Using RFT to Promote Generative Language: Derivation and Coherence
If you've ever watched someone—a learner, a parent, a friend—yourself— get stuck in a response pattern that clearly isn't working anymore, but they keep doing it anyway, you've been face to face with the clinical issues at the heart of today's post. In my last two posts, I've been working through the HDML framework—first the levels of relational responding, then the dimensions of complexity and flexibility. Today I'll cover the remaining two dimensions: derivation and coherence. Both deal with what happens as relational responses become more established over time, and both connect directly to the challenges we all navigate around rigidity, rule governance, and psychological flexibility.
Derivation
Derivation is a term you will be familiar with if you have learned anything about RFT, especially if you are using it in your language intervention programming: establishing repertoires that involve deriving new relational responses on the basis of previously learned ones (i.e., mutual and combinatorial entailment) is the crux of promoting more generative language. As such, it is also an essential foundation for psychological flexibility: observing the context and deriving rules related to valuing to guide your behavior when things are hard requires that first you are able to derive relations at any level.
Derivation as a dimension of relational responding refers to the extent to which a response is derived, vs directly taught or reinforced over time. This important to consider in intervention because it sits at the center of a key programming tension. We aim to establish derived relational responding as a repertoire across many relational framing patterns—and we are always looking for expanding novel responding within those patterns. Once we’ve established responding to “different” relations, we would keep generalizing and expanding that repertoire as we also move into new patterns, often using games and stories that require learners to respond to new relations of how things and people might be the same or different. In doing so, we strengthen the generalized repertoire of relating within that pattern, with novel stimulus relations.
But if completely novel or newly derived responses are meaningful and functional (as they should be!), they will also contact contingencies that strengthen those specific responses. It is actually important to learn the differences and similarities with meaningful, real-life content—academic learning is full of comparing and contrasting various concepts, as is discussion of all kinds of topics—talking to a friend about how your favorite book compared to the movie about it is likely to an enjoyable and reinforcing conversation. It’s socially useful to remember that your friend Susie likes different snacks than Becky does. It’s also important to “know” all the relevant attributes of common items that might be derived first through equivalence-based teaching—that a bird has wings, and flies, and you can see ducks in the park, and they say quack, and so on. As these responses contact real-life contingencies, they become more "practiced" and will become fluent and maintained over time.
At the same time, derivation is directly implicated in psychological flexibility and inflexibility. The more practiced a response is, the more difficult it is to change, and "rigid" response patterns often involve a long history of reinforcement. Providing opportunities to derive new responding can help introduce variability into a well-practiced repertoire. Both structured and natural environment/event-based teaching can be engineered to include novel stimuli — going to new places, trying new activities, encountering new items within familiar routines. And so, we need to provide many opportunities for derivation of new responses, and provide opportunities to "use" those responses in contexts where they are meaningful. This balance is an important issue in programming—and one we'll return to in more detail when we discuss case conceptualization.
Coherence
Relational coherence refers to the extent to which a response can be predicted given its previous history of reinforcement—that is, is this response consistent with one's learning history? Having learned that a rambutan is a type of fruit, a response relating rambutan as similar to other fruits, or as edible, would be highly coherent; responses relating rambutan as similar to a coat would be much less so.
From an RFT perspective, coherence is considered a powerful reinforcer and motivational variable, and it is a key proposition of RFT that coherence is sufficient to maintain derived relations. In early intervention, it may therefore be important to work on establishing coherence as a reinforcer. This can be pursued through different avenues and at different developmental levels—at first simply by pairing reinforcement with accurate completion of relational tasks that have an identifiable response product (such as completing a matching board), and later through teaching children to check the accuracy of their own work, and through teaching self-monitoring and self-reinforcement.
Coherence, like derivation, is also implicated in rule governance and psychological flexibility and inflexibility. Because coherence is so reinforcing, it can be very difficult to respond differently from what has been done in a given context in the past, even if that response is no longer "working" in the sense of having a reinforcing outcome in the current environment. Practicing responding in ways that are less coherent—such as by practicing "being wrong," or doing things the opposite of how they would normally be done—is one way to help increase flexibility. Even here, though, introducing the instruction to "be wrong" or "do the opposite" creates some coherence—paradoxical, we know! Similarly, responding to the relation between a rambutan and a coat would likely involve relational networking until you are able to relate both these items in terms of something else (like being fuzzy), thus creating a new context that adds coherence.
Coherence is also important in a teaching context because many educational tasks require evaluating whether a particular relation "fits" with what has been learned. Yes/no questions are the most common format for this—for example, I could tell you that unobtainium is rarer than kryptonite, and kryptonite is rarer than osmium, and then ask you if osmium is more plentiful than unobtainium. In early intervention, you're likely familiar with teaching programs that ask learners to answer yes/no about what an item is called—for example, while holding up a banana, asking "is this an apple?" Academic skills often require the evaluation of increasingly complex relational networks in increasingly varied contexts. RFT refers to this type of responding as a "relational coherence indicator"—a highly generalized repertoire that evaluates whether or not a relational network is consistent with previous learning history. If you think about how much of academic and daily-life learning involves evaluating whether something "makes sense" or "sounds right," you can see how foundational this repertoire is—and how important it is to establish early. We'll look more closely at how we can use this in assessment and programming in the next post.
Extending the dimensions to non-arbitrary relational responding
For early intervention practitioners working with early learners, considering how these dimensions could map on to non-arbitrary responding is critical, since much of our work is with NARR. While the HDML only addresses AARR, we would argue that non-arbitrary relational responding can be viewed in terms of these same dimensions. NARR can involve more or less complexity as determined by number of stimuli being related, number of discriminative stimuli and contextual cues, and types of relations. Similarly, relational flexibility and coherence can be identified at both non-arbitrary and arbitrary levels.
While only AARR can be viewed as having a dimension of derivation per se, we would view NARR as being characterized by a dimension that could perhaps be referred to as the novelty of generalization—is this a response to a completely novel set of stimuli, or have these stimuli been encountered before? Just as with AARR, it is important to provide opportunities to generalize and strengthen NARR with brand new stimuli, while also providing opportunities to contact reinforcement in contexts where functional use is relevant. For example, we have often suggested embedding responding to cues for same and different within the context of various natural educational activities such as arts and crafts projects (e.g., "can you get me a different crayon?" "Oh, I want to make mine the same color as yours!"), as well as prompting tacting of same/different relations in as many contexts as possible—from nature walks to doing the laundry. And at the same time, a well-practiced response, such as identity matching, may be very difficult to change—we have seen this when trying to teach responding to cues to select a "different" item in a match-to-sample format that has been well-practiced with selecting "same" items.
Within the context of early intervention in particular, we find consideration of these dimensions at all levels, including non-arbitrary relational responding, to be very helpful in developing curricular sequences and problem solving when progress is not being made. In my next post, I'll look at how we pull levels and dimensions together into a practical framework for assessment and programming decisions.
I hope you've found this helpful—as always, I'd love to hear your thoughts and questions.
Want to receive these articles via email? Sign up here.
Interested in training on applying RFT in early intervention? Check out the courses I run through Constellations.