Analysing written narratives: considerations on the ‘code-totality problems’

This article describes and discusses a number of fundamental aspects of analysing short written narratives. Of particular interest are the code-totality problems that arise during the transformation of several individual stories into a collective narrative. This article starts with a brief introduction to our previous narrative research on Swedish social-work students, which is followed by a description of textual interpretation according to Paul Ricœur’s theory of interpretation, and a discussion and elaboration on the different concepts of meaning within his theory. The core of this article is an account of four models for analysing narrative data from several informants. This is followed by a concrete example of the implications that follow from the implementation of these models. We conclude that the degree of heterogeneity in the narrative material affects the choice of the mode of textual analysis and the code compilation.


Introduction
The qualitative analysis of texts often implies some form of coding of text units, with the purposes to reduce, categorize or find the meaning in the material (Miles & Huberman, 1994;Robson, 2001).For example, in Grounded Theory, coding in several steps (open, axial, and selective) is necessary in order to build theory about the phenomenon under study (Strauss & Corbin, 1998).In narrative analyses, which this article employs, coding is a necessary step in order to uncover the meaning in a text (Riessman, 1993(Riessman, , 2004(Riessman, , 2008)).
The literature on qualitative analysis often discusses different aspects of the coding procedure in rather great detail, and is frequently based on the principles in Grounded Theory (Strauss and Corbin, 1998).Two common topics are how coding can reduce empirical material, and how codes can be created, sorted, and theoretically amalgamated (Bryman, 2001;Miles & Huberman, 1994;Robson, 2001).It is our experience that the literature less commonly addresses closely what we regard as the 'code-totality problems', and especially in relation to narrative analysis.We argue that these problems arise when the researcher 'transforms' several individual stories into a collective narrative during the coding process.
We can identify these problems by posing the following questions: 1) When during the process of analysis should a material be coded?2) Should codes be treated as unique singularities or as aggregations into more generic codepackages, and if so, how? 3) How can codes (single or aggregated) be connected to overall interpretations?4) What kinds of analytical totalities emerge in the final stage in narrative analysis?5) In what way are the answers of these questions dependent on how heterogeneous the empirical material is? Different answers to these questions will generate different qualities, and if a researcher is not aware of the choices that are made in the coding process he or she will not be able to contrast interpretations against alternatives.
These are the queries that we have dealt with in previous studies where we have investigated Swedish social-work students at Umeå University, and in this article we want to share some of our experiences.We do not present the final answers to these questions, but we present a number of principles and examples that we hope can be helpful to others who might be confronted by code-totality problems.
With the aim of contextualizing our discussion, this article starts with a short introduction to our previous narrative research, followed by a description of textual interpretation according to the French philosopher Paul Ricoeur.We also discuss the different concepts of meaning which form a fundamental but sometimes ignored part of his theory.The major part of this article consists of an account of the four principal ways of analysing narrative data that we have constructed by utilizing Ricoeur's methodological principles.This description is followed by a section in which we present a concrete example of the implications that may result from the usage of these models.Finally, we summarize some of the main points that we wish to pass on to the reader.

Analysis of short written narratives -a brief background
One point of departure for narrative analysis is that when we narrate, we put forward a message about the understanding or meaning we receive from (or attribute to) our experiences.An essential assumption within narrative analysis is that storytelling serves the purpose of creating meaning from one's lived experiences (Atkinson 1997;Czarniawska 2004, Riessman, 2008).The narrative method is particularly advantageous for research of activities where humans work with humans on a social, and thus abstract, level because mutual understanding becomes a central part of the result (e.g., Riessman & Quinney, 2005;Salander, 2002).Therefore, we assumed that an analysis of social-work students' stories about such situations provided a relevant point of departure.The students' stories have been used in studies on narrative analysis where we have focused on the meaning of critical situations during field studies as well as on the use of knowledge in social-work practice (e.g., in Blom, 2009;Blom, Nygren, Nyman & Scheid, 2007;Nygren & Blom, 2001).
The authors of this article were previously involved in the follow-up of field studies at the social-work programme at Umeå University in northern Sweden.
Social-work students in their seventh and last semester of their education (which then was preceded by 20 weeks of field studies that included training for working directly with clients) were requested to write a story.We asked the students to write down a 'narrative', a compressed story, where they described a critical situation from their practical training.We also asked them to reflect on the use of knowledge in relation to these situations.A shortened version of Sven's (false name) narrative offers an example of these short narratives.
One of the patients I regularly have contact with comes to a dayactivity centre where we have decided to meet.He is in a miserable shape: hollow-eyed, in a cold sweat, and filled with anguish.… He tells me of, for him very important, the beginning of school the night before, that according to him ended up in a catastrophe.The man is heartbroken; words and tears are flowing out of him, and for a few moments he loses his connection to reality.During the first part of the conversation I am almost completely quiet, only asking some brief questions.After he has ended his story, I take a more active role, where the purpose is, to some extent to try to tone down the un-controlled negative thoughts that govern him.Maybe his failure could be relativized.… … When it comes to the types of knowledge that I used, it is difficult to exclude any of the forms that the teacher of this course talked about.To say that facts to a great extent governed my acting feels, in a way, rather futile.Nevertheless, there is always a 'bank of knowledge', containing facts, which you … often have to start from.In this case, information from countless, more or less obscure, books on psychology and psychiatry concerning symptoms and other things surely was part of the picture.Moving towards a micro level, one might say that my knowledge of where the 'conversation rooms' are situated at the clinic partly forms the situation.My understanding is, to some extent, formed by my facts.The situation with a human in a state of breakdown, I would probably regard as frightening, if I were not to have any idea about why the person would be in that state.Here I believe that understanding (identification) and my ambition to understand are present in, for example, my questions, during the conversation.
Skill can in this case be about the choices I make during the situation, as well as the forms and the techniques around and in the conversation -for example, to choose a secluded place, not to govern the conversation, not to finish after an hour, and not to moralize and simplify.At the same time I do not recall anything that might be conceived as craftsmanship.
Familiarity might be a part of the picture.-Maybe earlier experiences from meetings with a number of persons in crisis are such a source of knowledge.The social-work education, especially the relational approach it conveyed, in combination with Christian values, could be another source.The psychodynamic frame of interpretation that I've been dragging around with me during my practice term, and that probably influenced my way of apprehending, is naturally from the literature (education, other courses, own reading), but also from influences from the supervisors and colleagues I've been in contact with.
The majority of the social-work students wrote narratives that were one-and-ahalf pages long, but the length of the material ranged between one and two pages.The resultant 144 stories were used as a pedagogical instrument, as a foundation for verbal presentation and discussion in the class, and as material for research purposes.They were analysed with a method of analysis that we have previously developed (Nygren & Blom, 2001).The method is based on Paul Ricoeur's theory of interpretation (Ricoeur 1976;1981), but also inspired by a method for research on lived experiences that was developed within nursing research (Lindseth & Norberg, 2004).

Examples of analysis and presentation of results
Chart 1 below illustrates how we coded the empirical material and presented the output of the coding.The example is built upon our experiences from working according to model 3, which is described later on in the article.Informal knowledge also important ---Among the 144 narratives that were analysed in the main study (Blom, Nygren, Nyman & Scheid, 2007;Nyman & Scheid, 2004), we could identify a number of different strategies concerning the students' use of knowledge in critical situations.These were merged and restructured in two new -more concentrated and chronological -narratives that according to Ricoeur is the final step of the narrative analysis.
2 These new narratives helped us to reach a more general and deepened understanding (comprehension).We have chosen to call one of these The opening of a locked situation, and we present a shortened version as an illustration.
The student (who is called Doris) in this story is preparing herself meticulously before meeting the client.She reads case records and statements about the client (here called Tom), whom she is about to meet.Doris therefore has a rather fixed opinion about Tom's life-situation beforehand.On the basis of all the information that she gathered before the meeting, she also has a clear idea of the appropriate intervention in this case.Nevertheless, Doris is quite nervous before the meeting, and she repeats to herself what she wants to achieve during the meeting.Nevertheless, she feels rather awkward and appears as though she is groping when she arrives to meet Tom.Doris enters the meeting with intention to solve Tom's problem quickly by suggesting the intervention she considers appropriate.However, the knowledge she brings with her leads to a cognitive and emotional blockage.She is stuck in her predefined conception about Tom.In fact, she does not bother to listen to his opinion of the situation.But suddenly, when Doris' previous knowledge does not seem to help during the meeting, she becomes very frustrated!And this frustration forces here to reconsider the situation, and new ideas arise.During the meeting she is confronted with the idea that she might be on the wrong track, because she has not been listening to Tom.Doris realizes that, her preparations and fixed conceptions are hindrances in this particular situation.
At this point Doris feels that she has partly loosened her grip of the situation.After a moment of confusion she has chosen to change her strategy: she now tries to disregard her own previous knowledge and her prejudices.
Afterwards, when Doris is reflecting upon this meeting, she feels a sense of satisfaction over her newly won insights.She has managed to get around the barrier created by her prejudices, and Doris thinks that from now on she will try to be more open-minded meetings with clients.
We believe that this new narrative is a representative example of the comprehension of the material as a whole, which was obtained after combining the previous naïve understanding and the structural analysis (according to model 3).In other words, we had reached a better understanding of our own social-work students' use of knowledge.
In one of our studies of the research method (Nygren & Blom, 2001), we concluded that there are several reasons for refining and developing further all the steps in the analytical procedure.This concerns, among other things, more consciously made processes in the steps from naïve reading to depth interpretation.And this is essentially what we shall deal with in this article.
2 Concepts and analytical principles are discussed further on in the article.

Textual interpretation according to Paul Ricoeur
The following sections present a detailed discussion about textual interpretation based on Ricoeur's theory of interpretation.A rather extensive explanation of the basic concepts and ideas is necessary, so that our later discussion about models and coding is comprehensible.Ricoeur (1976Ricoeur ( , 1981) ) argues that we can adopt two possible attitudes before reading a text: understanding or explaining.Understanding is about grasping or getting a feel for the whole chain of seemingly fragmented meanings in a merged manner; in other words, this attitude involves finding the meaning of the text.Explaining a narrative is the sorting out of the fugue-like organization of interlaced actions, its skein of movements.In other words, explaining is about uncovering the internal relations of the text through structural analysis.
Ricoeur considered explanation as well as understanding to be encompassed in the superior concept of interpretation.Put differently, interpretation is the dialectic between understanding and explanation.In this manner, understanding precedes, accompanies, and encloses the explanation.Conversely, the explanation develops the understanding analytically.
According to Ricoeur, the dialectic between explanation and understanding is an initial movement from understanding to explanation, followed by a movement from explanation to comprehension, which is a more sophisticated mode of understanding.Initially, understanding is a naïve (i.e., open, without prejudice) grasping of the meaning of the text as a whole.It is virtually a guess at the meaning in the text.Later on in the analysis the understanding (as comprehension) is more advanced.In between the two modes of understanding is the indispensable explanatory stage of structural analysis.Finally, the interpretation might reach what Ricoeur denominates as appropriation (Ricoeur, 1976;1981).These concepts are discussed more thoroughly later in the article.
We believe that Ricoeur presents convincing arguments for the importance of the dialectic between explanation and understanding with respect to ontology, epistemology, as well as methodology.On the basis of his theory we have developed a way of analysing narrative data where the different moments of the interpretative process are under control.Hence, both the analyst and the readers of an analysis can observe, quite explicitly, how the analysis relates to the original text.

Four different concepts of meaning in Ricoeur's theory
One question that it is necessary to ask when doing narrative analysis is what meaning is.Ogden andRichards (1989/1923) have written a book with the accurate title, The meaning of meaning, that presents more than 20 different definitions of the concept.Meaning can, by way of example, mean an intrinsic property, an essence, an event intended, the connotation of a word, and that to which a symbol refers.These examples demonstrate that it is all but selfevident how the concept of meaning should be defined, and lead us to understand that there is an obvious risk of misunderstanding when the goal is to uncover the meaning in texts.Analysts will quickly find themselves confronted with two basic queries: What is meaning?And, For whom is a meaning a meaning?According to Ricoeur (1976), the meaning of a text can be divided into four different concepts: one that he calls the utterer's meaning (utterer = the one that writes or says something), and three different types of utterance meaning (utterance = that what is written or said).We want to emphasize that these different types of meaning are not explicitly stated in Ricoeur's works exactly as we present them.Consequently, what we write are our interpretations of his somewhat muddled discussions on the subject.Our main sources are his books Interpretation Theory (1976) and Hermeneutics and the Human Sciences (1981).Below, we provide a short definition and a number of illustrative examples of each concept.

Utterer's meaning
This concept of meaning refers to the author's own explicit intention, purpose, motive, et cetera.Ricoeur writes that this is the subjective side of meaning.For example, with his book, Das Kapital (Capital, 1867), Karl Marx wanted to criticize the political economy of the nineteenth century, and Sigmund Freud's intention with the book Die Traumdeutung (The Interpretation of Dreams, 1900) was to show that it was possible to interpret and understand dreams.This way of defining meaning, which draws upon the author's conscious and often explicit intention, is what we usually connect with written texts.
Utterance meaning 1 (semantic meaning).Here, the autonomous meaning of the text, that what is said, is in focus.It is not a question of the author's intended meaning.The semantic meaning concerns the comprehensive and communicative aspects of language.An example of this type of meaning is to understand that the utterance 'wright' -write, right, rite to the right! is not a consequence of stuttering, but a concentrated request to a skilled worker to write down the correct ceremonial practices on a specific part of a paper.In other words, the semantic meaning is about the relation between the language symbols and what they denote.

Utterance meaning 2 (reference backwards or behind). This type of meaning refers to what the text talks about:
what exists, irrespective of the author's intention with the text.For example, one can read the books about Pippi Longstocking as entertaining stories about a very strong little girl with strange clothes, who lives without parents but instead with a monkey and a horse and so on.However, her books also have a historical reference, which, among other things, would include the cultural values that prevailed in Sweden, and many other countries, around 1946 when Astrid Lindgren wrote the first book about Pippi.The comical (and even shocking for some adults at that time) about a nine-year-old girl living on her own in a big house, being self-assertive and not afraid of adults and so on, should be understood against the background that children in the 1940s usually did not behave that way.Children were expected to be like Pippi's friends Tommy and Annika: obedient, disciplined, clean and well dressed.The point is that all texts are created at a certain time and under certain circumstances, thus one can say that, for example, Hamlet, Don Quixote, and Frankenstein, besides conveying their respective authors' intended message, also (unintentionally) refer backwards and tell us something about the conditions, norms, values, fears and joys, and so on, at the time and place that the text was written.
Utterance meaning 3 (reference forward or ahead).The meaning in this sense refers to the future possibilities that dwell in the text, that is, to that which can become: propositions or possibilities.The text is regarded as a medium for the reader to understand himself or herself and to develop, and, as with utterance meaning 2, it is a question of the meaning that exists in the text irrespective of what the author intended.For example, the books about Pippi Longstocking can be read as amusing stories for children.Furthermore, it is reasonable to believe that this was Astrid Lindgren's conscious purpose.However, it is quite possible that an adult person could also read the books and extract some guidance for his or her life.For instance, a person with a great deal of money and power could begin to consider that he or she -just like Pippi declares -should be generous and kind because of the possession of great resources.Therefore, what a text means for somebody's future, depends on the reader's individual characteristics and traits.

Key concepts reconsidered
According to our understanding, there are four basic alternatives (models) that can be used to analyse the kind of story we discuss.These models are described in the subsequent section.However, before describing the models, it is important to explain that we do use some of the key concepts in a manner that diverges from Ricoeur's definitions.
Our reconsideration of some of Ricoeur's theoretical concepts is partly an effect of our research itself, which is based on other types of empirical material (several short texts, instead of Ricoeur's singular long texts), and partly because we have developed Ricoeur's approach somewhat.For example, we think that it is important to make a distinction between the processes and the products of the analysis, and to distinguish between different kinds of analytical results, namely, material and cognitive products.These differences necessitate a demonstration of how we intend to use the concepts in connection with the models.
Naïve reading is an almost entirely cognitive process, which ends up in a naïve understanding of the text as a whole, and can be described as a changed cognitive state (i.e., one conceives something in a new way).Ricoeur describes it as 'a naive grasping of the meaning of the text as a whole ' (1976, p. 74).The naïve understanding of the text as a whole can also be documented and thereby handled as a material product (usually as a text).Metaphorically, naïve reading is like trying to grasp a superficial idea of the different motifs in a children's cube-puzzle (usually a puzzle with twelve cubes, and six motifs) without having seen the motifs on the front of the puzzle box.In such a case, one does not know what is 'hidden in' the cubes, only that there are six different possibilities.We assert that, from an ontological perspective, it is a realistic assumption that a text can have one or several meanings that are possible to unveil.
Structural analysis is something of an analytical artefact.Texts are deconstructed, restructured, and analysed, which results in explanations in terms of conceptual codes that, for example, answer questions about what, how, and who, and hence, mainly a material product (usually a text).However, this explanation also has a cognitive element because both the analyst and the person who reads the explanation must understand it.Therefore, the semantic meaning of the text (utterance meaning 1) is in focus.A structural analysis can resemble an extensive mathematical calculation.For example, it is made and 'stored' on a paper or a white board, but human reason is needed to carry out or evaluate the calculation.The point is that the structural analysis is a concrete artefact as well as an abstract cognitive procedure.
Comprehension.Ricoeur writes that the final stage of the analysis is a new story, and he describes the dialectical process of interpretation: 'first as a move from understanding to explaining and then as a move from explanation to comprehension.… comprehension will be a sophisticated mode of understanding… At the end… [understanding] satisfies the concept of appropriation.. (Ricoeur, 1976, p. 74).It is possible to read this as if comprehension and appropriation were almost identical, but we claim that there are methodological and epistemological reasons to differentiate between comprehension and appropriation.They are, in fact, two dissimilar results of the analysis, and it is important to pay attention to this point in the course of planning and conducting analyses as well as when reading the results.
Comprehension is a way to grasp the meaning in the empirical material in the form of a new text (both a material and a cognitive product).Appropriation, on the other hand, is only a cognitive product.Furthermore, we cannot assume that these two forms of understanding automatically coincide because a reader will not always appropriate a text (i.e., make it one's own, become changed, and discover new possibilities) even though he or she comprehends (i.e., understands, mentally grasps) it.In the kind of analysis we propose, comprehension is a reference backwards, that is, it shows what exists, how something is (utterance meaning 2).Moreover, it is a material product (usually a written document) that has two main components: one or several naïve understandings of the text or texts as a whole, and a number of conceptual codes from one or more structural analyses.The integration of these components implies an intellectual process as well as an artefactual one.In practice, it usually means that texts from different documents (1.notes on naïve understanding of the texts, and 2. 'lists' with conceptual codes) are reflected on and put together to form a new, more qualified and logical narrative.
Appropriation is something that we regard as the solely cognitive (but sometimes also emotional) end product of the analysis.As mentioned above, we distinguish between appropriation and comprehension, because they do not obviously coincide.Appropriation means a developed understanding (discovered future possibilities) from the empirical material, namely, the original stories.This kind of understanding can emerge when somebody reads the comprehension (the material product) of the analysis, that is, one or several new stories.The reader can also be someone other than the person who conducted the analysis.
3 Consequently, a human subject is necessary -a reader -if appropriation is to emerge.A new story in itself (comprehension) does not automatically imply increased understanding.In other words, there is a difference between information and knowledge, which means that a text is only information until the reader puts it in his or her own context and makes something of it (cf.Liedman, 2002).Subsequently, it is possible to assume that appropriation does not always occur.
Another metaphor might also clarify the differences between comprehension and appropriation: it is possible to fit the parts of a puzzle correctly (which is analogue to comprehension) without understanding what the motif of the puzzle means in a deeper sense (analogue to appropriation).Our point is that an analysis can be done quite instrumentally without bringing a deeper insight of the results to the researcher or another reader.
As a consequence, a text's reference forward or ahead -ideas of what can come to pass (utterance meaning 3) -is not ready-made and available to be picked up from the comprehension, like taking a pearl out of a mussel.In the way we understand this process, future possibilities only emerge when someone has absorbed the comprehension and has become affected by it cognitively and often emotionally.The degree of appropriation will consequently vary between different readers.The researcher who conducted an analysis might receive a deeper understanding than a reader who 'only' reads the analysis and the results (the comprehension) as a new story.
However, as already mentioned, it could be different because it all depends on how the reader interprets the comprehension.

Four principal ways of analysing short narratives
In the following section we describe the four basic ways of analysing short narratives that follow from our way of using Ricoeur's theory of interpretation.We shall also discuss what kinds of empirical material that each model is suited to analyse.Our survey shows the importance of bearing in mind that the choice of analytical model is very much a question of matching specific empirical material, and that different models lead to different results.

Model 1
Each story is naïvely read, which leads to as many naïve understandings (in the form of memoranda) of the text as a whole, as there are stories.Thereafter, every story is structure-analysed which generates X number of codes from each story.In the subsequent step, the naïve understandings (memoranda) and the codes from each story are merged into a comprehensive understanding of each story.Next, the comprehensive understandings of all the stories are merged together to an aggregated comprehension of the material as a whole, and this could, in turn, form the basis for an eventual appropriation.This way of analysing narratives can be adequate if the material is very heterogeneous, that is, if the stories differ from each other despite a common main theme.By way of example it is possible to imagine that within the theme, My greatest moment in life, it would be possible to write narratives about giving birth to a child, winning a billion dollars, or catching the largest salmon.Contextual contingencies and the events are in such cases so dissimilar that the narratives are only logically consistent within themselves.A reasonable connection between them is only possible to make at the end of the analysis, when comprehensions from each different narrative are merged into an aggregated comprehension.

Model 2
Every story is naïvely read and the results of all naïve readings (memoranda) are put together as a naïve understanding of the material as a whole.Then a structural analysis is conducted which generates X number of codes from each story.Next, the codes from each story are merged with the preceding naïve understanding of the material as a whole into a comprehension of each story (this is symbolized by a grey frame in the figure below).All the comprehensions are then amalgamated to an aggregated comprehension of the material as a whole, which makes a foundation for appropriation.This mode of analysing can be relevant if the narratives are somewhat more similar than in the previous example.In order to obtain a useful naïve understanding of the material as a whole, it is reasonable that the stories are fairly congruent in some structural sense, for instance, that they are about roughly the same things (e.g., a certain type of activity), or that the events occur within similar contextual conditions (e.g., in a certain environment).This model is, like the previous one, mainly suitable for analysing one story at a time, and thus aggregating at the end of the process.Narratives can be highly diverse even if there is an overarching main theme (e.g., social work in the 21st century).It is easy to imagine that stories within such a theme, despite common points of contact (e.g., social work with adults), are so dissimilar that it is most reasonable to conduct a separate structural analysis and comprehensive interpretation for each story at a time.This would, for instance, be true for stories about handling financial benefits to young adults, psychotherapy to older drug addicts, and counselling to abused women.

Model 3
With this model, each story is naïvely read, and memoranda from all of these naïve readings are compiled to a naïve understanding of the material as a whole.In the subsequent step, each story is structurally analysed.The codes from the structural analysis of all the stories are then put together into a totality.Next, the totality of codes and the naïve understanding of the whole are merged into a comprehension of all the stories, which in turn lay the ground for appropriation.This mode of analysis is adequate if one has relatively homogeneous empirical material.In our research based on social-work students' narratives about critical events during field studies, that was the case.Even though all the narratives were in some respect unique, there were many common denominators (types of clients, types of interventions, types of knowledge, etc.) that made it possible to compile a naïve understanding of the material as a whole, relatively trouble-free.However, the narratives also contained numerous unique components (descriptions of places, individuals, feelings etc.) that only made it possible to analyse structurally the stories one by one.In other words, each narrative's building blocks -words, concepts, sentences -were only possible to explain in relation to each story as a whole.Nonetheless, thanks to fundamental similarities in the stories (e.g., concerning strategies, use of knowledge, reflections), it was possible to put together the codes from the structural analysis into a rather consistent 'totality'.

Model 4
With model 4, each story is naïvely read, and memoranda from all the naïve readings are compiled to a naïve understanding of the texts as a whole.Then all the stories are put together to form a 'grand' story that is structurally analysed.In the next step, the codes from the structural analysis and memoranda from the overall naïve understanding are merged into a comprehensive understanding of the texts as a whole.This provides a foundation for somebody's eventual appropriation of the material.The fourth model can be relevant if someone wants to analyse material where all the respondents or informants have experienced exactly the same event, for instance, an accident in which many persons were involved.As the respondents or informants share the context and course of events (though individual experiences obviously vary), each person's story can be regarded as a part of a bigger narrative about the event.

Example of an analysis based on model 1
As a means of illustrating the analytical problem that this article focuses onthe code-totality problems -this section presents a short example of an analysis based on model 1 above.The analysis comprises three stories within the theme of what characterizes a survivor.The stories have this question in common, which makes it reasonable to include them in the same analysis.At the same time they are so heterogeneous that they can only be analysed according to the first of the four models.As an empirical starting point we have chosen three stories (Odysseus, Cinderella, and Robinson Crusoe) that we can assume are relatively well known, which makes it unnecessary to present the whole stories in the article.Moreover, these stories are easily found in books and on the Internet.
It is important to clarify that the stories are chosen for quite pragmatic reasons, that is, as a means of exemplifying the code-totality problems in a fairly condensed way.We are aware of that there are numerous versions of these stories, and that it is possible to interpret them in several ways.By way of example, it is not self-evident that Cinderella is a story about a survivor, but it is possible to regard it as a story about someone being rescued.However, it is not necessary that the reader agrees with our coding and interpretations -the point is not to understand these particular stories better.They are only here to serve as concentrated illustrations of our discussion.
We start the illustrative analysis with a naïve reading, followed by a structural analysis and a comprehension of each single story.Next, we present an aggregated comprehension and finally our appropriation.The stories do have a common theme, but they are too dissimilar to each other to aggregate parts of the stories (e.g., codes or understandings) in a meaningful way before the last stage of the analysis.The complete story about Odysseus can be read in Homer, Knox, and Fagles (2006).

Story 2
Table 2.The first steps in the analysis of the story about Cinderella.The complete story about Cinderella can be read in Perrault (2002).

Who
well as for the reader.The reader should bear in mind that this short example is meant to illustrate the developed understanding that the authors of this article obtained from the analysis.In reality, appropriation is not a text, but a process in somebody's mind.

Appropriation
Everyone fights a lifelong battle against outer and inner forces.One possible lesson from narratives about survivors is that it can be worth fighting on, even when it seems hopeless.Often it is possible to turn a setback into a success.And even if it does not turn out that way, the struggle in itself can be meaningful… Someone who reads our analysis might arrive at a similar insight, but this person might also get a partly or totally different insight.A further possibility is that the analysis does not offer a new insight at all.To a large extent it depends on who the reader is (previous experiences, expectations, motives, etc.) and in what context the analysis is read (during education, for research purposes or by chance, etc.).

Examples of an illogical analysis
Following the demonstration above of how an analysis might look when the empirical material and mode of analysis fit together squarely, we shall now illustrate how it may turn out with an analysis of a heterogeneous material that utilizes an inappropriate model of analysis.This is made by taking a look at the three stories during the second step of analysis (Naïve understanding of the material as a whole), as in model 2 to 4. The example below shows that it is indeed possible to create a story -a rather twisted one -without the kind of logic that is necessary in a scientific context.

Naïve understanding of the material as a whole
Odysseus, Cinderella, and Robinson Crusoe were sailing home when they were hit by a storm created by the enraged Poseidon.They ended up on an island, whose habitants consisted solely of Cinderella's stepfamily.They began to fight over cooking and cleaning, a struggle that lasted for ten years.Robinson and Odysseus fell in love with Cinderella's stepsisters, which at a closer look turned out to be the monster Scylla.The stepmother was a cannibal who was eventually killed by Odysseus before he went to the ball at the castle … From a scientific point of view, the analysis is going down the drain.The stories are too dissimilar to create a reasonable naïve understanding of the material as a whole.Although the example might seem ridiculous, it is nevertheless a possible result if one were to try to merge the naïve readings from the stories in question.Another purpose with this example is to demonstrate that it is inappropriate to use model 3 or 4 for analyses of such heterogeneous material.
If anyone, contrary to our expectations, should succeed in making a naïve understanding of such a heterogeneous material (as above), we assume that the analysis will fail at a later phase.By way of example, the Totality of codes that is the fourth step in model 3 could look something like this:

Totality of codes
Her life is sad and miserable, Disrupted slave-trade voyage, Odysseus' emotions got the upper hand, Lady Luck seems gone, He felt divine, Helped by natives, Her work never ends, Manly independence, Odysseus uses his cunning to manage, A combination of magic and persistence makes luck return, Faces the problem of cultural relativism, Back in the treadmill … As this example shows, this is just a mixed list of codes that can hardly be related to each other in a reasonable way.Therefore, the codes become a bunch of words without logical relations.It is possible to continue exemplifying this way, but we assume that we have made our point by now.

Summing-up
In this article we have discussed important aspects of meanings, models, and coding that we believe are important to consider when analysing written narratives.The starting point was our earlier studies of Swedish social-work students' use of knowledge in critical situations.In relation to four different models for analysing short written narratives, we have discussed the codetotality problems that arise when transforming several individual stories into a collective narrative.As a means of reaching a qualified understanding of the entire empirical material, the focus of discussion was on questions concerning the compilation of codes and aggregate comprehended wholes.
We are aware that the principles that we have discussed do not present a simple or final solution to the problem concerning how narrative data can best be organized in order to uncover different meanings, but we hope that this article can help researchers who plan to perform a narrative analysis to make more informed choices concerning the analytical process in relation to the empirical material.A conclusion that we want to convey to the reader is that the degree of heterogeneity in a narrative material affects the mode of analysing texts and compiling codes.In other words, the extent to which the stories are similar or dissimilar determines to a large degree the type of analysis that it is reasonable to carry out.
Moreover, we believe that there is a certain similarity with a statistical principle used when analysing quantitative data.Many readers may be familiar with the principle that it is not possible to perform all forms of statistical analyses with all types of quantitative material.To a large degree, the forms of analysis that can be carried out depend on the level of measurement to which the material belongs.In other words, the higher level of measurement (nominal, ordinal, interval or ratio scale) the material is on, the more possible it is to perform more sophisticated analyses.
Similarly, a researcher has to consider which mode of analysis can be employed logically in relation to the degree of heterogeneity of the qualitative material.Narrative material that can be analysed with model 4 can also be analysed with models 1 to 3. Further, material that can be analysed with model 3 can be analysed with models 1 and 2 as well, but not with model 4. In other words, it is logically possible to analyse material with a model 'lower' than its goodness of fit, but not with a model 'higher' than its goodness of fit.In spite of the fact that our discussion has been based on a certain type of qualitative material (short narratives) and a certain theory of interpretation (Ricoeur's), we believe that it encompasses a certain amount of generality.Many researchers work with qualitative materials that have basic similarities with our narratives, for example, transcripts based on interviews, observations, diaries, and so on.Furthermore, in the majority of cases the sorting and coding of the materials are necessary, and this would be done in the case of Ricoeur's theory as well as, for example, Grounded Theory.As we understand it, there are parallels between analytical steps in Ricoeur's theory (e.g., naïve reading and structural analysis) and steps in Grounded Theory (e.g., open and selective coding).Consequently, it is reasonable to assume that it is important to pay attention to the code-totality problems when performing other forms of qualitative analyses such as Grounded Theory.
We do not claim that our discussion has solved the code-totality problems once and for all, but we hope that this article will contribute to an increase in methodological consciousness in relation to this rarely discussed question.

Figure 1 .
Figure 1.The first model for analysing narrative material.

Figure 3 .
Figure 3.The third model for analysing narrative material.

Figure 4 .
Figure 4.The fourth model for analysing narrative material 4