The difference between ATLAS.ti's code families and code links can be 
sketched as follows:
1. by creating a code family you are simply creating a named and commented 
set of codes. Indeed a "family" is equivalent to assigning a dichotomous 
attribute to a code. By making codes members of a code family (e.g., 
"Emotions"), although you are not creating 'explicit' relations between 
these codes you are indeed creating an implicit equivalence relation 
between them. As long as your member assignments are semantically sound, 
you can always expect that they have something to do with "emotions".

2. Linking codes into networks is different. You are not creating sets of 
related codes, but you are creating explicit semantic links between pairs 
of codes. To represent an "emotion" model, you might create a code 
"Emotion", codes "Negative Emotion", "Positive Emotion", and link the 
latter to the former with an "is a" relation. You can proceed with such a 
categorization (which uses 'explicit' links!) until you reach the level of 
codes close to the data. The last level of links - between the codes and 
the data segments - is also implicit and could be termed as 
'is-incident-of'. Note that using this more elaborate scheme, new codes are 
introduced (either in the first place, when your approach is "top-down", or 
in a later modeling stage, when you are working GT like "bottom-up"-wise. 
Furthermore, by using explicit relations you are introducing another 
powerful and separate epistemological level alongside the "domain" level of 
the codes themselves. In the names of the relations, which you can define, 
you represent the "approach" of your enterprise (GT like, argumentation 
analysis, etc.).

Both approaches have their pros and cons.
The family approach is easy to understand and to do. Simply create a family 
and "throw" the codes into this "drawer". Use such families as efficient 
filters for constraining the number of items displayed, queried and 
otherwise processed. Create "super families" to negate or combine such 
categories.

The network approach is more sophisticated both methodological and 
procedural and lends towards a more modelling based approach. Using the 
Query Tool, you can now retrieve all incidences of "negative emotions", 
although this code has no "hard-wired" links to the data at all. In the 
Object Explorer, the links between the codes can be navigated. In the 
Network Editor you can display the model you created by linking the codes.
Indeed, the whole concept of networking in ATLAS.ti has been influenced by 
two sources: artificial intelligence, especially knowledge elicitation and 
representation, and by the Grounded Theory approach.

More information in the manual and online help system.

- Thomas

At 02:35 24.01.2005, you wrote:
>Hi all:
>
>I am a new correspondent on this list . . . if this question is best 
>addressed elsewhere, thanks in advance for letting me know.
>
>I am evaluating Atlas.ti, MaxQDA and NVivo for a project involving the 
>analysis of decades old psychosocial interviews with once-adolescents in 
>an ongoing longitudinal study.  I am doing a thematic analysis, one that 
>is primarily top-down – that is, not GT – at least on my first pass.
>
>I am using Atlas.ti first and have come upon the following question:  I am 
>trying to understand the difference between creating code families and 
>establishing relationships between/among codes.  Under what circumstances, 
>and for what theoretical reasons, would I do the first in Atlas?  The 
>second?  Both?
>
>Many thanks.
>
>Robbe Burnstine
>Pre-doctoral Researcher
>Judge Baker Children’s Center, Harvard Medical School
><mailto:[log in to unmask]>[log in to unmask]
>
>

_______________________________________________________________________
„Computers, like every technology, are a vehicle for the transformation
of tradition“ (Winograd & Flores, 1987)
ATLAS.ti Scientific Software Development GmbH - Berlin - www.atlasti.com
Dipl.-Psych. Dipl.-Inform. Thomas Muhr (CEO)  - [log in to unmask]