka
Researcher 4, 5, 6, 7 12 27% 1 didactic transformation
of content Ema
Total 44 100% 4
5.3 Phases in the analysis
e analysis of interviews was based on the protocol for ‘clean-ness’ validation when
conducting aresearch interview (Chapter3.5), which sets out four basic categories
evaluating the degree to which the researcher’s questions inuence the content of the
informants statements: (1) classically clean; (2) contextually clean; (3) mildly/poten-
tially leading; and (4) strongly leading. One trained researcher began (deductively)
analyzing the questions in the interviews according to these categories. She catego-
rized all other statements (comments) in the interviews, thus gradually (inductively)
creating the typology of the comments. is was the rst reading of the data conducted
by atrained researcher.
e categorization of questions and comments was subjected to re-analysis by two
other researchers (authors of this chapter) is was the second reading. It became ob-
vious that theoretically designed categories for the evaluation of questions were too
vaguely dened, and it was not possible to reliably distinguish in which category each
question belonged. Consequently, we started to approach the analysis inductively.
As far as the comments were concerned, it became apparent that the researcher in-
uenced participant statements to varying degrees—our hypothesis when we began
categorizing the comments according to the degree of inuence on the informant.
e originally categorized questions were revised (the third reading—again conducted
by the researcher based on instruction), and a new typology of questions and new
denitions of categories were created containing individual types of questions. e
comments were also categorized according to the degree of inuence on the partici-
pants’ statements.
e last phase was the fourth, nal, reading (by the authors of this chapter) where we
examined to what degree our division of individual types of questions and comments
into categories of ‘clean-ness matched that of the researcher’s. e categorizations were
amended and the partial denitions nalized.
Lawley’s original categories (Chapter3.5) were adapted to the context of our research
based on the iterative process described. First, the adaptation included adierent ‘un-
derstanding’ of the scope of each category. We took ‘scopeto mean what is logically
(and on aregular basis) understood as asummary of objects which fall under agiven
category (e.g. objects falling under Category1 dened by us); in other words the scope
of aterm. Our scope of categories is much broader than Lawley’s. Lawley’s category of
classically clean questions includes only prescribed strictly clean questions, whereas
our Category1 also includes some conversational ways of using clean language. On
the one hand this was due to the fact that we are not as experienced in conducting
clean language interviews, on the other hand, it was aresult of the fact that criteria
other than objectivistically dened clean-ness (where clean is dened only as clean
questions) were also important to us
e results of the analysis are summarized in the following sections—rst adescrip-
tion of the qualitative analysis of questions and comments (types and dierences
among them) followed by an examination of the quantitative analysis (the percentage
of individual categories of questions and comments analyzed in the interviews).
5.4 Qualitative analysis
In this chapter we describe the dierent categories considered in the analysis. e
questions are ordered according to the degree of content-inuencing on the partici-
pant—from Category 1, which includes the least content-leading questions or com-
ments, to Category 4, which includes the most aecting questions. e manner of
categorization is illustrated in specic examples.
5.4.1 Categorization of questions
Category1: ‘Clean’ questions—variants99
(a) ese questions included only the informant’s exact words supplemented by
some of the clean questions on the list created by Lawley and Tompkins (2000,
pp. 282–283). ese questions are ‘clean’ without depending on the context in
which they are asked (context-independent).
99 For each category of questions or comments where dierent variants are presented (this ap-
plies to Categories 1and 2 for questions and comments), we list the variants that repeatedly
occur in the interviews. ose occurring only exceptionally were not included in the list.
Example 1
Participant: Well, it was adisappointment that the plan, what Ihad expected, was not
Researcher: What kind of disappointment? (question What kind of? in the list by
Lawley and Tompkins)
(b) ese questions were variations of the basic question “What kind of X is it?” becau-
se they donot contain any topics, opinions, ideas, beliefs, etc. that the researcher
would bring into the interview through these questions. is is only avariation of
the wording of the questions.
Researcher: What is X about for you?
Researcher: How would you name X?
(c) ese questions contain words that donot contaminate the informant’s statement
in terms of content (at the level of the informant’s external speech). In essence,
they are paraphrases of clean questions in which, however, there is no semantic
shi. ese are clean questions uttered in one’s own words, where these words are
commonly shared expressions of communication.
Example 1
Participant: … When Igo into the classsroom, Ifeel rising tension ... then Igo to the
teacher’s desk and the tension uctuates … and when Isit in the teacher’s chair … it
goes away.
Researcher: And if everything goes like that, what happens next? (aparaphrase of the
question: What happen next? in the list by Lawley and Tompkins).
Category2: Contextually clean questions—variants
(a) Verifying questions—used by the researcher to verify they understand correctly
what the informant is saying. In fact, it is a paraphrase to clarify particular in-
formation in the informant’s statement. is is not an attempt to paraphrase the
meaning of the statement (as is the case in the category of medium-inuencing
Participant: e teacher tells me: Could you dothis topic and nothing more.
Researcher: So the teacher tells you what topic you should do, and it is up to you how
many texts, authors, there are?
Participant: When Iam in alesson and Ifeel aconnection with apupils.
Researcher: And now precisely you’re talking about maths or are you talking about...?
(b) Introductory questions—used by researchers to initiate an interview with an
Example 1
Researcher: When Isay “you and teaching”, what does it doto you, what could you say
in that respect?
(c) Questions aimed at the manner of expression—the purpose of these questions
is to invite informants to express themselves using the selected instruction
or technique. ese questions are not a direct focus on the experience of
Participant: When you think about it, its like Ilooked at it from afar.
Researcher: Now when you look from a distance and look at that, what it was about,
what was created, what’s happening inside you?
Participant: Is it like asphere with many connections. is is the most concise.
Researcher: And can you draw it here?
Category3: Medium-inuencing questions
ese are questions that contain words the informant has not said and that
introduce potentially new topics or links, or asemantic shi into the interview.
ese questions contain aparaphrase of the student’s statement.
Participant: Well Ithink that for many there is the eect that if they fail three times in
arow, there will come this: “Iwill fail again anyway.
Researcher: So the bad marks, the three failures make them give up, saying it doesn’t
matter anymore?
e example above contains aparaphrase, potentially bringing in the new topic that
the subjects do not care. Although this paraphrase may seem to correspond to the
student’s statement in terms of content, this cannot be said with certainty. e inform-
ant could have, in the background of their statement, implicitly perceived adierent
meaning (adierent topic), for example, that the teacher’s marking is unfair.
A paraphrase is always an interpretation because the same thing said using other
words creates the potential for asemantic shi in these other words. In this respect
, it is dicult to distinguish paraphrasing (medium-inuencing questions) from in-
terpretation (strongly inuencing questions), i.e. the extent to which the meaning of
the informant’s words was or was not changed, and the degree to which the inform-
ant was inuenced by the researcher. e inclusion of aquestion in agiven category
then depends on the researcher’s sensitivity to distinguishing the degree to which the
meaning of the informant’s statements was changed in the question. Disputable cases
(where researchers did not agree on the inclusion of a question in aparticular cat-
egory) were re-discussed among the researchers.
In the following example, the underlined words indicate apotential semantic shi in
the researcher’s question.
Participant: Ihave already given up on passing on to them everything Iprepare because
that has hardly ever worked out. So Irather hope that about aquarter of what Isay
sticks in their heads ... that maybe in the next class they will be able to repeat or
answer afew follow-up questions ...
Researcher: So if it sticks, at least that quarter, that means that you have something
to follow up in the next class that they will respond to your questions, that they will
actually remember the subject matter, what you had done?
e student’s wording everything Iprepare is paraphrased by the researcher as subject
matter. is represents asemantic reduction, and thus asemantic shi in the state-
ment—everything the student prepares for her class need not, in terms of content,
relate to the subject matter. e student’s wording about aquarter sticks is paraphrased
by the researcher as sticks, at least that quarter. e paraphrase carries an implicit
assumption that if less than aquarter is remembered by pupils, it would be impossible
to follow up on the previous lesson in the next one. e researcher thus introduces
apotentially new link, apresumption about “if-then” causality.
Category4: Strongly inuencing questions
ese questions include words the informant has not said, and explicitly introduce
acompletely new topic or link into the interview. ese questions contain an interpreta-
tion of the student’s statement.
In the rst example, the informant describes her experience in class where pupils are
unable to solve aMath problem without her support (specically her physical presence
and non-verbal signals). e researcher’s response was to encourage the informant to
think about whether she tried to change, to eliminate, this behavior in the pupils in
any way. However, such considerations were not present in the informant’s statements.
By introducing acompletely new topic, the researcher heavily inuences the content.
Participant: … they are not able to solve the problem without me sitting there with them
and nodding yes ... it seems to me alot of kids have problems with this.
Researcher: And doyou remove that somehow...?
Asimilar situation is also illustrated in the second example. e researcher introduces
an explicitly new topic in the interview—taking into account what the kids are like in
preparation for classes.
Participant: … Ithink that Inotice that, what the kids are like and what they doin the
class. Of course Idon’t notice all of them in one class ... but Ihad singled out afew
people [pupils] Iasked [other teachers] about...
Researcher: And then when you know, or you probably must have known, then what—
did you take it into account in preparation or how did you proceed?
5.4.2 Discussion on the categorization of questions
Variations in clean’ questions and Categories 1a, 1c and 2a refer to what Lawley and
Tompkins (2005) describe as a conversational conception of clean language. ‘Clean
conversation’ (dialogue) diers from the use of clean language in that:
1. in ‘clean conversation, the interviewer intends to achieve something (for themselves);
in the context of research, the intent of aresearcher is to explore the informant’s expe-
rience in acertain “framework, created by the research question in the interview;
2. it happens in the real world, and therefore it is possible for the interviewer to assu-
me more than in the metaphorical landscape; for example, in clean conversation in
the ordinary world we assume that the laws of physics apply, whereas this need not
be the case in the interviewee’s metaphorical landscape (cf. law of cartoon physics,
Harland, 2012a, p.56);
ese sub-categories (1b, 1c, 2a) refer to what is called, in the context of exploring ex-
perience in exposure interviews (pertaining to methods of examination of experience
that is the closest possible to the rst-person perspective), the deliberate inconsistency
of aquestion (Hurlburt & Schwitzgebel, 2007, p.15). However, this idea goes partially
against the clean language philosophy. Hurlburt & Schwitzgebel claim that if acertain
experience is suciently “robust”, then asking a question repeatedly and dierently
(inconsistently) will lead to asharpening of the meaning of the experience (Hurlburt,
2011, p.161). From the perspective of clean language, we can agree with this only in
relation to the repetition of aquestion (cf. Harland, 2012b) since, as Hulburt himself
says, each of these questions has its advantages and disadvantages and, from the per-
spective of clean language, the greater the consistency of the question, the more the
disadvantages are minimized.
Hurlburt thus assumes that our experience can be “robust” and we can vary questions
to describe the experience, and place the emphasis on “playful” phenomenological
variations in the questions (cf. Ihde, 2012). is means oering anumber of possible
questions that enable adeeper exploration of the experience from dierent angles.
By contrast, Clean Language assumes apotential for “fragility and uidity” in some
moments of inner experience, which may fall apart aer even the slightest inuence
on content.
5.4.3 Categorization of comments
Category 1: Positively inuencing comments
ese are comments that strengthen the relationship with the student and en-
courage open and detailed exploration of the structure of their own experience.
Although these comments usually also include words the informant has not
said, they focus on the process of the interview, not the content of the interview.
(a) Showing understanding and personal involvement
e research analysis categorized only the more apparent expressions of active lis-
tening. One-or-two-word expressions such as hmm and oh, good were not coded as
comments and thus donot inuence the overall frequency of the comments in this
Participant: Is like alight bulb … for many people is not clear.
Researcher: Clear, clear. Clear, light bulb. Yes, yes, yes.
Participant: Ifeel as sun in the middle of classroom.
Researcher: Hm, hm, hm, OK, good.
(b) Stabilization of attention through aliteral replica
Example 1
Participant: Yeah, in physics, the relationship between understanding and learning, Ire-
ally think it’s easy—if there’s no understanding, there’s no learning Ithink.
Researcher: No understanding, no learning {nodding}.
(c) Assurance leading to openness
Example 1
Participant: It’s illogical what Isay. Iknow … Ishould only use the correct term, that we
learned in school.
Researcher: … it doesn’t have to be completely logical ... if something is not right, you
maybe correct yourself or don’t correct yourself {gesticulating} simply if it isn’t
exactly as you have it, it’s not aproblem.
Category 2: Context-bound comments—neutral
(a) Pre-framing the interview—explanation of what the research is about, how
the interview will be conducted, etc.
Example 1
Researcher: ... Iwill be asking something, you will try to reply, just note that some of the
questions may sound abit strange ... whatever crosses your mind, whether it’s athou-
ght, afeeling, some whatever, it belongs here, that’s why we’re here ...
(b) Rening the instructions—these comments are the researcher’s attempt to direct
the participant’s attention so that it conforms with the research question. No new
content is introduced, only adeveloping of what has already been said.
Researcher: Iwould come back to you saying you explain it to them in very simple terms.
Researcher: Elaborate.
(c) Comment associated with instruction/technique
Example 1
Researcher: … we would try automatic writing, which means that on the topic Igive
you will write for three minutes without having anything for preparation, and what
is important is that your hand must not stop...
Category 3: Inuencing comments
ese are comments that contain words the informant has not said and that introduce
potentially new topics or links, or asemantic shi into the interview. ese comments
typically contain aparaphrase of the student’s statement.
Example 1
Participant: But Iknow that when Igo away, some of them solve the problem, and then
Icome back and say yes, great, let’s go on, or [Isay] Ithink we could doit abit die-
rently, abit better or its not supposed to be this way.
Researcher: Yeah, yeah, so actually you say it something like you don’t say, yeah, this is
wrong but let’s try it like...
In this interview, the researchers paraphrase introduces apotentially new topic into
the interview by emphasizing the level of feedback in the student’s statement. As is ap-
parent from previous statements, the student particularly emphasizes the inuence of
her presence next to pupils when they have to solve amathematical problem: “without
attracting attention, Igo, for example, to have asip of water and Itry to go away to
make them try on their own..., not the way of providing pupils with feedback.
Category4: Strongly inuencing comments
ese are comments that contain words the informant had not said and that explicitly
introduce acompletely new topic or link into the interview. ese comments contain
an interpretation of the student’s statement. In the rst example, the rst part of the
utterance is asummary of the contents of the informant’s statements so far, (this is
not inuencing because it contained words and semantic links used by the informant.
However, in the second part of the utterance, the researcher has interpreted the stu-
dent’s statement. e researcher thus created anew semantic link with an unexpected
situation and conrmation of the teacher role.
Researcher: We talked about unexpected situations, about situations which throw you
o your teacher role and return you to the other one. Now actually, in turn, again an
unexpected situation which reassured you in that role.
In the second example, it is aform of evaluation of the student’s statement and apres-
entation of the interviewer’s own opinion. Both the evaluation and the opinion intro-
duce anew semantic perspective, new links, into the interview.
Participant: … better if they admit they don’t understand it, and they doadmit that in
the seventh gradethen Itried to explain that further or explain it in adierent
way. Which Ithink is probably better, but thats the seventh grade, not sixth grade.
Researcher: Hmm, never mind, it’s in general like that, Ithink that also in the sixth class,
even if this happened, it would probably have the same course.
5.4.4 Discusion of the categorization of comments
Paradoxically our divergence from the traditional concept of Clean Language inter-
viewing in research, which deliberately does not work with comments, is most evident
in the comments. Categories 1a and 1c could seem undesirable from the perspective of
traditional Clean Language interviewing, while 1b best corresponds to the philosophy
of Clean Language interviewing.
e reason is that Category 1b uses only the rst two phases of the syntax of Clean
Language interviewing—the third part is not used (for details on syntax see Chapter 5.1
Conducting interviews using Clean Language). By including only two phases of syntax,
the question is missing and the repetition is only adeclarative sentence, i.e. acom-
ment. e comment does not include anything that would appear to contaminate the
respondents statement, but, on the contrary (from our experience), this repetition
reinforces the informant’s “immersion” in their own experience. is technique is
commonly used in other methods of interviewing close to the rst-person position
(cf. Gendlin, 2004 or Petitmengin & Bitbol, 2009). e purpose of most of the other
comments in Categories 1 and 2 is either to reinforce the relationship between the
informant and the researcher or to frame or pre-frame the space for the research in-
terview. is contributes to creating an atmosphere of trust and a secure interview
environment. Understandably, these comments can be perceived to have asuggestive
form but they are suggestive in relation to the process, not the content of the interview.
Categories 3 and 4 are comments we can label as undesirable in the context of an
interview because they unnecessarily stie topics brought up by the informant.
In summary, we can say that positively inuencing comments help obtain data from
aposition close to the rst person. Neutral comments help maintain the research inter-
view process in desirable dynamics. Inuencing and strongly inuencing comments
are undesirable in an interview because they have the potential to alter the focus of
the interview so that it is not in harmony with the informant’s previous statement(s).
5.5 Quantitative analysis
In this section we proceed to the quantication of the data analyzed. Researchers 1 and 2
conducted half of the research interviews (22 out of 44) asking on average 82% completely
clean’ questions (Category1) in arandomly selected interview. Researcher3 conducted
about aquarter of the interviews, but used n only 24% of all questions asked in the
interview analyzed were Category 1 clean’ questions. Arepresentative of the researchers
who conducted about aquarter of the research interviews combined (Researcher4) used
Category1 questions in 57% of cases (Table5, Figure6). With some degree of bias, it
may be deduced that these statistics also represent the level of clean-ness evident in the
clean language used in the research interviews that were not analyzed.
Table 5
e frequency of questions in each category of clean-ness in the interviews analyzed (absolute
Clean-ness rating of questions
researcher 1
+ Karel
researcher 2 +
researcher 3
+ Františka
researcher 4
+ Ema
Category 1:
Classically clean 57 65 20 66
Category 2:
Contextually clean 11988
Category 3:
Mildly leading 1 5 22 32
Category 4:
Strongly leading 0 1 32 10
Total 69 80 82 116
Interview length 80 minutes 95 minutes 85 minutes 77 minutes
Interview: researcher 1 Interview: researcher 2 Interview: researcher 3 Interview: researcher 4
83% 81%
11% 10% 7%
27% 28%
0% 1%
ategor 1: Classically clean ategor 2: Contextually clean
 3: Midly leading
4: Strongly leading
+ Karel
+ Ema + Františka + Ema
Figure 6. e frequency of questions in each category of clean-ness in the interviews
analyzed (percentage)
As far as the analyzed comments are concerned, Table 6 and Figure 7 show that
Researcher3 and Researcher 4 had ahigher number of comments in Categories 3
and 4 than the rst two researchers, who are more experienced in conducting Clean
Language interviewing.
Table 6
e frequency of comments in each category of clean-ness in the interviews analyzed (abso-
lute numbers)
Clean-ness rating of comments
researcher 1
+ Karel
researcher 2
+ Ema
researcher 3
+ Františka
researcher 4
+ Ema
Category 1:
Positively inuencing comments 18 38 26 34
Category 2:
Context-bound comments 12 15 8 22
Category 3:
Inuencing comments 2 0 15 15
Category 4:
Strongly inuencing comments 0 0 26 15
Total 32 53 75 86
Interview length 80 minutes 95 minutes 85 minutes 77 minutes
Interview: researcher 1 Interview: researcher 2 Interview: researcher 3 Interview: researcher 4
20.0% 17.4%
Category 1: Positively influencing comments
Category 2: Context-bound comments
Category 3: Influencing comments
Category 4: Strongly influencing comments
+ Karel (Participant 1)
+ Ema (Participant 2) + Františka (Participant 3) + Ema (Participant 4)
Figure 7. e frequency of comments in each category of clean-ness in the interviews
analyzed (percentage)
5.5.1 Discussion of the quantitative analysis
Overall, it can be said that Researchers 1 and 2 not only used more Category 1
questions but also signicantly fewer Category 3 and 4 comments. On the contrary,
Researchers 3 and 4 used cleanquestions less and used signicantly more Category
3 and 4 comments. We interpret these results in relation to the level of experience
of the researchers in applying clean language to interviews. Researchers 1 and 2
have more intensive experience with the use of clean language, not only in research
interviews but also in coaching and therapeutic interviews, interviews focused on
reective practice. is leads us to the conclusion that, in order to master the applica-
tion of clean language in research interviews, training for researchers is necessary.
is should include not only developing an understanding how the method works
but also repeated practice in asking questions, supported by feedback from amore
experienced practitioner. It can be assumed that more intense training will also lead
to areduction in the number of Category 3 and 4 comments. For researchers not
suciently familiar with amethod, it is important to concentrate when conducting
interviews on the exact process of asking questions. is may detract the researcher
from the interviewee’s lived experience and, at the same time, the informant’s atten-
tion may be detracted from the topic reected on.
So how... Idon’t know how to articulate this but what does it look like?
Well, um, what would that be, um, Idon’t know what to call it, Idon’t want to call it
indents or something similar, simply what is that you’d like to achieve, like this?
5.6 Conclusion
Overall, there are anumber of conclusions that can be drawn from our evaluation of
the use of Clean Language interviewing for data collection in research.
We dened the categories and sub-categories of clean-ness (1 to 4) for the questions
asked. Further research would be necessary, which would also proceed inductively,
and which could independently identify more categories. Such categories could be
subsequently compared, which could lead to arenement of the categories. e
rened categories could then be used for deductive coding of interviews for the
purposes of evaluating the interviews conducted.
We dened the categories and sub-categories of clean-ness (1 to 4) for the com-
ments used.
We discussed the nature of the categories of questions and comments in relation
to other interviewing methods that are also close to the rst-person perspective in
We quantied each category of questions and comments for four interviews to gain
an overview of the extent to which we were able to adhere to clean-ness in data
collection. When interviewing using clean language, less experienced researchers
had not only alower number of Category1 clean questions but also asignicantly
higher number of Category 3 and 4 comments than those researchers who had un-
dergone longer training.
A seemingly trivial nding, yet important from our perspective, revealed by the
quantitative analysis t was that the quality of an interview (and thus also the expec-
ted quality of data) depends on the level of experience of the researchers.
Auseful extension of the results of the evaluation could be the use of conversational
analysis. is could help us nd out how questions and comments inuence inform-
ant statements. For example, we judge from subjective observations that informants
are being educated while being interviewed. Researchers repeatedly experienced that,
aer acertain number of interviews, informants were able to predict what question
the researcher would ask. ey would oen ask and answer the question themselves.
We assume that this experience goes hand in hand with becoming more sensitive to
descriptions and the reection of one’s own experience. is could be seen as anatural
eect of long-term use of Clean Language interviewing. In the context of the examina-
tion of the experience of human consciousness, in some interview methods, for ex-
ample, descriptive experience sampling (Hulburt, 2011), informants are trained to be
able to capture their inner experience. is is an inspirational idea because it not only
educates the researcher but also the participant in how to approach their experience.
It seems that Clean Language is indirectly responsible for this.
Our experience shows that, at the beginning of an interview, ‘clean questions are
perceived as unnatural by the informants (they are puzzled by what they perceive as
‘strange’ questions). e informants then tend not to focus on the content of their
experience but instead comment on the actual question (is oen happened with the
question “What kind of X is that X?”).
However, in our concept of Clean Language interviewing, the clean-ness of aquestion
is not the same as the naturalness of aquestion. If we equated clean-ness with natural-
ness, the informant could perceive some questions as clean’ even though they contain
many assumptions. It is not important for aresearcher to explain to the informant
which question is clean and which is not. In the interview process, the informant
gradually (even subconsciously) learns the interviewing logic and becomes more sen-
sitive to their own experience. As discussed above, this also brings us, the researchers,
closer to the relational and contextual concept of clean-ness in an interview.
... The CLI method can be delimited with a certain degree of reduction to three rules (Nehyba and Svojanovský, 2017): ...
... Responses to leading questions can be removed from further analysis if the interviewee data are deemed to 'misconstrue their experience' (Froese, Gould and Seth, 2011, p. 47). Nehyba and Svojanovský (2017) compared the Cleanness Rating of four interviewers: two had gone through an intensive three-day training course and two had attended only a four-hour workshop in the CLI method. Their results showed that the better trained interviewers achieved ratings of 92% and 96% in the combined Classically and Contextually Clean categories, while the two less trained interviewers scored 34% and 64%. ...
This article reports on Clean Language Interviewing (CLI), a rigorous, recently developed 'content-empty' (non-leading) approach to second-person interviewing in the science of consciousness. Also presented is a new systematic third-person method of validation that evaluates the questions and other verbal interventions by the interviewer to produce an adherence-to-method or 'cleanness' rating. A review of 19 interviews from five research studies provides a benchmark for interviewers seeking to minimize leading questions. The inter-rater reliability analysis demonstrates substantial agreement among raters with an average intraclass correlation coefficient of 0.72 (95% CI). We propose that this method of validation is applicable not only to CLI but to second-person interviews more generally.
Interviewing is the most frequently used qualitative research method for gathering data. Although interviews vary across different epistemological perspectives, questions are central to all interviewing genres. This article focuses on the potential for the wording of interview questions to lead and unduly influence, or bias, the interviewee’s responses. This underacknowledged phenomenon affects the trustworthiness of findings and has implications for knowledge claims made by researchers, particularly in research that aims to elicit interviewees’ subjective experience. We highlight the problem of the influence of interview questions on data; provide a typology of how interview questions can lead responses; and present a method, the “cleanness rating,” that facilitates reflexivity by enabling researchers to review and assess the influence of their interview questions. This clarifies the researcher’s role in the production of interview data and contributes to methodological transparency.
ResearchGate has not been able to resolve any references for this publication.