Audience Dialogue

Software for qualitative research

Where statistical software manipulates numbers, qualitative software manipulates words. However, with qualitative software, much more user input is needed. You usually have to go through the text, code it, and enter these codes into the computer.

The qualitative programs vary a lot more than the statistical ones do, but in general, the way they work is that you feed in a lot of words - such as transcripts of group discussions, or of open-ended interviews. This mountain of text then needs to be coded.

The coding is done manually, and it can be a slow process. You then run the program, which sifts through and summarizes the codes. These programs don't really analyse the data. Computers are too dumb for that - they mostly sort it and group it, so that similar statements appear next to each other. The output often takes the form of a structure, which can be shown as a tree diagram.

Using these programs can give you a much richer understanding of your audience than anybody but a statistician can gain from a statistical program, but that understanding doesn't come easily. It takes a lot of time. I find that much of your understanding builds up while you're coding the data. Though the software doesn't save you much time in analysis, it can be a big time-saver when it comes to presenting the results. It also helps you think about the structure of your data, by forcing you to categorize.

A very useful resource for qualitative software is the book Computer Programs for Qualitative Data Analysis by Matthew B. Miles and Eben.A. Weitzman, published by Sage Publications. This is not a book that's quickly absorbed, because of the wide variation in qualitative software. Unlike statistical software, where a dozen packages all try to do the same thing, but in slightly different ways, no two qualitative programs take quite the same approach. By software standards, this book (published 1995) is now old, and some of the programs described have changed substantially. There was some talk of a new edition, but that was a year or two ago, and nothing seems to have happened yet.

Specific software

This section is a guide to some of the widely available software packages that can be used for qualitative research. It covers:

1. Entering data - by typing or talking

2. Analysing data

Word processing software
Software for linking and grouping concepts
Free-form text databases
Text-counting software

3. Presenting data

Written presentation
Multimedia presentation

4. Links to other websites

The section on graphical presentation grew up, and changed its focus, so now has its own page: concept mapping software.

Qualitative data entry

"Qualitative data" is another name for words. Vast numbers of them, perhaps, and perhaps arranged in a complex structure - but still just words.

The data entry software for qualitative data is usually a text editor (such as Wordpad in Windows-based systems), or a word processor, with data typically saved in ASCII format.

However, if your data is pre-structured (e.g. your respondents have each filled in a questionnaire, and each questionnaire asks the same set of open-ended questions) it's quite feasible to use a spreadsheet program, such as Excel, for data entry. Even though spreadsheets were designed to use numbers, they work quite well for qualitative surveys. For example, each line can represent one respondent, and each column the answer to one question. See our page on content analysisfor examples of using spreadsheets in qualitative research.

Speech recognition software
If you have long interviews or discussions to be transcribed verbatim, and you have a fast, recent computer, and you are a slow typist, consider using one of the newly improved speech recognition programs. There are three main competitors, most of which come in several different varieties (of steadily increasing power and cost).

With these, you "train" the computer to understand your voice, then speak into a microphone. (This is how the ads express it - but in fact the computer is training you!) The words appear on the screen, and you can type in corrections at any time. I know people who are using this type of software for data entry of group discussions. The trained computer understands only its master's voice, so the procedure is to play back a sentence or so of a taped interview, repeat it into the microphone, listen to another sentence from the tape, and so on.

Speech recognition software is progressing rapidly. Since I first wrote this page in 1998, the software has greatly improved in accuracy. They used to claim "98% accuracy". That which sounds great - but it means that one word in 50 is wrong, and unless you remember the interview you can't check the transcript without listening to the tape (etc) again. Amazing mistakes get through. A colleague of mine, telling Dragon Dictate the word "Hegelian" found that it was recorded as "Hague alien."

You need a recent, powerful computer to make this software work, with lots of memory and a huge hard disk. But any computer built in the last couple of years should be fine. Until recently there has been no voice recognition program for the Macintosh - which is strange, because the earliest Macs had speech synthesis software (in the other direction: text to talk), but now Dragon Dictate is available in a Mac version.

2. Software for analysing qualitative data

Like statistical software, qualitative software can take a lot of time to learn. If you're going to spend years working on your PhD, it will be time well spent. But if all you want to do is a simple study for your radio station, the time you'd spend learning to use something like Nud*ist(more below) would be disproportionate.

Unlike statistical software, which does analysis for you, qualitative software doesn't actually analyse the text. It simply makes it easier for youto do that. Many experienced qualitative researchers manage perfectly well without such software. Instead, they use scissors, sticky tape, large areas of floor, index cards - plus plain old insight. That insight is gained by immersing yourself in the data, till you know it practically by heart. Software can impede you, wasting your time, e.g. fighting your attempts to reformat text.

One key difference between qualitative and quantitative (statistical) research software is that the statistical programs are all similar. If you understand statistics, you can use a dozen programs interchangeably - if their file formats allow. But with qualitative software, no two programs do quite the same thing. We classify these into several groups:

Let's look at each group in turn...

Word processing software

You can do a lot with a powerful word processor, preferably one with outlining ability, word counts, pattern-searching, and ability to sort selected paragraphs. Word, Ami Pro, WordPerfect, and StarWriter are all widely used. These days, we use Word, simply because everybody else does too. Its power makes up for its many annoyances. Its constant attempts to be helpful waste more time than they save - and take forever to disable. (Word reminds me of a very willing, but very dumb dog - so much for computer "intelligence"! (A hint: to retain your sanity, go to the Tools menu and turn off all "auto" options when using Word for text analysis. They might save time when you're actually writing, but when the text is all there, auto-anything is a potential cause of trouble.

On a Mac, Word is the same as on the PC. Word Perfect has serious screen display problems, and Ami Pro isn't available. Appleworks is excellent in some ways, thoroughly annoying in others. Other possibilities are the powerful text processor Qued/M - if it's still around. Nisus can do anything - after you've spent a few months learning to use it. Our favourite, though, is BBEdit. With its powerful search/replace facilities, and its sorting ability, it does most of the essential work. And if you pretend you're writing HTML for a web page, you can even have different coloured fonts.

Apart from those general-purpose text-handling programs there's a wide range of special programs for qualitative research. Unlike the statistical and survey tabulation programs listed on our statistical software page, which all do more or less the same kind of thing, the qualitative programs all tend to do different kinds of things. Here are some that we've tried, and found helpful.

A type of software that fits between this category and the next - not really "word processing" in the usual narrow sense - is software for working with transcriptions. There's Soundscriber (free software by Eric Breck, for Windows only), which that makes it a lot easier to transcribe and label speech.

Software for grouping and linking concepts

You can think of this as the next level down from the graphical software just mentioned. With this group of programs, you're deeply immersed in words. Typically, you do a set of in-depth interviews, transcribe every word, and feed the transcript into one of these programs. You can then label the interviewees' statements, group them into categories, and draw conclusions about what your respondents have said.

Nud*istis perhaps the most widely used of this group. (In case you were wondering, the removal of clothing is purely metaphorical.) It specializes in manipulating words and text, and has the most powerful set of searching capabilities we've ever seen. QSR software, the publishers of Nudist, have a lively listserv email discussion group, which you can subscribe to through their web site. If you have hundreds of pages of text to make sense of, Nudist may be what you need. Like many of the other programs described here, it's not something you learn in an hour or two. There's a variant of Nudist, entitled NVivo, which seems to be aimed mostly at researchers who use grounded theory. Unlike Nudist, it handles RTF files, which allow different type fonts, colours, and so on. However its file sizes are more limited than Nudist's. A good way to learn Nudist (and NVivo) is to take a course. Usually these courses last several days: enough to get you started.

Two other programs similar to Nudist are Atlas/TI and The Ethnograph. Atlas/TI extends Nudist's hierarchical data model by allowing links between related data. We've never used The Ethnograph, but it's obviously very highly regarded by its users (always a good sign). A recent addition to this type of software is Xsight, a new package from QSR, the makers of Nudist. It's designed for market researchers, for analysing open-ended text data, and should be much quicker to learn than Nudist.

All of these programs (except maybe Xsight, which we haven't tried yet) are complex, not the sort of thing you'd master in a day - so when you have problems, a strong user group can be very helpful. The implication is the choice of software may be less important than the existence of a network of others who can help.

CDC (the makers of Epi Info) produce EZ-text, free software which is like a simplified version of Nud*ist. It seems quite good, but we haven't used it on a real project yet.

There's also a lot of text analysis software around, which focuses on much smaller units of text. This is used more by linguists more than by social researchers, but it has some applications to qualitative research, particularly content analysis. An example of such software is VBPro.

See also our introduction to content analysis.

Free-form text databases

Two database programs designed for free form text could be useful for qualitative research. These are AskSam and InfoSelect. InfoSelect handles scattered small chunks of information, so it's good for unstructured research - the preliminary stage, when you are collecting scraps of information on a new research topic, and don't yet know what to make of it all.

AskSam can handle either free-form text, or text in fields - which is more useful when you have a good idea of the concepts you are dealing with. A mildly disabled demo version is available, so you can download it and see if it suits your purpose.

Text-counting software

This group of programs is designed for use by literary analysts, for settling questions of disputed authorship and the like. Did Shakespeare really write "his" works? Could Bacon have been the real author? Stylistic analysis can offer clues.

Other programs in this group are designed for use by linguists and educational researchers. A good example of this software is TACT: a set of shareware programs for analysing texts, designed for use by literary scholars, but also usable by qualitative researchers. (Their approaches can be surprisingly similar.)

There are many more programs designed for qualitative data analysis, but we haven't used most of them. An excellent book by Weitzman and Miles entitled Computer Programs for Qualitative Data Analysis describes a wide variety of them. Though it's now outdated (published back in 1995) a new edition has been rumoured to be on the way - for several years. Maybe, one day...

3. Software for presenting qualitative data

Qualitative data is presented in two ways: in writing and in graphics. Of course, you can also talk, but you don't need software for that. And it's also possible to use software to create audio or video presentations - Powerpoint, Quicktime, and the like - but those are beyond the scope of this page.

Written presentation

Word processing software is the mainstay for this task - but full-strength word processors, such as Microsoft Word, have a lot of capabilities that many users don't know about. Here we suggest a few aspects that might help.

The practical problem, when you're producing a report or presentation on a qualitative research project, is the sheer volume of words that you have to reduce. A modest qualitative project might involve interviewing 20 people, for about an hour each. As each hour of conversation averages 6,000 to 8,000 words when transcribed, 20 one-hour interviews could produce about 150,000 words. If this was printed in the format of a novel, it would have about 500 pages. Unfortunately, interview transcripts aren't as interesting as novels. They have no plot, they're repetitive, and they're full of inaudible mutterings and half finished sentences.

If you produce a report or give a written presentation, the patience of your audience will run out after perhaps an hour. This means that you have up to 8,000 words (depending on how fast you talk) to describe the research process, give your conclusions, and present data to back them up.

Therefore, the main problem with qualitative research is how to pick out the relatively small amount of data you can actually use in a presentation. This is where two little-used features of word processing software come in useful: outlining and hyperlinks. Microsoft Word (and its high powered competitors such as OpenOffice Writer and Nisus) can all do this.

Outliningis a way of "collapsing" paragraphs of text into their headings and first lines. It lets you take an overview of a long document, and you can easily move paragraphs and sections around so that the order makes more sense. An outline is like a detailed contents page in a book, showing chapters, sections, and sub-sections.

Hyperlinksin word processors work like the Web. You create a hyperlink, then the readers can click on it and be taken to another part of the same document - or a separate document. In traditional terms, a hyperlink is very like a footnote.

When you use outlining, you create a hierarchy in the document, so that it could be shown as a tree diagram. Hyperlinks are more like cobwebs covering the tree. Imagine you're an ant: if you want to get from Branch A to Branch B and you're using outlining, you can crawl down to the main trunk, then back up again on branch B. (You are not allowed to be a jumping ant!) But if you are using hyperlinks, you can find a convenient cobweb and ant-walk across that straight to Branch B.

So outlining and hyperlinks are not substitutes for one another, but they work together very well. This is the strength of Atlas TI): it uses both together - its ants can jump.

Another feature of modern word processors that we find helpful - but few others seem to bother with - is to use the ability to have different sizes of type. If something seems important, put it in larger type. If it seems unimportant, reduce the font size. This is another good way of making sense of your data. Yet another creative way to use a word processor is to make use of different type colours, or coloured backgrounds. Selective indentation and the use of bullet points can be helpful too.

Multimedia presentation

Recently we've been compiling presentations on CD-ROM, for distant clients. It's surprisingly quick (if you have Web development skills) to make a multimedia CD. We've been using an HTML editor to produce the index page and linking pages, Powerpoint as presentation software, Quicktime to produce videos and audio soundtracks, and Acrobat to produce printable reports that look exactly as you designed them. It's not hard to add audio and video components to a Powerpoint show: a stamp-sized window can show the speaker, while playing back his or her voice, synchronized with the words and diagrams on the slide.

In case the recipients of the "report" don't have the necessary software, there's enough space on a CD to include the various free players, for Quicktime, Powerpoint, Acrobat, and so on.

One of the most enticing uses of video is to add colour and specificity to respondents' comments. However, if you've videotaped a group discussion or workshop session, sending off a CD of this raises some ethical questions. You can't control who might see the video. What if some unintended recipient sees a video on the CD, where a participant who might be identified makes a statement that might harm them in some way? For that reason, it's advisable not to show participants' faces - at least while they're saying anything that has the faintest change of rebounding on them. As the researcher can't always know what the consequences of a seemingly harmless statement might be, it's safer not to show the faces of respondents, or to separate the soundtrack from the video. Ways of not showing faces include having a video camera very high in a room, so that when people are sitting around a table the video shows mainly the tops of their heads. Even safer is to have a different person (on the soundtrack) read out the statements that respondents made, so the statement can't be associated with a perhaps-distinctive voice. Be very careful, because CDs last for years, and you'll never know who might see them. It would be better to put that video or audio on a limited-access website.

4. Other sources on qualitative software

These few links are to selected sites that are kept up to date, and also link to a wide range of other sites on software for qualitative research. Sites selling a single product are not included here.

A very useful link is CAQDAS: Computer-assisted qualitative data analysis software).

ISWorld: a site for people doing qualitative research into information systems - also useful for anybody else doing qualitative research.

Bobbi's Place at lots of resources on qualitative methodology.

Resources for qualitative researchers: Judy Norris's site at the University of Alberta.

Information on qualitative data analysis, softare, and references. James Drisko's site, from Smith College in the US.

Harald Klein's site that deals with analysing the content of human communication - mostly text, but also software that handles audio and video coding.

Related pages, on this site:
Content analysis software | Concept mapping software | Statistical software



If you value this content why not give back by contributing to the site so we can keep researching, testing and updating the information and answering questions?

You can do so by writing a review of this site/page on your website/blog and linking to us (let us know so we can thank you), or making suggestions/providing sources to help us improve its value to visitors, including by completing our survey, or making a donation via PayPal, any amount will be appreciated, just click the button at right.