I think that snowball sampling would fit together quite nicely with a grounded theory approach - using these, you would both develop your theory, and identify your sample as you go along. I can't help you with any literature off the top of my head, but encourage you to look into adopting this approach. Just remember to keep aware and engage in good critical reflection as you go about collecting your data and forming your theory.

It should not be snowball sampling per se, but purposive. Remember, you're aiming to create a theory, not get as many participants as possible as such. If your data reaches saturation and no new themes emerge, you can stop recruiting, but if there are gaps in other parts of your theory you need to recruit more people with that in mind, so it is not so opportunistic. That is why you need to use an iterative approach to analysis.

The following journal articles are helpful to your topic: 

• Biernacki, P. and Waldorf, D. (1981) Snowball Sampling: Problems and Techniques of Chain Referral Sampling, Sociological Methods and Research, 10, 2, pp. 141-163.
• Faugier, J. and Sargeant, M. (1997) Sampling hard to reach populations, Journal of Advanced Nursing, 26, 4, pp. 790-797. 

Kenneth

 I don't see a reason not to go for snowball sampling provided that the participants that are referred to you fit your inclusion criteria. 

Laura and Kiran's points of view complete each other to provide the appropriate recommendation. Also I thank Kenneth for sharing the papers.

I agree with Laura and Kiran's points of view. If you cannot readily contact all the subject but will need the support of those identified to reach others then you can used the snowball. 

The sampling in GTM is more complex than simple snowball sampling. Have a look at the work of Janice Morse, who is an outstanding specialist on sampling questions with regard to GTM.

From a GT point of view, I think the relevant concept would be "theoretical sampling." From that perspective, the new sources of data should be chosen according to how they can assist with the development of the emerging theory. In other words, your ongoing analyses of the existing data should help determine the new kinds of data that would be most useful for developing your theory.

For snowball sampling, this would mean not just taking any eligible person that a prior participant happened to know. Instead, you would be asking for the specific kinds of people who have the more explicitly defined criteria that meet your goals for theoretical sampling.

Thanks to all for comments.

Yes, it is well known that theoretical sampling drives selection of new informants in GT. However, the idea with snowball sampling is to collect contacts within the target population, and based on the current needs in developing theory to pursue by selecting the most appropriate among them in each stage of the research (theory development).

Yes, that is just what I had in mind.

If you are in early stage of your research, theoretical sampling will be inconsistent with the tenets of grounded theory. This is because the principles of theoretical sampling would not call for greater numbers to reach a particular sample size, but rather would call for the selection of participants to pursue emerging theory. So You can consider using snowball sampling in the initial stage, and then consider how subsequent sampling could be based on grounded theory principles.

In my opinion, snowball sampling can be used, with the notion that researchers are totally blind about the location of the study (population, category specifications, general even specific information). So researchers need insider involvement (coming from a research location) to help introduce anyone who understands the topic to be studied. So my view, hopefully can help. Thank you