Copyright Sociological Research Online, 1998

 

Christine A. Barry (1998) 'Choosing Qualitative Data Analysis Software: Atlas/ti and Nudist Compared'
Sociological Research Online, vol. 3, no. 3, <http://www.socresonline.org.uk/3/3/4.html>

To cite articles published in Sociological Research Online, please reference the above information and include paragraph numbers if necessary

Received: 28/5/98      Accepted: 3/9/98      Published: 30/9/98

Abstract

Choosing between Nudist and Atlas/ti, the main qualitative data analysis software packages can be difficult. To assist researchers in making this choice, I have conceptualised their differences along two dimensions, related to the qualities of the software and of the research project. The software dimension is structural design, and the project dimension is complexity. Software structure is dichotomised between structured, sequential, verbal versus visual, spatial, interconnected modes of operation. Projects are dichotomised between homogeneous sample, short timeframe, single data-type, single data analyst; versus multiple samples, longitudinal data, multiple data types and team data analysis. First I review the CAQDAS literature. Then I outline the different personalities and strengths of Atlas/ti and Nudist, and show how they match these dimensions. I offer suggestions as to how to choose, and whether to use in tandem with complementary conceptual network software.

Keywords:
CAQDAS; Chaos Theory; Computer Software; Conceptual Network Software; Hypertext; Qualitative Research

Introduction

1.1
Many qualitative researchers are deciding whether to use computer assisted qualitative data analysis software (CAQDAS) and if so which package to use. Common questions that are being asked include: What are the pros and cons of using such software? How will it affect the process and end product of research? And, If I choose to use it, how do I decide which package most suits my needs? In this paper I hope to go some way towards answering those questions.

1.2
In this paper, firstly I will explain how my own trajectory of experience influenced my opinions about the software. Next I present and discuss the perceived pros and cons from the literature. Then I introduce the idea that specific packages have differential effects on the research process. In the main body of the paper I focus on a comparison between what I see as the two most useful packages: Atlas/ti and Nudist. I have conceptualised the two packages along two dimensions: structural design of software and complexity of research project. I have found that this offers a system for choosing between the two for different researchers and for different projects. I then review the features of the two packages in terms of these two key dimensions of difference and make recommendations as to how researchers can choose between them.

1.3
As a qualitative health researcher I would like to have had more ways of differentiating between the packages three years ago, when I was setting out on the path to computerise my data analysis. Hence, this paper. I have no axe to grind, unlike may articles on this subject that have been written by the software developers. I speak as an ex- technophobe who has come to appreciate how technology can help me do my job better, faster and more creatively. I speak with some knowledge of how academics feel about the move to technology. I worked as the researcher on the Information Access project, an ethnographic case study of 18 academics and their take up of technology over a three year period (Barry and Squires, 1995; Squires et al, 1995). My final claim to knowledge in this area comes from three years using Nudist. Firstly I used it in education, and now I am using it in health research. During this time I have run several training workshops on Nudist and Atlas/ti , for other researchers (eg. Marshall and Barry, 1997) .

Pros and Cons of Computer Assisted Qualitative Data Analysis Software

2.1
The growing literature on computer assisted qualitative data analysis software (CAQDAS) expresses both hopes and fears. The hopes are that CAQDAS will: help automate and thus speed up and liven up the coding process; provide a more complex way of looking at the relationships in the data; provide a formal structure for writing and storing memos to develop the analysis; and, aid more conceptual and theoretical thinking about the data. In spite of these pros there are a good many criticisms and worries about the software in the literature. What Seidel (1991) calls 'the dark side of the technological advance'. The main worries are: that it will distance people from their data; that it will lead to qualitative data being analysed quantitatively; that it will lead to increasing homogeneity in methods of data analysis; and that it might be a monster and hi-jack the analysis.

2.2
The concern about distance from the data was raised by Seidel (1991) in discussing his own package, The ETHNOGRAPH. This fairly early package did suffer from a coding process that was very complex. This might have led its users to be so caught up with working out how to code that they may have lost sight of their data. Weaver and Atkinson (1994) certainly report this problem when they tested it out in tandem with some of the other packages.

2.3
Those who express this concern have often not used CAQDAS. Those who have tried the software have realised that it is not possible to analyse your data without reading and being familiar with it first. Continuing analysis relies on the necessity of re-reading data both in complete transcripts/fieldnotes and in categorised chunks, over and over again, to develop an analysis with any depth. In the worst case, it is possible that a researcher could only read the data in context during the process of initial coding (a prerequisite of any coding). After that, they could just read snippets of data that have been coded under each category to develop a final analysis, without ever returning to the full contextualised data. However this strategy is equally possible for those that use index cards, scissors and photocopies or word processor cut and paste functions. It may be possible to produce analysis using this superficial brush with the data but it is unlikely to yield a quality analysis. Technology does not increase this likelihood. There will always be some researchers who conduct superficial analyses that are lacking in rigour and depth, whatever their tools.

2.4
I think it is quite possible that some of these fears about CAQDAS do originate from those who have not worked with it very much if at all. The Information Access project identified a circle of non- use of technology breeding fears and misconceptions that were then reinforced by non-use (Barry, 1995). Once the same individuals started to use technology their negative perceptions were usually replaced by positive ones and an enthusiasm for positive ways in which the technology could help them. In the Nudist workshops that I have run there is a lot of anxiety and negativity amongst those who have not yet used this type of software. There has only been praise from those who have, whilst recognising its limitations as well as its strengths.

2.5
I have more sympathy with Mason's (1996) concern. She cautions about being seduced by the capabilities of software into treating categorically indexed slices of data as more concrete variables, and conducting quantitative variable analysis. It is true that many packages allow the researcher to categorise data according to demographics and other groupings arising from the data. It is also true that some software provides statistics about the proportion of text that is coded against a particular code, and the number of times a particular code has been applied. Some packages even allow for data in mixed qualitative-quantitative studies to be imported from and exported to SPSS (the quantitative data analysis package). These features might indeed lead researchers to perform types of analysis more suited to quantitative data. Counting occurrences, giving more weight to more frequent events, ignoring isolated incidences, and formulating and testing out rigid hypotheses are not sensible ways to analyse qualitative data[1]. This type of analysis would lead to clashes between method and approaches to epistemology and explanation favoured by qualitative researchers. However this was also a danger before computerisation. I have seen a number of qualitative data presentations that focus on counts at the expense of conceptual and theoretical explanations. The answer here seems to be good supervision and training of qualitative researchers and the recommendation of useful 'how to' books like Mason's (1996) to encourage novice researchers to reflect on these issues.

2.6
Another fear is that the rising use of CAQDAS is resulting in the convergence towards a single orthodoxy of data analysis (Coffey et al, 1996). I think there is some evidence that some inexperienced researchers cite CAQDAS software as a prop. In much the same way that some inexperienced researchers are laying inaccurate claims to Grounded Theory (Green, 1998). I have heard more than one novice qualitative researcher claim Nudist as their method of data analysis, as though Nudist were an epistemological standpoint. However I do not agree that this tendency equates with the suggestion that CAQDAS is forcing researchers towards a single ideal type of data collection analysis and storage (Coffey et al, 1996). As one of the authors of this paper explains elsewhere (Weaver and Atkinson, 1994: p. 165). Researchers will be more likely to take what they can from the software and use supplementary non-computerised methods, than to confine themselves to the limitations of computer methods. Perhaps first time qualitative researchers might be tempted to start with 'grounded theory' as a method and with computerised data analysis as a tool. As much as anything I think this represents a way of dealing with the anxiety of those unused to dealing with complex and unstructured datasets (Kleinman and Copp, 1993). However once qualitative researchers find their feet they will soon be happy to reject methods and tools that will not serve their type of data and their type of problem.

2.7
In my own research, I use Nudist as just one tool in my analysis armoury, as it only helps me to do part of the work of analysis. For cross-case thematic analysis I find Nudist a useful way to gather data together and then play with it. For the case study and temporal aspects of my data I am happier using a word processing package to gather data together into a holistic, narrative sequence. For theorising, and abstracting meaning from the coding I find hand-drawn diagrams and tables (in the style of Miles and Huberman, 1994) a very useful tool. Re-ordering the hierarchy of codes and linking memos to the data in Nudist is equally useful for conceptual development. For crystallising my ideas about the final analysis I return to writing with my word- processor to explore my thinking. I do not rule out using other CAQDAS software, such as Atlas/ti in the future for other types of data. I do not feel constrained by Nudist, only helped, and I feel sure that my colleague qualitative researchers will do the same. Some may not. They may be happy to be limited by what CAQDAS can produce for them. There have always been researchers who have been happy to develop simplistic analyses using the minimum of interaction with their data.

2.8
In general, I do not believe that the types of researchers interested in collecting qualitative data will 'lead to the uncritical adoption of a particular set of strategies as a consequence of adopting computer-aided analysis.' (Coffey et al, 1996: ¶7.3). I believe that they will adapt CAQDAS to their needs and add it to their expanding toolbag. I also believe that CAQDAS offers more than code and retrieve. It can be used at the stages both of more simple coding and of more complex analytical thinking. In any case, it is wrong to assume that all CAQDAS is homogeneous. As Kelle (1997) explains 'Theoretical and methodological concepts of developers and users of computer software for textual data management are much more diverse and heterogeneous than is often assumed.'

2.9
There are fears among non-users that CAQDAS software might be a monster and hi-jack the analysis. However, the consensus is that such packages are not monstrous but only exert some moderate degree of influence on the process of analysis (Buston, 1997). CAQDAS does some tasks for the researcher: data administration and data archiving, but only provides assistance in the theoretical thinking and analysis itself, which is as it has always been, the job of the researcher themselves (Kelle, 1997).

2.10
I think the views expressed by Lee and Fielding (1991; 1996) are more consistent with the views of the majority of existing CAQDAS users. Their reply to Coffey et al. (1996) is that the software represents not a developing orthodoxy but 'the multitooling of qualitative researchers, making available to them more or less at will, a wide range of analytic strategies.' (Lee and Fielding, 1996: ¶2.2) . They also see 'the ability effectively to manage data may be a considerable improvement over the ad hoc procedures we suspect frequently underpinned manual analysis.' (Lee and Fielding, 1996: ¶2.3) . They also stress other advantages: the speed and flexibility with which coding can be achieved, freeing up time for analytical tasks. The greater ease with which deviant cases or small bits of significant material can be extracted; the fact that researchers are encouraged to 'play' with their data and the fact there are better opportunities for teams to analyse together and for replication to be performed. They also point to disadvantages of the software. These do seem to be more realistic ones however, such as the dangers of the convergence of quantitative and qualitative methods, the limitations for analysing temporal or sociolinguistic data; and the loss of the 'untypable' (Lee and Fielding, 1991).

2.11
To sum up this review of the CAQDAS literature, I think it is important to stress that these programs are just another tool, with faults and with benefits. As Lee & Fielding point out from their own research, users of CAQDAS programs 'have found little hesitancy over abandoning program use in cases where software did not meet the analyst's needs or where it was perceived to be at variance with the researcher's epistemological presuppositions.' (Lee and Fielding, 1996: ¶3.3) . Not every piece of software will be relevant to every task and researchers will often be able to achieve their ends using non-technology solutions or simple word processing cut and paste. I would however encourage those who have yet to try them out to get familiar with what they can do. In this way they will know when it might be useful to pull this particular tool out of the bag.

2.12
At the end of the day, it is up to the individual researcher to take responsibility for deciding how useful the software will be for them, which package they should use, and how they will integrate this into their existing analysis methods. Talking to other researchers who are using it, reading Weitzman and Miles (1995) excellent review, and accessing the CAQDAS website at Surrey university <http://www.soc.surrey.ac.uk/caqdas/> are all advisable approaches for getting more detailed insight.

Differences between CAQDAS Packages

3.1
A lot of the literature treats CAQDAS as though it were one unitary product but this is not the case. Many of the generalisations made by authors may well depend on which types of package they have had experienced. Weitzman and Miles (1995) for example, talk about five different categories of software and review a total of 24 packages. The situation is very dynamic and not only are there new programs being developed all the time, but there are new versions of the same packages being released regularly. Each offering substantial changes to the last. This means that the literature reviewing CAQDAS has a very short shelf life. Reviews such as those of Tesch (1990) and Weitzman and Miles (1995) whilst both excellent, were both out of date quite quickly. For example, Nudist was not even a contender when Tesch wrote her book. The version reviewed in Weitzman and Miles (1995) version 3, has now been superseded by a substantially improved and quite conceptually different version 4.

3.2
Different packages and different versions do have differential effects on the analysis process. Weaver and Atkinson (1994) feel that different packages transform the data in different ways and that this in turn encourages different ways of thinking about data and theory. They analysed a common data set using five packages, representing four different categories of data analysis approach. They tested ETHNOGRAPH as a pure code-and-retrieve package; FYI 300Plus as a lexical searching program; Kwalitan and Nudist as coding and theory building programs and GUIDE as an example of a hypertext program. (The same issue of newer versions applies here, as for example, they did not review Atlas/ti which has become one of the major players and they used version 3 of Nudist). Whilst their findings about the way the software shapes the analysis are out of date, their more general points about the effect of the package on the process still stand.

3.3
Their central argument is that hypertext[2] is far superior for research than the other types of software. Their criticism is that the 'coding' principle of organisation, with its contingent loss of contextual information, seems to conflict with the goal of holism. They feel that hypertext retains more holism, yields a richer description, is more amenable to the creative process, is more flexible and dynamic and encourages reflexive modes of thinking.

3.4
They also recognise the potential pitfalls of hypertext: too little predefined structure and too much flexibility might lead to cognitive overload. This openness makes it disorienting and difficult to know where to begin with analysis, whereas other packages have more obvious starting points and set strategies. They talk of the dangers of getting 'lost in hyperspace'. They suggest these hazards demand organised thinking skills. They also feel that a novice researcher with no existing procedures for analysis might find this type of software more exciting than those with established routines for analysis. I would argue that novice researchers may equally likely be scared off by hypertext. I certainly had this experience when offered HYPERCARD (a Macintosh based hypertext program) as a means of analysing my data as a novice qualitative researcher. I was looking for more reassuring structure and set routines, and found Nudist far more satisfactory in this respect.

3.5
In some ways Weaver and Atkinson's negative views of coding packages are exacerbated by their choice of package to review. The ETHNOGRAPH is more inefficient, inflexible and labour intensiveness than others in its class. The newer versions of the more sophisticated software such as Nudist and Atlas/ti are far less cumbersome than the original ETHNOGRAPH (I can't speak for later versions) and both have incorporated elements of the hypertext media into the current releases of their programs (Atlas/ti from the start and Nudist only in the latest version 4, and to a lesser degree). I will return to this issue in the comparison between these two packages below.

Focusing on Two Packages - Atlas/ti and Nud*ist

4.1
Weitzman & Miles recognise that the newest category of coding and theory building software offers most to researchers. Within this category they place Atlas/ti and Nudist at the head of the field:

'The only program in its category [Atlas/ti] is not a clear winner over, is Nudist and the choice between the two is not clear-cut. You will have to weigh up the strengths and weaknesses and match them against your needs.' (Weitzman and Miles, 1995)

4.2
I would agree that Atlas/ti (version 4.1) and Nudist (version 4) appear to be two of the most serious contenders, in meeting the requirements of researchers. How the two differ is the subject of the remainder of this paper.

4.3
The feature by feature comparison of the software packages provided by Weitzman and Miles are helpful in gaining a view of the strengths and weaknesses of the two packages. However, I found that at the end of reading I was still unsure of which to recommend to others and on what criteria. A comparative workshop that Martin Marshall and I ran for the Annual Conference of the Association of University Departments of General Practice (AUDGP) allowed me to build up a picture of the two different personalities of Atlas/ti and Nudist (Marshall and Barry, 1997). In this paper I hope to provide a more qualitative feel for the software to supplement the more quantitative analysis of differences of Weitzman and Miles (1995).

Two Dimensions of Difference

5.1
There are many similarities between Nudist and Atlas/ti, which is why it is difficult to choose between them. I will concentrate here on defining what I see as the main differences. My systematic analysis of the differences has led me to conceptualise them along two dimensions. The first dimension is the structure of the software, and how this impacts on analysis. The second dimension is complexity of the research project.

5.2
The structure of the software is a matter of researcher preference, which although it might change over the career of a researcher is likely to be fairly static across different types of research. The degree of complexity of the project will, however, change according to the type of work a researcher is engaged upon. Therefore in reading this you will need to think about your preferred methods of working with computers; the type of project on which you are currently working; and, the types of project you may work on in the future.

Interconnected versus Sequential Structure

6.1
Software packages have different personalities and different structures. In part this is a result of two long- term trends. The first of these is the move along the continuum from simple, binary, serial, sequential processing at one end to complex, connectionist, networks at the other. The second is the improvement in 'user-friendliness' and the human computer interface. The two developments are happening in tandem (Winograd, 1996). As software becomes more complex in its structure it is also becoming more user friendly, more intuitive and more easy to learn and use. The world wide web is a good example of a complex interconnected network where many, many computers are linked together through a web of connections. The documents on the web are also organised in hypertext, another form of complex connectivity. The Netscape interface to the world wide web is a good example of software that is very user friendly, intuitive and easy to learn.

6.2
Whilst both Nudist and Atlas/ti are relatively sophisticated in terms of software development, Atlas/ti seems to be further along this continuum. It has a more complex inter-connected, hypertext structure and it is more intuitive and easier to learn. Nudist has a more sequential, linear structure. It also has a clumsier interface, is less intuitive and less easy to learn. In some ways these reflect the differences drawn attention to by Weaver and Atkinson (1994). Nudist represents a sophisticated coding and theory building package and Atlas/ti is more of a hypertext package. Atlas/ti operates in a more visual and spatial medium with data and software functions organised in pictorial form, while Nudist's operation is predominantly verbal.

6.3
This does not mean to say that Atlas/ti is necessarily better than Nudist. There are other differences in terms of the actual functions of the software that are discussed under project complexity below. Also there is a matter of personal preference. Just because software is further along the evolutionary path does not mean that individuals will prefer it. Apple Macintosh computers were for a long time a lot further along this road than IBM PCs (until the advent of Microsoft's Windows interface which has closed the gap). However individuals often still had a preference for the IBM system and preferred the more sequential, simpler, less user-friendly DOS system to the more organically structured Windows. It may be too, that novice users prefer simpler sequential structured software to organic connectionist software, given the more obvious structured routines, fewer ways of doing things, and less room for becoming lost in hyperspace.

6.4
The different types of software structure may map onto differences in individuals' preferred modes of information processing styles. For example, some individuals use different styles of learning (Honey and Mumford, 1992). Also, some people differ in their ease and ability to visualise, create and work with images (Isaac and Marks, 1994). Others like to work in more than one way with information, drawing on a variety of information-processing methods.

6.5
Where individuals do have a fixed preference for highly structured, sequential forms of data processing, this may link to a preference for order and having a limited capacity for uncertainty. It may also link to the type of academic training that individual has undergone. Mathematicians and scientists are trained to utilise logical, (in the pure mathematical sense) ordered and sequential forms of information-processing in their work. (Although the importance of creativity and unconscious processes in science is also recognised, Poincare, 1908).

6.6
The questions potential users need to ask themselves are: What software structure do I prefer, simple and sequential or complex networks? Verbal or visual/spatial? How much of a novice am I where qualitative analysis and software use is concerned? Do I have a preference for highly structured ways of working or am I happy to stay with uncertainty and be patient in the development of creative solutions?

Project Complexity

7.1
Qualitative projects differ markedly in the levels of complexity of their research designs. The most simple of projects consist of one homogeneous group of participants, often quite small sample sizes, only one mode of data collection, e.g. in-depth interviews, and are carried out over a relatively short period of time. Conversely, the most complex of projects can have one or more of: different participant groups, a longitudinal component, varied data types, linked data, large sample sizes, and large research teams.

7.2
To give examples from my own experience one simple project I conducted was to look at the difference between men's and women's definitions of morality. My data consisted of one straightforward in-depth interview about views of morality. The analysis involved thematic comparisons between the women and men to look for similarities and differences (Barry, 1994).

7.3
In contrast, my current study of doctor-patient communication[3] involves most of the components of a complex project. There are two participant groups, general practitioners and patients; a large sample (20 doctors and 70 patients); a number of different data types: doctor interviews, patient interviews before consultations, patient interviews a week after consultations, audio-taped consultations and fieldnotes. These different data types are linked by case; the fieldwork is being done by two researchers based at two different sites and a multi-disciplinary team of five is involved in the analysis.

7.4
Atlas/ti and Nudist have different features to offer projects of differing complexity and in the next two sections I want to explore how the two fit along both these dimensions of preferred structure and complexity.

Structural Features of the Two Packages

8.1
The two different structures of the two packages result in different benefits and disadvantages for both, arising from the different way in which they have been designed (see Tables 1 and 2).

8.2
Atlas/ti's strengths are related to its immediacy, its visual and spatial qualities, its creativity and its inter- linkage. To quote from the developer's internet homepage <http://www.atlasti.de/atlasneu.html>: 'The main strategic modes of operation can be termed 'VISE': Visualization, Integration, Serendipity and Exploration.' Having all aspects of the data and analysis on screen at once and being able to visually map out relationships between different parts of the data and theoretical ideas, and to form links between them and jump back and forth, all seem to encourage that creative process of sparking ideas and pattern recognition.

8.3
The free-form, loose structure of Atlas/ti may evoke anxiety in people and some may prefer the 'safer' more structured approach of Nudist. Conversely, those who are easy with a mass of data and comfortable with uncertainty may prefer the variety of options in Atlas/ti. It may facilitate seeing links between different aspects of the data and theoretical ideas.


Table 1: Strengths of Atlas/ti arising from complex, hypertext software design
Atlas/ti's strengthsNudist's weaknesses
+ Visually attractive / creative

+ All features on screen at once

+ Well designed interface

+ No limits on units of coding: can
code any amount of text as one unit

+ Hypertext links between data
/codes /documents (i.e. click
on one jump to the other)

+ Conceptual network displays
which can be manipulated

+ Can include text segments and
memos in network displays

+ Assists creative and visualised
theoretical thinking

+ Memos can be multiply linked

+ Can process audio and visual

+ Best used on screen
- Less attractive/smooth interface

- Less immediate: several steps to
each process

- Fewer links between data/
codes/ documents

- Memos can only be linked to
codes, one per code

- Units of coding have to be decided in
advance and used throughout project

- No option to print or annotate 'tree'
hierarchical display of codes

- No option to develop visual
relationships between data apart
from hierarchically

8.4
Nudist's strengths are its more structured organisation, its project management functions and its sophisticated searching. It allows hierarchies to be built and developed. One of its organising principles is system closure, where any first stage of analysis (eg. memoing, annotations or coding) can become codable material in an iterative, potentially never ending cycle. It facilitates to-ing and fro-ing between on-screen and paper analysis, and it has many clever options for searching that allow sufficiently complex questions to be explored in the data.


Table 2 Strengths of Nudist arising from sequential, structured software design
Nudist's strengthsAtlas/ti's weaknesses
+ Hierarchical structure which can be re-organised
in stages

+ Project management tools e.g. progress
log, lists of codes plus definitions

+ Structure makes it easy to familiarise and apply
coding scheme

+ Memos can be searched and coded

+ First pass coding can be re-coded

+ Documents can be annotated, edited or appended
after import

+ System closure: results of searches become codes
which can be memoed and searched

+ Can switch between on- screen and paper coding
and theorising

+ Systematic exploration of ideas, questions in the
data with sophisticated searching

+ More limited structure may provoke less anxiety
- Unstructured, few limitations:
loose structure and uncertainty

- Limited searching capacity

- Limited case and project
management tools

- Can't alter or annotate files


Nudist , Atlas/ti and Project Complexity

9.1
Nudist's more systematic nature comprises a number of features that aid more complex project work. It is easier to keep track of progress on complex projects with the project management emphasis. It is suited to the handling of different data types, for example data can be linked by case; or longitudinal data can be analysed by time of data collection. Where a project employs both quantitative and qualitative data Nudist has a number of features for importing and utilising the quantitative facts known about a case in analysing the qualitative data. Nudist allows multiple researchers based at different sites, even those using different hardware (PC or Mac), to work on coding separately and then combine the work done at stages using the Merge function. For example, on my current project, my colleague and I are doing the first pass of analysis on half the sample, 35 cases each. We will then combine the analysis into one dataset and share out different aspects of the data to conduct a more detailed analysis across the whole 70 cases.

9.2
Nudist's sophisticated searching options can be used to explore the data in real depth by cutting the data on any number of different dimensions (always remembering the caution about not being lured into conducting inappropriate variable analysis (Mason, 1996)). This becomes more useful the more heterogeneous the data and the more complex the project structure. So for example, any information about the project can be used as a way to sort and view the data. This can be demographics, on our project: gender of doctor, size of practice, patients' educational level and so on. Or it can be concepts that arise from the data itself, doctors view of the role of medicine, nature of relationship between doctor and patient, attitude to complementary therapies and so on. Whilst simple splits like gender could have been achieved in the days before CAQDAS, complex mixes of ways to cut the data would not have been so possible without endless sorting and re- sorting through transcripts. Nudist does this in a matter of seconds. It produces a file with all the results that can be printed out and studied for patterns.

9.3
Another advantage when working on complex projects with large numbers of data documents is the option to automate parts of the coding process. On our project for example we have 70 cases each comprising 5 different data types: 350 documents. Command files can be built to automate tasks like introducing the files and preliminary coding according to case and demographic codes. Or in semi-structured interviews a command file can look for all instances of answers to 'Question 1' and store them under one code. In focus groups a command file can be written to gather together all the speeches of a particular member of the group under one code.

9.4
The drawback of all these complex options in Nudist is that it does take time to learn how to do all of them. It would be over ambitious to expect to be able to use all these functions on the first project without allowing plenty of time to learn. Technology and the new skills that it requires, take time to acquire (Barry, 1997). It would be more realistic on any of these packages, to only expect to use the basic functions for the first few months.

9.5
Having said that, part of Atlas/ti's immediacy is its simplicity of operation. It is easy to grasp what it does and how it works, certainly at the basic level of operation. It is easier to understand what it is trying to do than it is with some of the Nudist operations on first acquaintance. Certainly if you have a simple project and little time, Atlas/ti might be an easier package to learn.

9.6
Atlas/ti, however, does not have so many options for complex analysis. You can analyse data by data types, but there are more limitations than with Nudist . Multiple research teams working on Atlas/ti have to work on the same copy and if members use Apple Macintosh they will not be able to use Atlas/ti.

9.7
I have summarised the software features that support different levels of project complexity in Table 3. In the next section I will tackle how to choose between the two.


Table 3: Features of Nudist and Atlas/ti that support complex versus simple projects
COMPLEX PROJECTS
NudistAtlas/ti
+ Good for different: data types, case
types, longitudinal research, large
samples, mixed qual/quant studies

+ Multiple researchers/multi site projects
can repeatedly merge separate coding

+ Easier to keep track of progress with
project management tools

+ Case management allows analysis of
complex linked case data

+ Sophisticated searching, 18 options,
including matrices comparing text for
sub-categories of 2 codes

+ Command files allow automation of all
functions

+ Can input data in tabular form

+ Transferable Windows to Mac systems
- Only simple / Boolean searches

- Multiple research teams have to work on
same copy / site

- Not Mac compatible

- No option to automate commands except
pattern searching
SIMPLE ROJECTS
+/- Suitable for simple projects,
but lacking ease and creativity
+ Good for straightforward, simple
sample, one timepoint projects

+ Probably easier to learn the basics

How to decide between Atlas/ti and Nudist?

10.1
If you have a clear preference for one type of analysis style or if your research projects tend to be inherently more or less complex, this might clarify the choice. Certainly you should try out both in an attempt to gauge the personality and structure of each and see to which one you are instinctively drawn. Demonstration versions can be downloaded from the Surrey CAQDAS.

10.2
I have summarised the choices you might make using my taxonomy of software structure and project complexity in Table 4. Atlas/ti would be the best choice for simple projects with researchers who prefer to work in a more complex software environment or where the software needs to be learned in a hurry. Nudist works best for those who prefer to work in a sequential structured style, particularly for complex projects. If you wish to combine the two different strengths of the two different software structures I would recommend using Nudist in tandem with other Network software[4], such as Decision Explorer <http://www.scotnet.co.uk/banxia> or Inspiration <http://www.inspiration.com>.

10.3
Inspiration is a mind mapping tool that allows you to draw high quality pictures and diagrams of your data and ideas. These diagrams can also be converted into outlines for writing. It's easy to use and not too expensive. Decision Explorer is a cognitive mapping tool that allows you to map concepts and then use a whole range of tools to explore and analyse your map. It is a powerful qualitative analysis tool that is easy to use but not as good at drawing as Inspiration. It is also quite a lot more expensive. Both Inspiration and Decision Explorer are compatible with Nudist . The hierarchical tree of codes can be exported to these packages and then manipulated and built into complex relationships, and re- imported into Nudist for further coding. They can also be used to structure writing outlines, based on the coding structure, and to produce conceptual diagrams to aid presentation of data.


Table 4. Suitability of Atlas/ti and Nudist for differing software preferences and project types
PREFERRED SOFTWARE STRUCTURE
Interconnected/
visual/spatial
Sequential / structured / verbalCombination



Simple


PROJECT



ATLAS/TI


ATLAS/TI

OR

NUDIST

NUDIST

and
NETWORK
SOFTWARE

TYPE



Complex

NUDIST

and

NETWORK
SOFTWARE



NUDIST

NUDIST

and

NETWORK
SOFTWARE

Chaos Theory and Qualitative Data Analysis Software

11.1
This analysis of software structure raises an interesting issue of which work styles are preferable for a quality outcome. We may currently lean towards one or other type of software structure and this may lead us to work in a particular style. For example I am more comfortable with more sequential methods of data processing, which is why I have found Nudist so useful. However, sticking as analysts to one or other type of software and style of analysis may be limiting in itself. Perhaps the best option is to alternate between both types of software as tools to aid different styles of working.

11.2
Research on problem solving has suggested that there are two elements involved. Even in the last century there was awareness of an alternating conscious and unconscious aspect to mathematical problem solving (Poincare, 1908). More recent work on problem solving lies within the framework of chaos theory. Here problem solving is seen as an inherently chaotic activity with oscillation between two brain stages, one laminar, smooth and predictable and one unpredictable and turbulent.

' The brain searches for the best solution to a problem through a unique combination of order and randomness that optimises the scope of that search.' (Torre, 1995: p. 186)

11.3
People with the capacity to maximise these two different ways of operating are thought to better at solving problems in chaotic systems.

'Within psychology there is a considerable evidence of personality correlates that cluster within people who seem to have more of a capacity to beat the chaos odds than others. Traits such as open-mindedness, creativity, imagination, tolerance for ambiguity, brain hemispheric equipotentiality and systems sensitivity.' (Loye, 1995: p. 351)

11.4
The chaotic social worlds that most qualitative researchers are investigating, require striving to operate at both levels.

11.5
Different software can assist us to do both types of processing. Whilst we can, and do use many different types of mental processing whichever tools we use, we might want to capitalise on the ordered, sequential qualities of packages like Nudist to assist with ordered thinking tasks, like data management, and then use visuospatial, hypertext software, like Atlas/ti to prompt us to think more creatively, conceptually and perhaps more holistically.

Conclusion

12.1
Whichever route you take, I believe that using any software can benefit and enrich the analysis process. It may seem an awesome barrier acquiring the new information skills that the new technology requires (Barry, 1997). There is however, sufficient pay-off in terms of enriched data analysis and more comprehensive development of coherent theoretical ideas. Both packages will help to automate and speed up some analysis tasks, allow you more instant access to data once coded, facilitate more complex questioning of the data, and provide creative aids to stimulate theoretical development.

12.2
I have tried to build up a picture of the different personalities of the two main packages, Atlas/ti and Nudist in order for you to choose from a better informed position. Nudist tends to win out on sequential structure, project management and sophisticated searching while Atlas/ti strengths lie in its inter-connectedness and creative interface.

12.3
Considering software structure and workstyles in the context of choosing CAQDAS may also lead to greater awareness of the need to integrate the different modes of analytic thought into your qualitative work.

Notes

1Some forms of qualitative anlaysis are more quantitatively inclined such as content analysis and counting features may be of use here.

2Hypertext structure relies on non linear links and is best demonstrated by the structure of the world-wide web. Weaver and Atkinson cite Landow's (1992) definition of hypertext : 'Chunks of text are joined together in meaningful ways by electroinc links, which are activated by "soft buttons". The latter offer the hypertext reader various pathways or trails through the text. As such, text in a hypertext sytem is multisequential, multilinear, and multidimensional' (p. 113). You have just experienced a hypertext link if you located this footnote by pressing on the highlighted text of the word 'hypertext' in the main body of the paper.

3This project is funded by the Department of Health. My co-investigators are Professor Nick Barber, Professor Colin Bradley, Dr Nicky Britten and Dr Fiona Stevenson.

4Clare Tagg has written an article illustrating the different uses of Inspiration and Decision Explorer with NUD*IST. Click here to access <http://www.taggoram.co.uk/nudist/nudinsde.html>.

Acknowledgements

I would like to thank Martin Marshall, Exeter University, whose experience with Atlas/ti allowed me to develop this comparison when we developed a workshop for the AUDGP conference. Thanks is also due to David Afia who brought the chaos systems view of problem solving to my attention, and Barry Cassidy who read and gave useful comments on an earlier draft.

References

BARRY, C.A. (1994) 'A Q-Sort Study of Men and Women's Definitions of Morality'. Unpublished psychology undergraduate dissertation.

BARRY, C.A. (1995) 'Critical Issues in Evaluating the Impact of IT on Information Activity in Academic Research: Developing a Qualitative Research Solution', Library and Information Science Research, no. 17, pp. 107 - 134.

BARRY, C.A. (1997) 'Information Skills for an Electronic World: Training Doctoral Research Students', Journal of Information Science, vol. 23, no. 3, pp. 225 - 238.

BARRY, C.A. AND SQUIRES, D. (1995) 'Why the Move from Traditional Information-Seeking to the Electronic Library is not Straightforward for Academic Users: Some Surprising Findings', in D.I. Raitt and B.Jeapes (editors) Online Information '95: 19th International Online Information Meeting Proceedings (pp. 177 - 188). Oxford: Learned Information

BUSTON, K. (1997) 'NUD*IST in Action: Its Use and its Usefulness in a Study of Chronic Illness in Young People' Sociological Research Online, vol. 2, no. 3, <http://www.socresonline.org.uk/2/3/6.html>.

COFFEY, A., HOLBROOK, B. AND ATKINSON, P. (1996) 'Qualitative Data Analysis: Technologies and Representations', Sociological Research Online, vol. 1, no. 1, <http://www.socresonline.org.uk/1/1/4.html>.

GREEN, JUDITH (1998) 'Commentary: Grounded Theory and the Constant Comparative Method', British Medical Journal, vol. 316, pp. 1064 - 65.

HONEY, P. AND MUMFORD, A. (1992) Manual of Learning Styles, 3rd edition. Maidenhead: P.Honey.

ISAAC, A.R. AND MARKS, D.F.(1994) Individual Differences in Mental Imagery Experience Developmental Changes and Specialization', British Journal of Psychology, vol. 85, no. 4, pp. 479 - 500.

KELLE, U. (1997) 'Theory Building in Qualitative Research and Computer Programs for the Management of Textual Data', Sociological Research Online, vol. 2, no. 2, <http://www.socresonline.org.uk/2/2/1.html>.

KLEINMAN, S. AND COPP, M.A. (1993) Emotions and Fieldwork. (Qualitative Research Methods Series, volume 28). Newbury Park : Sage.

LANDOW, G.P. (1992) Hypertext: The Convergence of Contemporary Critical Theory and Technology. London: John Hopkins University Press.

LEE, R.M. AND FIELDING, N. (1991) 'Computing for Qualitative Research: Options, Problems and Potential', in N. Fielding and R.M. Lee (editors) Using Computers in Qualitative Research. London: Sage.

LEE, R.M. AND FIELDING, N. (1996) 'Qualitative Data Analysis: Representations of a Technology: A comment on Coffey, Holbrook and Atkinson', Sociological Research Online, vol. 1, no. 4, <http://www.socresonline.org.uk/1/4/lf.html>.

LOYE, D. (1995) 'How Predictable is the Future? The Conflict between Traditional Chaos Theory and the Psychology of Prediction, and the Challenge for Chaos Psychology', in R. Robertson and A. Combs (Editors) Chaos Theory in Psychology and the Life Sciences.. Mahwah, New Jersey: Lawrence Earlbau.

MARSHALL, M. AND BARRY, C.A. (1997) 'Analysing Qualitative Data using Computer Software Programmes' (Workshop). Association of University Departments of General Practice Annual Scientific Meeting, Dublin, July.

MASON, J. (1996) Qualitative Researching. London: Sage.

MILES, M.B. AND HUBERMAN, A.M. (1994) Qualitative Data Analysis: An Expanded Sourcebook. 2nd edition. Thousand Oaks, California: Sage.

POINCARE, H. (1908) La Science et l'Hypothese [Science and Hypothesis]. Paris: E. Flammarion.

SEIDEL, J. (1991) 'Method and Madness in the Application of Computer Technology to Qualitative Data Analysis', in N. Fielding and R.M. Lee (editors) Using Computers in Qualitative Research. London: Sage.

SQUIRES, D., BARRY, C.A. AND FUNSTON, H.T. (1995) The Use of IT-Assisted Information Systems in Academic Research. (British Library R&D report No. 6215). London: British Library.

TESCH, R. (1990) Qualitative Research: Analysis Types and Software Tools. London: Falmer Press.

TORRE, C.A. (1995) 'Chaos, Creativity and Innovation: Toward a Dynamical Model of Problem Solving', in R. Robertson and A. Combs (Editors) Chaos Theory in Psychology and the Life Sciences. Mahwah, New Jersey: Lawrence Earlbau.

WEAVER, A. AND ATKINSON, P. (1994) Microcomputing and Qualitative Data Analysis. Avebury: Aldershot.

WEITZMAN, E. AND MILES, M. (1995) Computer Programs for Qualitative Data Analysis: An Expanded Sourcebook. 2nd edition. Thousand Oaks, California: Sage.

WINOGRAD,T. (1996) Bringing Design to Software. Stanford, CA: Stanford University and Interval Research Corporatio.

Copyright Sociological Research Online, 1998