Thanks for the OOW article!

So let's talk through the question of exclusive and nonexclusive coding.

Exclusive coding is typically used when you want to categorize something based on a single scheme. The underlying assumption is that the categories are either orthogonal (they can't be both) or hierarchical (in a case in which they are both, one dominates). 

• Orthogonal ex: We'll code this word as a noun and that one as a verb.
• Hierarchical ex: This person has a BA, an MA, and a PhD. We'll code them as a PhD.

Exclusive coding also has the advantage that researchers can use Cohen's Kappa to calculate inter-rater reliability. For instance, if you and I both code a list of people based on their highest earned degree, we can calculate to what degree the two of us categorize the same person as having a PhD.

But exclusive coding is often applied poorly. For instance, I saw a paper in entrepreneurship studies that coded management messages by rhetorical appeals. They concluded that management usually started with pathos, then went to logos, then ended with ethos appeals. The underlying assumption was that these appeals were orthogonal—an utterance couldn't be categorized as more than one appeal. We in rhetorical studies would disagree and suggest that exclusive coding is inappropriate here.

Notice that exclusive coding is usually (always?) based on existing categories and assumptions. It's top-down.

Non-exclusive coding is typically used when you're not applying orthogonal or hierarchical codes. For instance:

• Coding an utterance with multiple characterizations (logos and ethos)
• Coding an utterance with multiple overlapping categorical schemes (for instance, logical appeal + audience + tense)
• Exploring data vis-a-vis grounded theory, when you're just trying to hang different descriptors on the data in hopes that patterns emerge

Non-exclusive coding facilitates bottom-up exploration. It's also not tractable to inter-rater reliability. For that reason, it's common to see people who use this coding scheme discuss how they generated codes collaboratively. Here's an example from Propen and Schuster: "We each then developed inductively codes, or categories and subcategories of topics, from the samples, models, and heuristics. We then developed a merged list and recoded across the entire data set to create a comprehensive list of features of the VIS" (p.12)

With that in mind, I don't think the mix of exclusive and non-exclusive codes is that big a deal as long as we briefly address what purpose each one served. I also don't think it's necessary to calculate Kappa for the exclusive codes -- because even the exclusive codes are there to help us map the genres. 

More broadly, I don't think we'll need to spend a lot of time justifying this approach. Non-exclusive coding is a broadly used approach in writing studies and I don't think readers will bat an eye when they see it being used in this way. (Clark's integrative literature explains in detail -- and frankly probably overexplains -- in part because the editor of IEEE TPC at the time was a bit of a micromanager. Remind me and I'll tell you some stories on Thursday.) 

Hope this helps, but hit me back if not. CS