Qualitative Text Analysis with MAXQDA Anne M. Kuckartz, Udo Kuckartz This paper is a written version of a presentation given at the CentrA Seminar on advanced computer aided qualitative research, November 2001 in Granada. The main targets of the conference were to get an overview of some of the most advanced software packages supporting the researchers in qualitative data analysis and to link this overview with a debate of some key issues of the strengths and weaknesses of qualitative research. We wanted to keep the character of the conference presentation in the present paper. Therefore most of the original sheets but completed it by explanations and comments are put at the reader’s disposal. The presentation is divided in three parts. In the first part some key points of qualitative research will be marked out before we scope the methodological background of the software tool MAXQDA, follower of winMAX in the second part. The third part is dedicated to a closer look on MAXQDA in practice. SHEET 1 For further information to both software packages see: www.MAXQDA.de and www.winmax.de 1 I The Paradigm of Qualitative Research Qualitative research is part of the answer that social sciences give to their very basic question: SHEET 2 We will recall the complexity of the real world for a minute by looking at the definition Anthony Giddens gives of social life (Giddens 1984:XXIVf.). SHEET 3 2 Within the spectrum of social research methods the Qualitative Research Paradigm claims for a great variety of issues to correspond particularly well to this complexity of the real world and to bring out reliable results for the issues in question. Qualitative Data Analysis takes advantage of the chance to reach an in-depth understanding of the social reality and thus to be able to develop relevant middle-range theories about a particular social phenomenon. The essential gain of reliability in qualitative research is convincingly expressed by Denzin and Lincoln (Denzin/Lincoln2000:3). SHEET 4 3 Let’s again take a look at Denzin and Lincoln, to see some of the weak points of Qualitative Research, weak in the sense of giving special challenges for the researchers (Denzin/Lincoln2000:6). SHEET 5 Doing qualitative research implies an increase of methodological creativity but also an increase of ambiguity and uncertainty. There are no convenient rules and fixed procedures at the researcher’s hand which she or he has doing quantitative research and which make life safe and somehow easy. The main methodological advantage of qualitative research - to provide research results of great reliability - is confronted by a reproach resulting from this very strength. The researcher accumulates a great amount of data (observations/interviews/field notes etc.) and has to reduce and to condensate all these information to a data grounded answer to the research question. The „lack of transparency combined with a tendency towards an “anecdotal approach to the use of data in relations to conclusions or explanations are some of the key points of Qualitative Research criticism“ (Bryman 1988:77, cit. in Silverman 2000:11). 4 The use of software tools is a powerful possibility to handle these weaknesses of qualitative research. The gain of transparency and the possibilities to incorporate also elements of quantitative analysis procedures can be a feasible way to proceed in overwhelming the still existing dualism of quantitative and qualitative research. We will to finish the short view on the methodological basics of qualitative research by a rough sketch of its procedures. Anselm Strauss and Juliet Corbin distinguish three major components of qualitative research which is shown in the following sheet (Strauss/Corbin 1998:11 ff.). SHEET 6 In all three parts of the qualitative research process the computer has become a basic tool. Whatever is the type of data, however they have been created, they are brought into a computer as text, audio or video data. In a more and more self evident and extensive way the computer is also used in the second part, the analytical core of the research process where the interpretation and organisation of the data takes 5 place. It is the great challenge for the researcher in this phase of the research process to preserve the embodiment of the field data in the “real world” of words and/or situations as one of the basic strength of the qualitative research. Reaching this target they have to find the best possible balance between maintaining the context of the data and the unavoidable selection and reduction of data during the analytical process. Looking at the qualitative research methods we can identify two major fields with different sources each: The classical methods which have been developed since the birth of the qualitative paradigm and the new methods, which appeared only a short time ago and are still in the process of elaboration. SHEET 7 SHEET 8 Looking at the procedures which are at the researchers’ disposal we find that the basic ones are suitable for computer aided qualitative data analyses. Again we can differentiate between those assigned to the classical and those assigned to the new methods. 6 SHEET 9 All listed procedures are applied on the data being analysed (e.g. texts). The definition of codes, the writing of memos, and the definition of variables are the operationalizations of these methods. All of these procedures are supported by MAXQDA. II. Methodological Background of MAXQDA The main foundation of MAXQDA is found in the methodological works of Max Weber and Alfred Schütz. They aimed both for a methodologically controlled construction of typologies. Schütz indicated that the cognitive process of typification is a fundamental anthropological technique, which enables us to understand our everyday world as well as to conduct scientific inquiries (cf. Schütz 1962:7 ff.). Abstractions and typifications assist our thinking, consequently the procedures of classification, quantification 7 and constructing typologies are always present on a subliminal level, even if is no formal classification or act of counting, in the mathematical sense, occurs. Max Weber said that the construction of understandable types of action was the central goal of empirical social science. Action types are a connecting link between the hermeneutic method, which tries to understand the single case, and statistics in the social sciences, which focuses on regular or law-like relationships (cf. Weber 1964, Kuckartz 1991). To a certain extent this is the position of unified science: it aims to integrate qualitative and quantitative research methods. However, this does not mean overlooking the differences between the two methodologies, and the different criteria for reliability and validity associated with them. But the methodological aim is to identify patterns in social regularities and to understand them in the sense of controlled Fremdverstehen (understanding the other). Doing qualitative data analyses with MAXQDA supports the researcher in his effort to construct empirical types in the process of a case-oriented analysis. Typification always relates to individuals, i.e. the respondents. The types do not necessarily have to be personal, in the sense of people having certain typical features or character traits, but can also mean types of actions, situations, experiences or attitudes. The overall purpose of the case-oriented- analysis is to classify and, if necessary, quantify the qualitative data or parts of it. In doing so, the ‘case’ (i.e. the respondent in question) forms the referential basis of this method. Both software tools, winMAX and MAXQDA are based on this methodological model: The case oriented quantification. 8 During the process of analysing two levels can be distinguished. The scheme of the first level is presented in the next sheet: SHEET 10 The title “1srt level coding” (SHEET 11) should not be misunderstood in the way that the analysis starts right away with the decontextualization of text segments. Quite the reverse: The first steps in this model are hermeneutic, it parallels to a large extent procedures used in the fields of phenomenology or ethnography (Geertz, 1973). First of all it is essential to make sure about the subjective meaning of a text. The principal of “interpretive understanding”, based on Weber’s concept of social interaction is the underlying theorem. 9 Social interaction is a form of human behavior to which the actors involved attaches subjective meaning and which is related to the behavior of others. The term ‘meaning’ is related to the subject and is not, according to Weber’s definition, ‘any kind of objectively correct or metaphysically explored true meaning’ (Weber 1964:4). ‘Verstehen’ therefore primarily means ‘understanding’ the behavior of those individuals who take part in social interaction. Only on this basis the search for regularities, for ‘understandable types of action’, makes sense. For the same reason the first steps of the case-oriented method are not a segmentation of the texts and a decontextualization of text segments, but an ‘overview’ of single cases. The codes that are developed in the following two methodological phases and then allocated to the text segments can, in this form of analysis, be characterized as thematic categories. They are used to identify themes in the transcribed interviews. Coding is the preliminary step, the necessary prerequisite for the following evaluation of the data, in which first of all text segments of the same category are compared. By means of contrasting comparisons this comparative analysis of themes aims at identifying similarities between individuals, peculiarities of single cases and relationships between categories. Although these coding and retrieval techniques are relatively simple, they consume a lot of time and effort on the part of the researcher. The results initially appear to be rather meager: But the data has been clearly structured and ordered. A close inspection and interpretation of the data can now reveal typical patterns and guide to further questions. The bigger the sample, and therefore the data, the harder it is to identify complex patterns. It is common to conduct 20 or more interviews during a qualitative study, but even with this population size (which is small compared to the sample size in mainstream social research) it becomes increasingly difficult to determine whether the apparent or intuitively observed relationships, which are often labeled with quasistatistical terms (typical, mainly, pattern, orientation pattern, model, leitmotiv), are indeed verified by the data. The evaluation method of case-oriented quantification addresses this problem. By means of careful classification of the data and by the rating of individual statements on scales the aim is, on the one hand, to extract the maximum amount of information from the data and, on the other hand, to identify complex relationships on the basis of these newly constructed variables. 10 The 2nd level coding process is schematically shown in the following Sheet. SHEET 11 The 2nd level coding is starting with the compilation of text segments belonging to one thematic category. It ends up by the rating of individuals, the respondents, on scales. The simplest possible case would be a process of coding in which only a single variable on a yes/no scale is rated. Interpretation and classification processes, requiring repeated readings of the data, take place between the starting and finishing point of the second-order coding. After the first step, a dimensional analysis is conducted whose natural starting point is a re-examination of the material. The process of dimensionalization was described in detail by Anselm Strauss (cf. Strauss 1991:44 ff. and 204, Strauss and 11 Corbin 1990: 69-72). The principal aim is to identify distinctions in a particular section of the data. The distinctions do not necessarily have to be inductively drawn from the data, but can also be based on existing contextual or theoretical knowledge (cf. Strauss 1991: 41). The process of dimensionalization can be conducted in several steps and on several levels, so individual distinctions can be further distinguished to construct sub dimensions. As a result of the dimensionalisation the construction of variables can be done as the next step. An examination of the material is necessary at three points: at steps 1, 4, and 7. If the material is extensive, it is recommended that the analysis of dimensions and the definition of variables and their values are conducted with only a part of the material. For case-oriented coding an examination of all the material is of course necessary. The model is not conceived of as an inflexible, sequential, forward progressing schema. At certain points it might be necessary to take a step backwards. It might happen that the ‘case-oriented coding’ reveals that the nominal scale values are not sufficiently differentiated so the coding instructions have to be changed. Another possibility is that the original ordinal scale exhibits a degree of differentiation which is not appropriate to the quality of the data. In this case steps 5 to 7 have to be repeated. After a complete examination of the material and the entering of the code values the initial statistical evaluation, a frequency evaluation, can be conducted. If necessary the values the variables take should be modified after this evaluation. This is the case if the frequency of certain variable values is 0 or negligible, and it does not seem meaningful, in terms of the research question, to continue. If we now ask which univariate and multivariate statistical methods can be used to analyze the newly created variables, we must first of all look at the general criteria for statistical analysis. The main criteria are: • • • • Type of scale Distribution of characteristics or variables Size of the sample Appropriateness of the statistical model For data generated by second-order coding the scale is usually not metric, the distribution is not known and the assumption of a Normal distribution usually cannot be justified. The size of the samples is rather small, and compared to the size of well known surveys (ISSP, EUROBAROMETER etc.) is in fact tiny. What avenue of statistical analysis remains open in the face of these adverse conditions? 12 Univariate evaluations of the items by means of frequency counts with the results expressed as percentages are certainly possible. Measurements from robust statistics, such as the median, are more appropriate as measures of central tendency. The arithmetic mean or variance cannot usually be used because of the type of scale. As bivariate methods of evaluation, cross-tabulations with the figures also expressed as percentages are most appropriate. With such tabulations it is easy to recognize conspicuous cells in the table which stick out because of their over-proportional frequency. To identify such configurations, whose content can also be meaningful, the chi-square test can be used and the chi-square residual of the single cells can be calculated. Cells that show a particularly large deviation between the expected and the observed frequency can then are more closely investigated by means of, for example, retrieval of the appropriate coded text segments. As far as the multivariate methods are concerned the classical inferential statistics; variance, regression and factor analysis, as well as the modern methods of linear structural analysis and confirmatory factor analysis are certainly not suitable for this type of data. However, exploratory multivariate analysis methods, such as correspondence analysis and especially cluster analysis are well suited (cf. Anderberg 1973, Aldenderfer and Blushfield 1987, Bailey 1983). The significance and the purpose of the case-oriented quantification method is not just the frequency evaluation, or quantification, made possible by the second coding process. The real advantage is at the level of analyzing relationships and the possibilities for selective retrieval, in which the values of the variables for different respondents serve as a selection criterion for accessing the text segments. Analysis of relationships, employing proven statistical methods, can now be conducted without any further preparatory work. The integration of quantitative operations into the analysis of qualitative data occurs in case-oriented quantification with the objectives of comparing the investigated cases and the creation of types. There are parallels between the simple analysis techniques of coding and retrieval and the method of grounded theory; this method does not stop after coding and retrieval at description and interpretation, but continues the coding. In principle, work on the codes does not stop there, but new codes, integrating codes etc. are defined. As opposed to this work on the codes (content analytic categories), the case-oriented approach works with explicit coding of the values of a variable. The unit of analysis is the person, the case, for whom a score on each variable is calculated. The advantage of classification is immediately applicable; assumptions of and statement about correlation can be verified 13 immediately. This does not have to be oriented towards the significance of relationships. The method would have been totally misunderstood if the only concern were the significance of relationships, this would also be a scarcely adequate strategy with samples that, in most cases, are not random. However, the coefficients, as a descriptive measure, could highlight the strength of relationships within the sample. Their value lays, therefore, not so much in hypothesis testing, but they have a heuristic value in terms of constructing a theory. The second coding process also highlights a problem which, until now, was valid for qualitative data analysis and the coding processes involved, namely the question of reliability and validity was largely neglected. A task for future research will be to develop practicable strategies for controlling the quality of coding. III. MAXQDA in Practice The Desktop The Desktop of MAXQDA keeps the clearly structured main screen with four main windows which is well known from winMAX. 14 SHEET 12 MAXQDA - FOUR-WINDOWS-SCREEN The basic structure of the MAXQDA desktop is the simultaneous presentation of four windows, each of them containing the essential data for the analyses. Size and arrangements of the four windows are not fixed. Each of them can be changed in it’s size, switched on and off depending on the users’ needs. The DOCUMENT SYSTEM WINDOW contains all the texts of the project. All the categories are included in the CODE SYSTEM WINDOW. MAXQDA offers a hierarchical structured code system with up to ten sublevels. The results of the retrieval procedures are always brought into the RETRIEVED SEGMENTS WINDOW. In front of each segment the source information (Text group/Text/Paragraph and code) is given beside the segment. Finally the TEXT BROWSER WINDOW shows the text which is currently in work. In case you want to jump from a text segment shown in the RETRIEVED SEGMENTS WINDOW, a double click on the source information of that segment will bring the complete source 15 text into the TEXT BROWSER WINDOW; the segment is highlighted in the text, so the “quotation” shows up immediately. The basic management functions for the code system and the text document system are integrated into “their” windows. A click on the icons in front shows all the options that are offered and the choice can easily be made. Thus in MAXQDA it is really difficult to get lost - whatever manipulation shall be done – the place in the software where an action has to be done is always clear and mostly found intuitively. The menu bar in the second line of the desktop gives access to all subordinated procedures like import and export of complete projects or parts of it (e.g. the contents of one or more of the four windows), print outs and backups. It gives immediate access to the overviews and management of the codes, the memos, and the variables. The lexical search function (SEARCH) - with an unlimited list of words/strings to be searched for - is invoked from the menu bar as well as from the tool bar. The management of text retrieval in the coded segments of the datasets is completely done in the ANALYSIS section which is accessible from the menu bar. Also from the menu the user has access to the online help: A short introduction, sketching the basic functions of MAXQDA and a comprehensive 300 pages manual with convenient possibilities for selective search is at the users’ hand. The tool bar of MAXQDA, situated directly under the menu bar on the left side, gives easy and immediate access to frequently used features. The right side of this line contains the code bar where the quick-coding and the in-vivo-coding are done. A mouse click on the little text icon “A” changes the code modus into the edit modus. The status line on the bottom of the MAXQDA desktop keeps the user always informed about the mode the program is working in: The number of activated texts, codes and segments in the RETRIEVED SEGMENTS WINDOW. It is shown whether the sub codes are included into the search procedures or not. The sort sequence (sorted by texts or by codes) is indicated as well as the currently chosen search procedure. All icon buttons on the MAXQDA desktop are explaining their meaning by quick info menus which occur when a button is touched by the mouse. 16 Bringing Texts into MAXQDA SHEET 13 No matter what types of texts are brought into MAXQDA - whether you have transcripts from interviews, protocols from observational studies, typed articles or texts copied from the internet - they all have to be brought into the .rtf format (this is not necessary if you use the windows clipboard!). MAXQDA is based on the real .rtf format that means that all font attributes are retained: no matter what font, colour, size or language (special characters) they are. The documents are ordered in text groups. In case of strictly structured data, e.g. answers to open ended questions in questionnaires, guided interviews or data with a fix number of speakers who shall be identified, MAXQDA provides the procedures of the PREPROCESSOR. This specific procedure allows to pre-code the data before it is brought into the program. The coding will automatically be done according to the precoded categories during the process of inserting the data into MAXQDA. 17 The text units in MAXQDA are defined as paragraphs. A paragraph is marked by using the return key – after each RETURN a new paragraph begins. Regarding the limits, the number of texts can not pass over 99.999 thousand, there is no limitation on the software side for the number of texts nor for the number of projects which can be managed in MAXQDA. Working with Texts in MAXQDA SHEET 14 One of the very basic principles in MAXQDA is the Activation of texts and codes. On each level of the code sytem and the text system management in “their” windows activation and deactivation is provided. Thanks to that specific control option the researcher can focus highly flexible on whatever selections of texts and/or categories she/he needs for the analysis of her/his problem in question. A reset button in the tool bar (first icon on the left) helps to avoid unintended activations: Clicking on it will deactivate all texts and codes. 18 Each text stays always immediately available in the text browser window; no matter where you are working, it takes only 2 or 3 mouse clicks to get into any of the source texts. MAXQDA allows selecting single texts and rearranging them in text sets. Text sets can be saved; they are an additional option for the user to focus on specific topics. They also allow saving the results of a complex activation process (e.g. only the texts with the distinguished features: female, under 30 years, divorced, more than 1 child, satisfied with the situation of life). The structure of the text groups is not touched by the Text Sets. MAXQDA provides an edit modus in the text browser window. A mouse click on the icon button “A” in the code bar changes it into the edit bar. Now you can add, delete or correct parts of the text in the text browser window, shortly any changes can be done. In case you want to create a new text, you just have to define a new text, open it in the text browser window, switch on the edit modus and anything can be typed in or brought in via the Windows clipboard (STRG/V). MAXQDA offers particular features to support teamwork during the analysis process. The whole projects with all codings, memos and variables can be transferred by disk or email to the different members of the team. The second transfer option is: Just the codings and memos of any selected text can be exported by a team member and imported by another one. Memos can easily be marked in different colours and/or symbols to support the communication between the team members. To enable a clear and easy data organisation, MAXQDA provides overviews over the entire coded segments, memos and variables. The overviews are organized in highly flexible tables. The structure can easily be changed by the user; one mouse click on a quoted text brings it into the text browser windows. All tables in MAXQDA can be copied (by CTRL+C) to the Windows clipboard and imported (by CTRL+V) into other programs as MSWORD, Excel or Power Point. 19 Coding in MAXQDA SHEET 15 MAXQDA supplies a variety of code- tools, allowing a specific, convenient and easy coding work. The code system is hierarchically structured and allows creating up to ten levels for the sub codes. The text segments are marked by using the mouse, the coding is done either by drag & drop, choosing directly a code out of the code system or by using the quick-code-list in the code bar. Single words or expressions in the source text which highlight specific phenomenon can directly be incorporated into the code system by using the In-vivo-coding. Clicking on the in-vivo-icon in the code bar will define the marked word/expression as a code and insert it automatically into the code system. The coding is fully visualized, the beginning and the end of a coding is marked out, the indication of the code is optional. Touching the little rectangle in the code marking shows a quick info quoting the complete code. The code volume can easily be fitted to the users need by enlarging or minimizing. Five different colours make it easier to differentiate code types. 20 A unique feature in MAXQDA allows the researcher to weight the coding. Clicking on the right button while touching the rectangle of a code mark will bring up the weight menu showing a fuzzy variable “weight”. The importance of a segment can be indicated by giving in a value between 1 and 100. The weight can be used as a selection filter in all retrieval procedures. The code system provides the researcher with the flexibility which is needed in the complex process of coding, analysing, recoding etc. By simple drag & drop operations any changes in the structure of the code system can be done. An undo-button in the code bar makes it easy to delete the last coding – they can be chosen out of a pick list showing the executed code actions. Again, MAXQDA guarantees a clear management of the coding in the overview tables: The frequencies, the coded segments and the linked memos are organized in tables of the same appearance, structure and properties as we described for the text overview. MAXQDA has no limits in the number of codes. A coded segment is limited to 65 kb, which are approximately 30 pages. Text Retrieval in MAXQDA SHEET 16 21 MAXQDA distinguishes two different types of text retrieval: Code related retrieval procedures and “free” retrieval procedures: The lexical searches. The results of the code related retrievals are presented on the list of selected Segments window. The sources of each retrieved text segment are visualized in the source volume on the left side of the window. All information regarding the segment is shown: the text group, the name of the text, the line, the name of the code, the weight. A mouse click on the info pad highlights the segment and brings at the same time the source text into the text browser window. This is one of the highly effective tools guaranteeing that the researcher never looses the contact to the source text, supporting her/him to find the best possible balance between contextualization and decontextualization. The researcher is very flexible in arranging the retrieval results. The order sequence can be chosen: A sequence order by texts or by codes. MAXQDA offers the weight filter as a powerful tool to reduce the data under the researcher’s control: Only those segments are listed in the list of selected Segments window which correspond to the weight defined – so a relieving reduction from more than 1000 to less than 100 text segments can be achieved. MAXQDA offers a variety of retrieval options. The basic way to run a retrieval procedure is activating the text(s) and code(s) which shall be included into the search. The result of all codings which are found in the selected texts will be shown in the list of selected Segments window. For an ambitious in-depth analysis MAXQDA offers complex types of retrieval procedures: Besides the Boolean operators 10 different analytical strategies are provided to guarantee an out-standing level of the hypothesis testing and theory building assistance. The results of retrieval procedures can automatically been coded if the researcher finds it worth to be saved. All retrieval options are accessible from the menu bar. Pressing the Analysis button brings a pop-up-menu where the selections can be chosen. The index and the frequencies of the coding are again presented in MAXQDA overview tables of the same appearance, structure and properties as described above. MAXQDA Analysis pop up menu 22 The lexical search function offers the possibility to search independently from the so far defined codes and prepared coded text segments. The search function is accessible from the menu bar (“Search”). The search can be done in all the texts, in the memos, in the text segments currently presented in the list of selected Segments window or in the activated texts. It can be searched for complete words or just the strings; the number of search words is not limited. The user can define how many lines before and after the finding should be shown. The result is presented in a MAXQDA table which has the same properties as the ones described above. The results can be saved and/or copied to the Windows clipboard and directly be used for further analysis. Likely as for the complex retrieval results, the automatic coding is also provided for the lexical search. Memos in MAXQDA SHEET 17 23 The memos can be regarded as the researcher’s storehouse for their ideas, background information and first hypothesis. They are an important tool to increase the reliability of coding and they can also play an important role as a communication tool for the members of a research team. Memos can be attached at each text line of a text shown in the text browser window. A double click in the memo volume right beside the text brings up the memo dialog box. Up to 30 pages text (64 kb) – in .rtf format - can be written in the text field. Codes can be taken out of the code system and linked to a memo. MAXQDA offers 11 different colours/symbols to differentiate specific memo types (e.g. theory/communication/literature….). Memos can also be attached at codes. This allows writing down definitions of codes and anchor examples which stay always immediately accessible. MAXQDA Memo Overview Table MAXQDA Memo Dialog Box MAXQDA provides a refined management of the memos. The researchers can immediately and spefically access their memos. Overviews on different levels are at the users’ disposal. The MAXQDA tables can be requested for one text, for a text group, for memos linked to a single code or the code system, as well as for all the memos of the project project. Line in the table corresponds to one memo. Each memo can be opened directly out of the table line. The memo manager for all projects is accessible from the menu bar, the selected overviews are accessible from each level where the selection shall be done (text manager, text group manager etc.). The memos are completely incorporated into the lexical search function. Thus, the user can even search for specific words or strings over all his memos. There is no limitation for the number of memos. 24 Variables in MAXQDA – Selective Tool Inside & Bridge to Merge Qualitative and Quantitative Analysis SHEET 18 Variables in MAXQDA can be defined as a set of attributes which is linked to each document. There is no limitation in the number of variables which can be defined or handled in the program. Four types of variables are offered: string (text), number (integer), floating point number and Date/time. The variables are managed in tables as we described above. The table has as many rows as there are documents in the project and as many columns as variables are defined. The order of the volumes can be changes by the researcher, the display of each volume can be switched on or off. 25 MAXQDA Variable Table The variables are directly accessible from the menu bar as well as from the tool bar. Once a variable is defined 4 internal variables are automatically created: The text group, the text, the creation date of the text and the number of coded segments are generated and shown in the variable table. If a variable is created and the process of giving in the values is started, MAXQDA will save the entered values in a pick list. After a few inputs the values just have to be chosen out of the pick list which avoids mistakes and makes the input much more convenient. The variable table or selected parts of it can easily be exported to statistical software like SPSS, EXCELL and others, as well as variables can be imported to MAXQDA from statistical packages. Besides the merging of qualitative and quantitative data analysis the variables have an important function within MAXQDA. They are an easy and efficiently to use selection tool for the documents. On the level of the project manager specific variable values can be selected and MAXQDA will accept this selection as a filter for all search procedures – only those documents are activated which correspond to the selected variables values. Regarding the limits: There is no limit for the number of variables. The Strings of the variables are limited to 64 characters – but the limitations of the quantitative programs should be taken into account, especially when the exchange of variable is planned. 26 MAXQDA: Some Special Strength SHEET 19 Trying to mark out some of the special strength of MAXQDA we consider 4 characteristics to be highly significant. According to our understanding all of them are very important to an efficient support of qualitative data analysis. MAXQDA is a highly flexible tool, which supplies the researcher with an outstanding range of options to organise and reorganize the data corpus and the code system, to make any changes at the source texts and the memos. The researcher is supported very well in carrying out any changes she/he thinks to be necessary. The second dimension concerns the openness of MAXQDA to surrounding software programs and to modern forms of working together. Qualitative data analysis is more and more done in teams, is more and more done in national and international coop27 eration. An increasing number of projects consist of several parts, data are gathered in different cities, regions or even countries and the researchers try to proceed a comparative analysis. MAXQDA offers with its teamwork features several strong and powerful tools helping the researchers to realize an effective cooperation, to share the data corpus, the code systems and to keep each other on track in all aspects of the analytical process. The hostility of the qualitative and quantitative research paradigm softens more and more. An increasing number of researcher accept that the results of empirical research don’t profit of the cultivation of an ”either….or” in the choice of research methods. They realize that there a lot of research subjects where a triangulation (c.f. Flick 1998:229f.) of methods is appropriate – and more and more quantitative elements are incorporated into qualitative research studies. The features offered in MAXQDA guarantee a smooth transferability from quantitative to qualitative software tools and aim to be a helpful contribution to this development. Another dimension with an increasing importance is the accessibility and transparency of the data. Qualitative research projects tend not only to be shared out into different geographical places, but we can also find a development towards the enlargement of the data corpus. Data sets with several hundred documents are not anymore exotic exceptions. The development of powerful software tools to support the qualitative data analysis had as one effect the possibility to work with larger and more complex data sets. In this situation it becomes a great and difficult challenge for the researcher to reach this best possible balance between contextualization and decontextualization. MAXQDA supports the researcher in facing that challenge by providing a variety of highly efficient tools and offering a basic structure of the software itself which guarantees the researcher • • • • to keep always in narrow contact to the source text – it is never more that two mouse clicks away – to keep in touch with his own ideas, written down in the memos to provide at any place and stage of the analysis the information about the sources of the decontextualized parts of the texts (i.e. the codings) to enable a transparent and comprehensible Management of the entire information gathered on the different levels and in the different sections (texts, variables, codes, memos) of the project (overviews in the MAXQDA tables). Last but not least we are convinced that the power and technical standard of a software program has not at all to correspond with a complicated and hard to learn handling of the provided tools. In the development of MAXQDA it is - and it was 28 already in winMAX - an important criterion for the internal validation of the programs’ structure and the single tools to make it as intuitive to handle as possible. The demands of the software tool itself were minimized in order to save the researchers their energy and intellectual power for a sophisticated, fruitful analysis. As qualitative researchers we are sure that it should be one of the basic aims of a software tool to provide an easy to handle tool which allows the researcher to concentrate on his real task: To analyse his data, find the best possible answers to her/his research questions and to develop theories which help to understand better what is going on in the real world. Literature Giddens, A. 1984: The constitution of society. Berkeley and Los Angeles Denzin, N.K., Y.S.Lincoln 2000: Handbook of qualitative Research. Second Edition. London/Thousand Oaks/New Dehli Flick,U. 1998: An Introduction to Qualitatve Research. London/Thousand Oaks/New Dehli Strauss, A., J. Corbin 1998, 2nd. Ed.: Basics of Qualitative Research. Techniques and Procedures for Developing Grounded Theory. London/Thousand Oaks/New Dehli Silverman, D., 2000: Doing Qualitative Research. A Practical Handbook. London/Thousand Oaks/New Dehli Aldenderfer, M. S./Blashfield, R. K., 1987: Cluster Analysis. Beverly Hills/London/New Delhi: Sage Anderberg, M., 1973: Cluster analysis for applications, New York: Academic Press Bailey, K. D., 1983: Sociological classification and cluster analysis. Quality & Quantity 17: 251-268. Bryman, A. 1988: Quantity and Quality in Social Research. London/New York. Corbin, J., Strauss, A., 1990: Grounded Theory Research: Procedures, Canons and Evaluative Criteria. Zeitschrift für Soziologie 19: 418-427. Geertz, C., 1973: The Interpretation of Cultures, London. Kuckartz, U., 1991: Ideal Types or Empirical Types: The Case of Max Weber's Empirical Research. Bulletin de Methodologie Sociologique published by the International Association of Sociological Methodology, No. 32. Schutz, A., 1962: Collected Papers I. The Problem of Social Reality. The Hague:Martinus Nijhoff. Strauss, A., 1991: Grundlagen qualitativer Sozialforschung. Datenanalyse und Theoriebildung in der empirischen soziologischen Forschung. München. Weber, M., 1964: Wirtschaft und Gesellschaft. Grundriß der verstehenden Soziologie. Studienausgabe, hrsg. von J. Winckelmann. Köln/Berlin. 29