First, there are at least two versions of open coding. In the fourth edition of Corbin & Strauss, Corbin presents a version of open coding that looks much more like memoing rather than the traditional process of attaching codes to text. Of course any software package would also allow you attach memos to text, but that would not use very much of what those programs are designed to do

In contrast, Charmaz (20114) generates a free flowing set of "initial codes" along side the original text, and that could be more suitable for capturing in software. The main problem there is that her version of initial coding would produce literally hundreds of codes. Indeed, I have had students tell me, "I generated XXX number of codes, what do I do next?" Unfortunately, the answer is that they should have been refining their codes all along, rather than waiting so long. 

FYI doing a word frequency search would certainly not be a standard approach to GT coding.

Theoretical sensitivity is not lost  because in both cases the codes are generated by the researcher. However, NVivo can enable the researcher to perform many tasks which can not be performed manually .These include; producing visuals showing the number  of times an item has been referred to or mentioned in the study (coding reference bar charts) .NVivo can also express these visually in percentages.In the query mode, it can also visually  show  the most  frequently used words in the study and the contexts in which these were used ,which is difficult to achieve manually.

Dear Lysias 

Thank you very much for your answer. It was very enlightening and useful. 

First, there are at least two versions of open coding. In the fourth edition of Corbin & Strauss, Corbin presents a version of open coding that looks much more like memoing rather than the traditional process of attaching codes to text. Of course any software package would also allow you attach memos to text, but that would not use very much of what those programs are designed to do

In contrast, Charmaz (20114) generates a free flowing set of "initial codes" along side the original text, and that could be more suitable for capturing in software. The main problem there is that her version of initial coding would produce literally hundreds of codes. Indeed, I have had students tell me, "I generated XXX number of codes, what do I do next?" Unfortunately, the answer is that they should have been refining their codes all along, rather than waiting so long. 

FYI doing a word frequency search would certainly not be a standard approach to GT coding.

Professor Morgan

Thank you so much for your answer. Considering Charmaz constructivist approach, the researcher assumes a role in constructing the meaning. Correct me if I’m wrong but by I could understand on your wonderful explanation is, if Charmaz emphasizes that focused coding is the second major phase of coding, you don’t need a very sophisticated software to work with your data because you (and not the software) will have to analyze and give the meaning to the first open codes. In other words, you could even use a word processor to register your first open codes and eventually could come back to do them just to perform the comparisons, as stressed by Strauss. Does it make sense?

In Grounded Theory approach to research, care must be taken not to get carried away with word frequency counts which may  sometimes be misleading. However, I suggest that you also  some contextual mapping of words with high frequency counts. This will give you more clue  to the actual  values of the words within the context of the data you are analysing.  (NVIVO does give a good contextual maps of words)

Dear Yves,

first of all, if you have ever considered manual coding of qualitative data, forget it and use CAQDAS (specialized software, like nVivo you mentioned), it's easier, analytically way better (you can results of your coding instantly and do some further analysis with codes, such as interconnection of codes), and, although it may sound strange, more fun. And from practical point, analysis of a big data corpus is nearly impossible without  CAQDAS.

As for your question regarding grounded theory, this analytical approach is very demanding on a researcher because the ideal outcome of grounded theory should be a new theory or modification of the current one. Change of a theory should done on the basis of knowledge gained by inductive analysis of the created data, so a researcher should create data (interviews, textual production etc.), then analyze it without any initial theoretical approach, create myriad of codes and memos, and finally take together the codes and memos and created or modify a theory which is connected to the researched topic. Of course, people are not robots so there is an influence of their pesonality, experiences, life philosophy, social position, stances etc. However, you, as the researcher, should reflex this influence (and possibly bias) and be aware of it during the analysis and coding. As you can see, the analytical process is very complex and doing it without CAQDAS is, in my opinion, madness :). 

The only difference between "more conventional qualitative research strategy" (as you call it) and grounded theory is the existence of initial analytical theories you approach with to the data in case of qualitative content analysis, and ideally no (or very strictly limited) approach if you use grounded theory. Also, during QCA you (should) have some expectations and qualified guess of what you will find, on the other hand your mind should be completely positioned within the data when you are doing grounded theory. Thus, QCA is more like combination of deduction and induction while grounded theory should be about pure induction. 

And one more thing about coding - it's a multi-layered process which has multiple rounds (Johnny Saldana offers good reading about it) so I recommend you to get familiar with some coding techniques and try to be creative. At the same time, creating and using code book (try to read some texts by Philip Mayring) where there are definitions of what is represented by each code and what passages should go there. Again, within QCA you can create some codes deductively before the analysis based on your experience and expectations but during using of grounded theory the codebook should be created procedurally aout of the data and codes.

Hope this will (at least) somewhat helpful.

Happy coding and analysing :).

Yves asks if there is any advantage to using CAQDAS software for focused coding, as used by Charmaz, and I personally am ambivalent about that. The main thing I can think of is using the search capacities to aid in making comparisons across existing codes. But if you are immersed in your data and your coding process from the beginning, as GT insists you must be, then this should not be a problem.

As for the "madness" of not using software, people did analysis manually for many decades before CAQDAS became available, and the software is itself designed to mimic the same kind of marking and searching processes that were always the basis for this kind of analysis.

Where I find software most useful is when you have a well-defined codebook that needs to be applied to a large amount of data. That does not necessarily describe GT, since your coding evolves throughout the data collection process. 

Dear Jiří , thank you for your contribution. In fact I tend to think about a CAQDAS software as tool to keep the codes. I'm using RQDA to organize the codes and categories. I'm having the same perception at this point of my research(following Charmaz procedures) of professor Morgan  when he says that when you use GT, the search capacities seems to be the main utility  of these software