Relational frame training

The short answer

Relational frame training is practice at working out how things relate when nobody has told you directly.

You get a few short statements, such as "Zek is more than Vox" and "Vox is more than Qir". Then you answer a question the statements never answered, such as "Is Zek more than Qir?"

The made-up words stop you from leaning on what you already know. The only thing being practised is the reasoning.

It comes from relational frame theory, a psychology account of how people learn language and reason.

Try three questions to see how it feels.

What relational frame theory says

Relational frame theory (usually shortened to RFT) was set out by Steven Hayes, Dermot Barnes-Holmes and Bryan Roche in their 2001 book Relational Frame Theory: A Post-Skinnerian Account of Human Language and Cognition.

Its central idea is that a large part of human thinking comes down to one learned skill: relating things to each other in ways that don't depend on what the things look like.

A young child learns that a bigger coin is "more" than a smaller one because they can see the size difference. Later the child learns that a small coin can be worth more than a big one.

At that point, "more than" no longer depends on anything you can see. It has become a relation you can apply to anything, including things you have never met. RFT calls this arbitrarily applicable relational responding.

The word "frame" refers to one of these learned patterns, such as "same as", "opposite of" or "more than". You can place a frame around any pair of things.

RFT says every frame has three properties. They are what the training questions test.

1. Mutual entailment

If you learn that A relates to B in some way, you also know how B relates to A, without being taught it.

Told "Zek is more than Vox", you know that Vox is less than Zek.

2. Combinatorial entailment

If you learn how A relates to B and how B relates to C, you can work out how A relates to C.

Told "Zek is the opposite of Vox" and "Vox is the opposite of Qir", you can work out that Zek is the same as Qir. Nobody told you that. You worked it out from the two statements.

3. Transformation of function

When relations link things together, what you feel or do about one thing can carry over to the others.

If someone tells you a Zek coin is worth more than a pound coin, you'll pick the Zek over the pound, even though you've never seen a Zek.

Nothing about the coin itself changed. The relation changed how you treat it.

The relations you derive this way are called derived relations. Nobody trains them directly.

The claim that makes RFT interesting for brain training is that this deriving skill sits underneath a lot of what intelligence tests measure: analogies, verbal reasoning, "A is to B as C is to ?" items, and working through a chain of conditions in your head.

The main relational frames

Researchers group frames into families. The table uses made-up words, as the training does.

Distinction is the vaguest frame. "Different from" tells you less than "opposite of" does.

If Zek is different from Vox and Vox is different from Qir, you can't tell how Zek and Qir relate. Distinction questions have to be built carefully to have an answer at all.

Deictic frames are the ones behind perspective-taking. RFT researchers have built training aimed at understanding other people's point of view on them.

What a training session looks like

Relational frame training programmes differ in the details, but they share a basic loop. Here it is as it runs in Terminal, our relational frame training game.

1. You see two or more short statements (the premises) about made-up words.
2. You get a question about a relation the premises never stated.
3. You answer Yes or No and get immediate feedback.
4. The next question comes straight away. As you get more right, the questions get harder.

Here is a worked example of one harder question.

Zek is the opposite of Vox.

Vox is the opposite of Qir.

Qir is the same as Taz.

Is Taz the same as Zek?

The opposite of an opposite is the same, so Zek is the same as Qir.

Qir is the same as Taz, so Zek is the same as Taz. The answer is Yes.

How questions get harder

You get more premises to hold at once, and frames get mixed in the same question.

The premises can arrive out of order, so you have to rebuild the chain yourself.

Some questions add an extra premise that has nothing to do with the answer. You have to spot it and ignore it.

Later questions move from single frames to networks, where several frames interact.

Why the words are made up

With real words, "A cat is bigger than a mouse" can be answered from what you already know about cats.

With "A Zek is bigger than a Vox", memory is no help. You have to do the reasoning.

What the research says

Pages about relational frame training tend to overstate the evidence one way or the other. Some promise big IQ gains. Others say brain training never works.

The research sits between those two positions, and it has shifted over time.

The early results

Bryan Roche, Sarah Cassidy and colleagues at Maynooth University in Ireland developed a relational training programme called SMART (Strengthening Mental Abilities with Relational Training).

It trains the frames of same, opposite, more and less using made-up words.

A pilot study by Cassidy, Roche and Hayes, published in 2011 in The Psychological Record, reported IQ increases in children after relational training.

The study that drew most attention was Cassidy and colleagues (2016) in Learning and Individual Differences. In the first experiment, 15 children aged 11 and 12 did several months of SMART training.

Every child showed a full-scale IQ gain on the WISC-IV of roughly one standard deviation. The smallest gain in the sample was 14 points.

14 pointsThe smallest full-scale IQ gain among the 15 children in Cassidy et al. (2016).

A second experiment with 30 adolescents aged 15 to 17 found gains on the verbal and numerical reasoning parts of the Differential Aptitude Test.

Colbert and colleagues (2018) repeated the training over 12 weeks with 26 secondary school students and reported IQ gains on the WASI, a shorter IQ test.

A 2020 active-controlled trial in the Journal of Behavioral Education reported that relational skills training raised adolescents' scores on a standardised matrices test.

The problems with those results

The early studies were small, and some compared trained children with children who did no training at all.

With that kind of comparison it's hard to rule out other explanations, such as practice at test-like tasks or the extra attention that comes with being in a study.

Bryan Roche, a co-author of the 2011, 2016, 2018 and 2023 studies, declares a commercial interest in RaiseYourIQ, the company that sells SMART training. That doesn't make the results wrong, but it's one more reason independent replications matter.

In 2022, May and colleagues pulled the SMART trials together in a systematic review and meta-analysis in the British Journal of Educational Technology.

They found five trials that met their criteria, and rated all five as having a high risk of bias.

Across those trials they calculated a moderate effect on nonverbal IQ. They found too little evidence to say anything about SMART's effect on any other cognitive or school skill.

In 2023, Roche and colleagues ran another SMART study, and this time measured attention skills and starting IQ. Some of the co-authors had worked on the original studies.

Participants' IQ scores went up after training. But once the analysis accounted for attention skills and baseline IQ, the training no longer explained the gains.

The newest trial

In January 2026, Meijer, Delabie, De Houwer and Cummins published a randomised controlled trial in Collabra: Psychology.

They gave 105 adults aged 62 to 93, all with self-reported memory or thinking complaints, either SMART or dual n-back training.

People in the SMART group got better at relational reasoning. There was no sign that this carried over to nonverbal IQ, neurological status, depression or quality of life.

Both groups improved about the same amount on executive function.

Where that leaves things

We think relational frame training is worth doing for the skill itself: holding several relations in mind and working out what follows from them.

That skill shows up in reading arguments, following instructions, planning and solving puzzles.

If someone promises you a fixed number of IQ points from it, the research doesn't support them.

How it compares to other brain training

Relational frame training and n-back get compared most, since both have been studied for effects on fluid intelligence.

N-back trains how much you can hold and update from moment to moment. Relational training trains what you do with the information: combining it and drawing a conclusion.

They are different exercises. The 2026 trial above found neither did better than the other on executive function in older adults.

Our dual n-back guide covers the n-back side, and our guide to the wider evidence on brain training covers the rest.

How to practise

The studies ran their training over weeks or months, not single sessions. There's no agreed best dose, so these are our suggestions rather than research findings.

Little and often

We suggest short sessions on most days rather than the odd long one.

Ten to twenty minutes is easy to fit in, and it's easier to keep going for weeks.

Accuracy before speed

If you rush questions you don't understand, you end up guessing, and guessing doesn't train anything.

When you get one wrong, read the premises again and find where your chain broke before you move on.

Training that adapts

If every question is easy you aren't practising anything new. If every question is too hard you're back to guessing.

The useful zone is questions you get right most of the time, but not without thinking.

How Terminal does it

Terminal is Neurobuff's relational frame training game. It follows the same method as the research: made-up words, short premises and a Yes or No answer, with immediate feedback.

You start on the core frames: more and less, same and opposite, same and different, before and after, directions on a grid, and "all" and "some" statements.

Once you're consistently accurate, Terminal brings in advanced frames. These include perspective (I, you, here, there), hierarchies, cause and effect, frames that change meaning with context, and networks where several frames interact in one question.

Difficulty adapts to you. Terminal looks at how accurate you are and how quickly you answer, and sets the next question from that. There's no level to pick.

Terminal is one of five training games in Neurobuff. Start training to play the full game.

Frequently asked questions