Dear Andy,

your question is a little bit more complex than trying to distinguish condition A from B...
I cannot comment TDT, having never used it, you should probably contact the authors/developers with your question.

Are you interested in each subject's perception? Or a the group level?
With the former, it would be like trying to categorize each event X individually, based on the series of events A and B for that subject.
With the latter, you would summarize the response to events A/B/X for each subject and check how X is classified in each subject given the perceptions of A/B from all the subjects.

With PRoNTo, both are doable but will require some tweaking of the current version, as only standard cross-validation schemes are implemented.
The next release ("coming soon" for quite a while now, hum hum) should offer some more flexibility in term of training and testing (kind of flexible cross-validation) and would let you test what you're interested in.

[Start self-advertisement]
This paper could interest you: http://orbi.ulg.ac.be/handle/2268/174912
We did something similar to your question to assess working memory
[End self-advertisement]

Good luck!
Chris

---