Interlaza

What Interlaza is

From literature to session: the ideas Interlaza uses — and the limits it keeps

DTT, match-to-sample, Sidman, RFT, Varela, Ribes, BKT and stimulus pairing do different jobs inside Interlaza. This is the map.

For: Instructors

By INTERLAZA 5 min read Updated

Research language can make a product sound impressive while explaining almost nothing. “Based on ABA, RFT and artificial intelligence” leaves the reader with three labels and no idea what will happen in the next session.

Interlaza is easier to understand as a set of layers. Some arrange teaching. Some define relations. Some measure change. Others exist only to answer research questions.

DTT sets out how a trial is organized

Discrete Trial Training organizes an opportunity: an instruction or sample, a response, a consequence and the pause before the next opportunity. It tells us how a teaching event is arranged and recorded; it does not decide by itself what relation should be taught.

Figure — One discrete trial, part by part

Trial begins

  1. Instruction

    The material or request the child responds to — a sample to match, a word to find.

    The decision: Is it clean and consistently delivered, or wrapped in commentary that changes?

  2. Prompt (optional)

    The help arranged when the child needs it to respond correctly: a faded distractor, a gesture, a model. Absent once a response is independent.

    The decision: Present on a written schedule and withdrawn by a fixed criterion, or improvised when it looks hard?

  3. Response

    One observable behavior, defined before the session starts.

    The decision: Could a second person score it the same way without asking you?

  4. Consequence

    Feedback, correction and reinforcement are three different things: feedback says right/wrong, correction is the procedure after an error, and reinforcement is what follows a correct response and is meant to make it more likely again.

    The decision: What follows a correct-but-prompted response — the same reinforcer, or less?

  5. Inter-trial interval

    A short pause that closes this trial before the next instruction opens the following one.

    The decision: Is there one at all? Without it the boundary the format depends on disappears.

Trial ends · one scored response

The five parts of a discrete trial, in order, each with the decision it silently answers. Two programs can share the format and differ at every one of these points, which is why procedural fidelity is checked separately from the trial structure itself. The boundary at each end is what turns an interaction into one scored learning opportunity.

Interlaza uses that structure when a discrete response is appropriate. It also supports observation and Pavlovian pairing where a scored trial would be the wrong unit.

Match-to-sample is a task, not a theory of the child

In MTS, a learner sees or hears a sample and selects from comparisons. The format can teach identity, non-identical examples, spoken-word-to-picture relations and more. Prompting and errorless procedures can support early practice, while later probes ask what the learner can do without that support: a few trials with no help and no feedback, which Interlaza’s own mastery rules deliberately exclude from progress — that exclusion is this product’s design choice, not a fact true of probes everywhere.

Here is where each of the ideas below actually sits in the product, sorted by the job it does rather than by how well-known its author is:

Figure — Four different jobs, not one ladder

Organize

Teach & test relations

Measure

Investigate

Four jobs an idea from the literature can do inside Interlaza, sorted by function rather than by fame. Varela and Ribes are research frameworks, not curricula — and right now one of the two is switched off entirely rather than merely unlinked.

Sidman asks whether new relations emerge

Stimulus equivalence moves beyond trained pairs. After selected relations are taught, probes test symmetry or transitivity without feedback. A correct untrained relation is different evidence from another correct repetition of the trained one.

Interlaza separates training from probes so help given during teaching does not leak into a claim about emergence.

RFT adds context to the relation

Relational Frame Theory asks how contextual cues change the relation that applies: same as, opposite, more than, part of, here/there and other frames. This is not “harder matching.” The context tells the learner which kind of relation matters.

Most of these frames can be authored directly from a pathway stage today — coordination, distinction, opposition, hierarchy, comparison, temporal, spatial and causal each have their own setting. Deictic relations (here/there, I/you) are the one exception: they run only from existing templates, since the free-form scenario builder that once authored them from scratch was removed. The Instructor remains responsible for selecting meaningful examples and interpreting performance either way.

Varela and Ribes are research frameworks, and not in the same state right now

Varela’s transfer matrix varies factors such as modality, instance and relation to study changes in performance. Its conditions are not a staircase from easy to difficult. Ribes’ taxonomy distinguishes forms of contact and requires measures that fit each one; it is not a curriculum for teaching a child through five stages.

Neither is an automatic clinical prescription, but they are not equally reachable today. Varela is switched off product-wide during a methodological review — no route renders it, and it cannot currently be run. Ribes has a live, tier-gated route for Researcher accounts; it simply is not linked from the main Pathways navigation, so reaching it means going to it directly rather than following a menu.

Measurement and decision support form another layer

Bayesian Knowledge Tracing maintains a probability estimate from eligible trial evidence. Trials-to-criterion asks how much teaching a target required. Fidelity asks whether the procedure ran as designed. Decision latency asks how long a meaningful signal sat before action.

These measures do not replace judgment. They make the cost and basis of a decision visible.

A different procedure for learners who are asked for very little

Forward stimulus–stimulus pairing, based on Freitas, Henry, Kelley & Tonneau (2020), does not ask the child to respond. An adult says a target sound built by recombining sounds the child already produces in free play, then delivers a preferred item. The order is fixed; there is no correct/incorrect score.

That procedure belongs in the same platform precisely because Interlaza does not force every interaction into DTT.

What this means in practice

Interlaza gives families a calm guided route, Instructors a defensible workflow and researchers controlled ways to ask harder questions. What they share is a record you can follow back: which idea shaped a procedure, which evidence fed a result, and the point past which the software says nothing.

That is more useful than “science-based” as a slogan. It is science made inspectable.