Assessment through the HDML lens—Part 1: Relating

The situation that generally leads to people reaching out to me for consultation is that they’ve been working with a young learner who seems to have some (or even a lot of) language skills, but something seems “off”—responding is rigid or progress is slow, or they’ve been working on academic skills like reading or math, and just continually getting stuck. Or maybe they’ve been working through RFT-based protocols but the child isn’t progressing as expected. My first questions always have to do with assessment—what skills can this child demonstrate, and under what circumstances?

In my last few posts I worked through the HDML — first the levels of relational responding, then complexity and flexibility, then derivation and coherence, and most recently the ROE-M as a way of holding all of it together as a dynamic operant unit. Today and over the next couple posts I’ll take a look at how that maps onto a framework for assessing and teaching relational framing skills.

In any therapeutic setting, we need to understand what influences client behavior; this is no different when considering the behavior of relational framing—once we embrace the notion that relating is an operant, it is easy to see that that behavior will be influenced by a whole range of antecedent and consequent variables. And, understanding these potential influences becomes particularly critical in early intervention, when we are working with early learners who may not have any (or may have a very limited) history of derivation within a given pattern of relational responding. If we are to accurately assess their skills in order to sequence programming appropriately, we need to determine if the relational framing repertoire in question simply needs strengthening and generalizing, or if it is a repertoire that needs to be newly established, because each of those two situations involves a different programming path.

In the first case, when a repertoire is demonstrated in some but not all contexts, your task involves generalizing to new contexts, adding discrimination requirements, and building fluency (just like with any other skill you are trying to establish). But if the repertoire has not yet been established at all—there are no contexts in which your learner demonstrates the relational response—then that is your primary task: using multiple exemplar training in the pattern to establish relational responding at least in one context.

Decades of research have shown that demonstration of derived relational responding (DRR), particularly equivalence, may be more likely given a number of factors. Most of this research has been with frames of coordination, but we have also found it to be helpful to our analysis and programming with non-sameness relations. What follows in this post series are the variables I always try to keep in mind, and which we have considered in developing our protocols for assessment and training. Some of these factors are well established in the equivalence literature; some of them are our reading of that literature through an RFT lens.

Factors relevant to the “relating” part of the ROE-M are ones that make sense to most practitioners even if they are using different language to describe them. For one, consider the developmental level you’re working at. You wouldn’t expect a child still learning to imitate gross motor movements to be able to follow a dance video. In the same way, we need to assess and establish mutually entailed orienting and evoking (joint attention) and non-arbitrary relational responding (NARR) before mutual entailment, relational framing, relational networking, relating relations, and relating relational networks.

Another aspect to consider is relational complexity — this includes what type of relation is involved (with coordination being simplest, and complexity increasing from there) and how many relations need to be trained. Derivation always involves responding to a novel relation on the basis of related responses having been trained: to derive A=C, a learner has to have learned A=B and C=B. So the number of trained relations is itself a marker of complexity, both in terms of how many separate relations are involved (A1-B1-C1 and A2-B2-C2 are two separate three-member relations) and how many members each contains (A1-B1-C1 has three members; A2-B2-C2-D2 has four). Fewer trained relations make training easier and successful testing more likely—very young children may have difficulty learning or remembering more than two conditional discrimination relations before being tested for equivalence. At the same time, fewer relations means more likelihood of chance responding, so pass criteria need to be fairly strict.

In practice, when I’m working with early learners I use small sets (usually two three-member classes), and I don’t assume that NARR has been established or well-generalized (unless of course the learner is spontaneously demonstrating these skills, like if they are going around talking about how their shoes are a different color than mine or they ask for a bigger pot in their play kitchen to make soup). Testing for NARR goes very quickly for a learner that has the repertoire, and saves a lot of frustration in trying to test for DRR with a learner that does not. I’m also not going to test for combinatorial entailment without first testing mutual entailment, and so on up the levels. Similarly, I will work in a developmental sequence in terms of the complexity of the relation itself, moving from coordination first to distinction, comparison, opposite, spatial, and temporal relations.

Sometimes just determining more clearly whether programming is at an appropriate level is enough—it becomes clear that non-arbitrary training is needed, for example. When I was first working on teaching arbitrary same and different responding, for example, every teacher told me their students “understood” same and different. But universally, more work was needed to discriminate those contextual cues in new contexts and with new stimuli at the non-arbitrary level before we started playing games about animals that liked the same and different food as one another! It was only once we did the NARR work that we were able to get somewhere with multiple exemplar training for deriving new arbitrary relations. Along the same lines, you need to know what relational patterns are involved and if your learner has the necessary relational responding skills to complete the math word problem, or answer the reading comprehension questions that you’re stuck on. If a learner can’t readily respond to questions about difference, then comparison or temporal relations will be very difficult.

However, as the ROE-M analytic unit highlights, relational responding does not only consist of "relating". We’ve talked today about which relating skills to look for and how the complexity of those relations affects assessment decisions; in my next post we’ll look more at how different contextual variables influence whether or not those skills will be demonstrated.