A page to keep footprints from an experience of getting to grips with !AtlasTi for interpretive research by a group of field researchers working on the same research programme.
Q: Why Atlas.ti?
The Atlas.ti tool is directly usable to support abductive or grounded approaches to theory generation in interpretive research (see GLASER, B. G. & STRAUSS, A. L. (1967) The Discovery of Grounded Theory : Strategies for Qualitative Research, Chicago, Aldine.). Usage can be directly mapped to the methods of grounded theory including; open coding, axial coding, selective coding, the use of memos and the overarching process of GT.
Flexible analysis and associations can be employed on different layers, by marking quotes, by inventing and assigning codes, and by writing notes or memos. Objects at each of these layers can then be associated with objects in the other layers, although the association between PD and quotes are more intimate than between the others because quotes are direct extracts or parts of a primary document. Each object type is provided with an object manager, a panel which offers a context sensitive mode for accessing, listing, viewing and editing these associations.
Quotations can be inserted into the full selection of media; transcript, text, image, video or other data. All codes and memos are linked directly to quotations. Quotations can be generated automatically as you apply a code or memo to a selected segment of data.
The tool can be used as a way of marking up transcripts, supporting the use of Jeffersonian Transcription for carrying out Conversation Analysis to annotate speeches, speech, performance, observed events, acts, texts (see ATKINSON, J. M. & HERITAGE, J. (Eds.) (1985) Structures of social action: studies in conversation analysis, Cambridge University Press.).
Importantly Atlas.ti projects are readily shared between researchers collaborating on the same studies and can be used to facilitate a common approach to coding during analysis. For example a group of researchers working on shared material might consider agreeing protocols for naming the codes and other objects during the process of analysis and reflection.
Atlas.ti provides excellent electronic support for interpretive research by accepting multiple document types (video, audio, texts (including doc, rft, plain text, html and others) as primary documents (PD).
- HU : Hermenutic Unit, a storage unit which organises all files and objects associated with the research programme.
- OM : Object Manager, a generic interface to managing the
objectsin Atlasti, i.e. !PDs (or P-Docs), quotes, codes & memos.
- Object Explorer : This is a kind of root level object organiser for all objects in your HU, it includes the !PDs, Codes, Memos, Primary Doc Families, Code Familes, Memo Families and Network Views. You arrive at the Object Explorer via the menu Tools | Object Explorer
- PD : Primary Document
- Families : You can organise your objects into familes of objects, this offers a way of creating your own hierachy of these objects, usually because you have so many they need to be grouped into agreed classifications, e.g. male respondents, media source etc.
- assign : adding objects in the HU
- code : both the
codeobject in the Atlas program AND the act of coding your data
- network : a flexible graphical representation of the relationship between objects in Atlas, automatically generated links are generated between for example a code, the corresponding quotation(s), and the primary document(s) the quotations reside in. Other kinds of relationships defined by the research can be drawn (i.e. draw lines or connectors) between pairs of objects.
- quotation : a selection of transcript, text, image, video or other data, you can't code or add memos without quotations, quotations are often generated automatically as you apply a code or memo to a selected segment of data.
- annotation : A way of marking up a transcription. JeffersonianTranscription is a scheme employed in the Conversation Analysis community for annotating speeches, speech, performance, observed events, acts, texts...
Running Atlas for the first time.
After installation, the usual trepidation and uncertainty as you approach a new application for the first time. Trepidation because you fear you won't "get it", won't get the big-idea about which this package is all about. Uncertainty because you are thrust back into the position of rank beginner, a complete novice in some new skill or knowledge, returned (unwilling it seems) to an open-ended formative position, personally torn between total ignorance of something encountered against the need to master and overcome the unknown.
Take the quicktour and the sample project
What is the central metaphor embodied in the design of this product? Consider it an environment within which to generate or create loose associations between artefacts. The core artefacts are !PDs (texts, images, video, sound recordings). Analysis and associations can be employed on different
layers, by marking
quotes, by inventing and assigning
codes, and by writing notes or
memos. Objects at each of these layers can then be associated with objects in the other layers, although the association between PD and
quotes are more intimate than between the others because
quotes are direct extracts or parts of a primary document. Each object type is provided with an object manager, a small window which offers a context sensitive mode for accessing, listing, viewing and editing these associations.
The Object Explorer is a kind of
root level view of all the objects stored in your hermenutic unit, allowing you to browse the HU (conceptually an object repository), to browse and apply filters on the entire HU, and finally to access the
network views you have created. These
network views provide a graphical representation of relationships generated or those you have established between objects.
Jump in and Code
Start with a new HU and create a new Memo. Change the title of the memo, replace ME with a meaningful title, for example
Research Process Journal - 20/04/06, select the tick icon to save, use the insert menu to insert the current date and time. This memo could be used as a daily journal for your engagement with the data in your HU.
Starting a project
You'll create a new HU (Hermenutic Unit, a storage unit) for the research project. The first lot of toolbar items deal with the project at the HU level. The second lot of toolbar items relate to the objects that you add to or create in the HU, these objects are p-docs, quotes, memos and codes.
Having saved your HU (the ATLAS title bar displays something like HU:!CaseACS you need to add (assign) at least one primary document. Do this through the menu or toolbar, i.e. Edit | Assign, or click on the P-Docs toolbar icon. A comprehensive list of document types are supported (tiff, doc, rtf, html, mpg, avi, jpg, mp3, bmp, wav, wma, wmv, au, snd) so it should be possible to import or convert to import most kinds of electronic media.
Once a primary document is assigned we commence a process of interpreting and coding the material, I guess this could be termed the engagement in a practice of qualitative research through a reflexive process of reading, interpreting, marking and coding. An enacted process of investigation, analysis, identification, relationship linking, this is inevitably done in several passes over a period of time, in isolation or with others in a dialog. The authors of this coding and relationship building themselves create an interpretation or a mutually constituted reading which may either extend or distill the primary material, a kind of process of transformation which in turn gives rise to interpretations which might be reified as other objects for subsequent use.
One possible way to commence working on material derived from field work or desk research as soon as possible as follows; Compile a single document which containing interview transcripts or press clippings (for example) on the subject of investigation. Add a single PD (e.g. a press clipping's document) and conduct a
first pass coding exercise on the material in the document. The code manager collects the results of our first attempt at coding and the PD can be released (or disconnected from the database, from PD Manager | Documents | Disconnect) we might now re-enter the relevant press items as separate PD's and associate the quotes or entries with the codes devised earlier. Q: Eoes this approach unnecessarily duplicate effort? What's the benefit of doing things like this?
The user should become aware pretty quickly what is happening
under the hood when they jump in and commence coding the PD immediately. Coding implies the direct identification and marking of quotes, that's the relationship, code maps to a part of the PD, which is in turn identified as a quotation and will appear in the quotation manager.
- b. A word of warning for the unwary, when managing the relationships between codes and quotes, never choose to
deletethe association between a quote and the code, always use
unlink, this way you avoid the possibility of accidently deleting your code (and its associated properties).
The PD viewer can zoom the text of a PD, nice to zoom in and out to see how the coding, memoing is going.
Drag and drop works as expected - but only if you select text or media in the PD viewer with which to create the association to code or memo, this could work better for the situation when you don't have material selected in the PD view and nothing happens.
The environment really needs a large (very large) monitor to work efficiently. I evaluated on a 1024x768 size screen and it really wasn't big enough to work in. To work efficiently you will need all viewer windows open at the same time and visible (i.e. not overlapping) to really get a sense for the
shape of your data, the associations and relationships between PD, codes, quotes, memos, network views etc. 1920x1200 works reasonably well in this regard.
Can be used with the following primary types, e.g. codes, documents, memos. Think of families as another level at which you can organise these objects in some useful ways, e.g. organise interview transcripts (primary documents by gender, age, role), or organise codes by level of analysis (e.g. respondant answers to questionnaire by Q1, Q2)
Document Family Manager
Documents>Edit Families>Open Family Manager
Creating document families allows code manager reporting document by document (or document set basis) rather than over all the documents.
Outputs can be generated from any of the object managers, they are reports on the currently retrieved objects (that is you may have a filter running - biege). From the code manager most of the output reports are based around the primary document, if you want to see all quotations organised by code rather than by PD then select
All codes with quotations.
With the PD's added to the HU. Go to Documents>Miscellaneous>Word Cruncher and tweak settings as required. This gives you a csv formated excel spreadsheet counting the number of times each particular word in your text files occurs. The count can be over a single PD or over the entire set of PDs. You might then review the generated csv file, for example adding a totals column, sorting by total, excluding common grammatical elements which might be independent of the phenomena you are researching, e.g. AND, OR, A etc. If you do this a lot then edit your Atlas installation's
stoplist, a text file located at Application Data\Scientific Software\ATLASti\stoplist.txt
Having produced a satisfactory report you might then save a copy as a memo into your HU. Give it a meaningful title, and it is stored now as a memo of record which might be useful if you employ it as part of the body of evidence you are establishing. Another alternative is to insert the spreadsheet into an RTF file; Warning! Is it possible to manipulate the spreadsheet subsequently in an RTF PD and avoid that Atlas thing where it
forgets the PD?
Hyperlinks can be created in either the default PD view or even better, through a new network view. It is nicer typically to use a network view because of the nice context sensitive pop-up list of link types.
You can also right-mouse on a link and open the Hyperlink Relations Editor for added control; e.g. create own relationship types which may be specific to your own field work.
Backing up and moving the HU around
In the knowledge that we have determined on a well known location (folder) for the working HU and the associated data files there are a couple of approaches available for managing backups.
- In the first instance you merely need a backup copy of the *.hpr5 document for your project. Store a recent copy in another (well known) location on alternative media.
- A more comprehensive form of backup is available via the Tools menu. Choose Tools | Copy Bundle | Create Bundle to create an archive of the entire project including primary documents. Use Install Bundle to restore or migrate a HU. The bundle approach is purely for backups, not managing a multi-user installation.
Team working -
To share or not to share
The key question is, are there absolutely compelling reasons for a team to share all the material and all the codes and comments and memos and networks etc. If not then it may be feasible for a team to work independently on their own separate !HUs.
There are two main approaches a team can take, they can 1) split the data, or 2) split conceptual areas. Split the data corresponds to differing documents sets but shared coding schema. Split conceptual areas describe the situation where the team shares common documents but utilises independent coding schemas.
Merging !HUs together into a consolidated central repository when team working or collaborating. See the on-line help Collaboration and Project Management topics. From the main menu select Tools | Merge with HU to open the merge manager.
A Common Approach to Coding
A group of researchers working on shared material might consider agreeing a shared approach to naming the codes and other objects. This could be simply a prefix system or similar, just so long as we have a shared understanding of how it works and any percieved benefit.
Do include an ID as an indication of what types of
things the object they name actually represent, for example:
What names to use for codes, file names, protocols for use of memos (i.e. naming them!), use of comment field (displays in lower pane of the family manager).
Standard preparation of documents: spelling (English vs American), spaces, title, date-time, speaker or informant identification, header, hard returns, font characteristics colour in text, naming of informants.
Agreed and/or documented prefixes and names are a really big thing, for filenames, memos, codes, the special use of Capitalisation, user initials, agreed format for dates (e.g. yymmdd).
m: milestone, something like a noteworthy event fixed in time
a: actant, an actor or object which appears in the field, referred to by others and attributed with effects and identity
t: translation, perhaps when translations between actors and/or actants are observed or claimed?
A regime for working together
Ultimately the group of people working together (and possibly employing Atlas) must establish a viable way of working which facilitates the production of their goals. How do we manage this process and what artefacts selected to support the process.
Working with Multimedia
Just be aware of the distinction between media and textual content. The same metaphore of quotation-code applies to media as to text, the difference is that the main window is unable to present which parts of the media are quoted and which is not.
Tip: When marking quoted sections of media (video and audio) the quotations are auto-entered with the filename of the media file, the tip is to rename the auto-entered file in the quotation manager giving it a meaningful title.
Working with network views
You may open a network view on ANY object (PD, code, quotations, memos). It is relatively easy to expand from a single object by bringing in the other objects it is connected or related to, e.g. when you select an object in the network view, right-mouse and select
import neighbours, objects which are immediately adjacent to the first (connected) now appear in the network view with corresponding linkages.
Working with Queries & codes:
Codes and therefore the results produced by codes can be aggregated through a variety of means, e.g. merge, code families, supercodes (through Query Tool). The
* is automatically prepended on the new supercode to indicate that this is a super code. A supercode is not a static
code, it is a search expression waiting to be rerun, the results will then be dynamically generated. These results are not the same as the list of coded quotations provided through the code tool, however you can generate such a
coding by selecting (from the Code Manager not the Query Tool) Miscellaneous | Create Snapshot
Tip: Update a snapshot name with the yymmdd-id of when it was run. This is important because a snapshot is just that, a snapshot in time of the data you have and its state at that time.
The Scope option (from within the query tool) overlays the query tool and allows you to dynamically apply bounds based on primary document or PD families, notice that the query tool window (operating behind the
Scope of Query window) displays the refined results with the scope of query criteria applied, n.b. use the printer icon from the Query Tool window to output a report of these results.
Tip: Consider rarely deleting a code, rather, if you feel a code is redundant or useless then prefix it with
z, this renaming shoves the unneeded code down to the bottom of the default sorted view of the code list, out of sight - out of mind, but available later on if you needed it.
The really good thing about using the query tool and is that superfamilies produce results which are regenerated whenever you use them, based on the current behaviour of each of its sub-codes.
Finding co-occurance of codes.
COOCCUR is a good one to get your head around. UP, DOWN, SIB. help to identify relationships between codes which are connected by transitive links (e.g. causes).
- Scientific Software's Atlasti website for support, product info, pricing, community membership.
- The Getting To Grips with Atlas.ti workshop held in the UCD School of Business, Quinn Building, from 2006-04-20 to 2006-04-21, presented by Christine Silver from the CAQDAS group at the University of Surrey..
- The CAQDAS Networking Project managed by Ann Lewins at the University of Surrey. Read the working paper by Ann Lewins & Christina Silver reviewing software support for interpretive research field-work and field-data tools, particularly from the perspective of a group collaborating on the same programme.
- Online QDA at the University of Surrey are pretty much the same group as CAQDAS, but pitch entry level training information to novice users of qualitative researcher tools.
- AIS qualitative research site for notes on software tools for qualitative research
- Digital Insight is a research programme coordinated by the Wisconsin Center for Education Research (WCER) and other partners, part of this effort is devoted to supporting the development of an Open Source package (Transana) for transcription and data management of video and audio data for interpretive research. See www.transana.org
- ESRC National Centre for Research Methods
- Qualiti at Cardiff
Successes? | Work changes? | Attitude changes? | Experience of introduction? | Opinion practices static or moving? | systems or structures necessary?