Starter Codes: A short list of descriptors representing things you expect to find. These could include phenomena, tools, or roles.
Example: You're interested in how people manage projects in high-tech organizations, so you decide that you'll code interactions by categories such as "project management," "leadership," and "project management software," all of which are mentioned in your IRB.
Open Coding: An open list of descriptors representing things you find in the data. Think of these as free-form "tagging" as you might do on a social networking or social bookmarking site (Flickr, Delicious). After coding a few pieces of data, you'll start reconciling these codes.
Example: You start coding field notes and observations, and you quickly realize that much of the project leadership is shared. People agree on a small set of steps at the beginning, but day-to-day, they decide their own tasks and keep in contact via IM. So you tag these interactions with the starter code Leadership, but you also tag them with an open code, say, Autonomy.
Axial Coding: A set of codes that describe persistent connections between some of the existing codes.
Example: You notice that every time you code Autonomy, you also code IM. Maybe they can be autonomous because they are in constant contact? You develop an axial code, Connected_Autonomy, to express the connection, and you see how often that connection exists.