Hi Barbara, you've picked a tricky topic and one which we are still some way from resolving. I'd like to make some observations on the paper and make an attempt to answer your final question.
If I've got it right, you're assessing whether or not we can resurrect CBI through the implementation of aspects of Merrill's Component Display Theory (CDT) and the concept of instructional transactions. My answer is a definite YES, but with the rider that we cannot do this if we continue to apply the traditional classroom or paper-based instructional methodologies to interactive learning applications. In other words (and this is where your criteria continua come in), we need to think more carefully about how we present instruction through the interactive medium. Although we've been doing it for some 30 years, there would appear to be only limited examples of truly creative techniques which are motivational, attention getting, interactive and effective.
Merrill's "pearls of wisdom" are very important, and I, like many others, have always strived to implement dynamic interactions between the instructional material and the user. However, as I've observed before, inaccurate implementation of these "pearls" and CDT elements has resulted in the exact opposite of dynamic instruction.
You introduce the notion of interaction, which I believe is critical to any computer-based instructional system, although it would appear that it is not clear what interaction is. From my recent research, I would argue that it is intricately linked to elements of learner control, navigation, interface, and interactive level (see Schwier & Misanchuk, Interactive multimedia instruction, for a taxonomy of interaction), and that the criteria you define are also critical. However, rather than seeing a transaction as consisting of three criteria, I would see the transaction as being the interaction, and therefore the transaction could also be seen as a continuum from electronic page turning to virtual reality and beyond (wherever that may be).
With reference to your comments on the criteria, it would appear that with (1) there can be no guarantee that we actually cause the stimulation of higher levels of mental activity as we are not there to witness it--unlike the classroom. So, we need to specify potential strategies which will elicit this response--which is assumed to be "better." Criteria (2) focuses on the above comments on interaction. In a paper presented at APITITE'94 in Brisbane, I identified some levels of interactivity which were defined by a combination of response and mental processing (cognitive load). While these levels need refining, I think it is very clear that the way in which students process information through interactive instruction is critical to the success and effectiveness of the product. Criteria (3) is based on the computer's response, and I would argue that branching is something that occurs between transactions rather than within, and is more a function of learner control variables. On the other hand, the technique of comparing a learner's response with those of an expert has proved successful (we did an interactive video study in 1986 which verified this) and can also be used so learner's can compare each other's responses--so we get the implementation of self-evaluation rather than simple answer matching, which presumably is way to the right in criteria (1). It is also great to see that your class is not following guidelines to the letter--we need lots of creativity and imagination to take full advantage of the technology--which is quite different to that when Merrill was working with TICCIT!! So one question is whether the guidelines specified by Merrill can be directly applied to IMM or if they need refining and modifying?
I agree and disagree with your conclusion that there are three conditions required for a transaction, the components are necessary alright, but I wouldn't want to introduce a "checklist" for every transaction within an interactive application. I think I said before that a transaction can be equated to an interaction (or series of interactions).
So, what is a transaction? Well, I see the transaction as being a discrete piece of instructional or information content with which the student must interact, and which is essentially under instructional control--that is, the interaction required of the learner has been defined by the developer/designer and must be completed (or skipped if that is an option) before the next transaction is introduced. I would argue that transactions are separated by branching options, which may be program or learner initiated. A transaction could also include feedback or updates depending on the instructional strategy employed. I also see the transaction as independent from the controls provided to navigate, i.e., the user navigates among the transactions provided in the application.
As an example, a very simple transaction would be the presentation of a specific piece of information which "disappears" after the user elects to proceed; a more complex transaction might be the same piece of information which remains on the screen until the learner has made a response to a question or completed a task (e.g., constructing an object from its component parts). A highly-complex transaction might require the learner to complete a task by maneuvering through a virtual environment, collecting objects/information until the task is complete. These examples suggest that a transaction may consist of one or many interactions, and I suppose this depends on the instructional concept of the lesson and the parameters which define when a particular task has been completed.
Perhaps the answer is that a transaction is what the instructional designer defines to enable a learner to progress from one state in the instructional process to the next. Of course, if you were to add the ideas of Dave Jonassen where the software becomes a tool for learning, I think the concept of Tx is quite different.