Generativity: When New Skills Emerge on Their Own

Generativity is when a learner shows new skills that no one taught them directly. The learner puts known pieces together and makes something new. This shows up most often in language, but it also shows up in play and problem solving.

This matters because we can never teach every response a person will need. There is not enough time in a day. When teaching builds generativity, the learner keeps growing on their own. That is the real goal of good instruction.

What generativity means#

Generativity comes from the study of verbal behavior, which means language treated as learned behavior. A learner is given some direct teaching. Then new responses appear that were never trained. Those new responses are the sign of generativity.

Dr. Jessica Osos frames it in clear terms in her talk on teaching for generalization. She defines generativity as the emergence of new verbal relations or operants that were never directly taught.

An operant here just means one unit of learned behavior. So the learner produces new units of language without a teacher setting up each one. The teaching seeds a system, and the system grows past what was planted.

Generativity is not the same as generalization#

People mix up generativity and generalization all the time. They are close, but they are not equal. Generalization means the same skill shows up in new places or with new items. A child who says "dog" for a poodle and a lab has generalized. The skill is the same. Only the setting or the item changed.

Generativity goes further. It is a brand new response, not the same skill in a new spot. Osos makes the line sharp with one short rule.

All generativity involves novelty From the talk — Dr. Jessica Osos

But she adds that not all novel responding reflects generativity. So novelty alone is not proof. A response can be new but still just be simple variation. Generativity is the kind of new response that shows a whole new relation the learner worked out.

Generalization is needed, but it is not enough#

Osos describes generalization as necessary but not sufficient for generativity. That is worth slowing down on. You almost always need solid generalization first. The learner has to use skills across many items and settings.

But strong generalization by itself does not promise new relations will emerge. You can have a learner who uses a skill everywhere and still never produces something untaught. Generativity is the higher bar. It asks the skill to become a launch pad for new skills.

Novelty is the key sign, but read it with care#

The clearest marker of generativity is a correct response that was never trained. The child names a category no one drilled. The child asks for an item using words in a new order. Those moments tell you the teaching took root.

Still, you have to read novelty with care. Not every new answer is generative. Some new answers are noise, or lucky, or just small tweaks. The skill for a teacher is telling real emergent relations apart from surface variation. That judgment is part of what makes this concept hard and worth studying.

How teaching methods build generativity#

Some teaching methods are built to grow generativity on purpose. Multiple exemplar instruction, which means teaching a skill with many varied examples, is a core one. This is the focus of Osos's course. You teach across many examples so the learner pulls out the shared rule, then applies it to new cases.

Matrix training works in a similar spirit. It uses a teaching grid, where you cross words like actions and objects. You teach only some of the boxes in the grid. If it works, the learner performs the untaught boxes too. The untaught boxes are the generativity. Research has looked at matrix training and verbal generativity in children with autism for this reason.

Derived relational responding is another route. This means links the learner forms without direct teaching. Teach that A goes with B and B goes with C, and the learner may derive that A goes with C. These derived links are a source of new language that no one had to train step by step.

What developmental research adds#

Generativity is not only a teaching target. Researchers also use it to understand how thinking differs across groups. Studies have measured how many new ideas a person can produce on a simple task. This is called ideational fluency, which just means how freely new ideas come.

Some studies find that people with autism generate fewer novel actions on certain tasks. One example is inventing new things to do with a toy car. Other tasks, like play with a doll, show no such gap. Work on children with Down syndrome found fewer new actions on new objects compared to matched peers. These findings suggest generativity is a real and measurable difference, not just a teaching outcome. They also remind us that the same word covers language, play, and idea generation.

What the research says#

Several peer-reviewed studies inform this page. Most did not have full citations on file, so they are named here by topic and finding.

One paper does carry a complete reference. It ties language generativity to response generalization and to derived relational responding (Stewart, McElwee, & Ming, 2017). That lines up with the teaching routes above. The full reference is: Stewart, I., McElwee, J., & Ming, S. (2017). Erratum to: Language Generativity, Response Generalization and Derived Relational Responding. The Analysis of Verbal Behavior, 33(1), 176-176. https://doi.org/10.1007/s40616-016-0060-8

The other studies cover matrix training and verbal generativity in children with autism. They also cover generativity in play, object play in children with Down syndrome, and ideational fluency in children with communication problems. Readers who want to cite those formally can find the full references in the original journals.

Frequently asked questions#

Is generativity the same thing as creativity?

They overlap but they are not the same. Creativity is a broad, everyday word. Generativity is a more exact term for producing untaught responses from taught parts. In research it is measured, like counting new ideas or new relations. So generativity is one careful way to study part of what people call creativity.

Can you teach generativity directly?

You cannot teach the new responses one by one, because that would defeat the point. What you can do is set up conditions that make new responses likely. Teaching with many examples, using matrix grids, and building derived relations all raise the odds. You teach the system, then watch for untaught skills to appear.

How do I know generativity happened and not just a lucky guess?

Look for untaught responses that hold up over time and across items. One right answer could be chance. A pattern of correct untaught responses points to a real emergent relation. Good data collection is how you tell the difference between a fluke and true generativity.

Keep learning#

Dr. Jessica Osos walks through how to build these outcomes step by step in Multiple Exemplar Instruction (MEI): Training for Generalization. It is a solid next step if you want to turn this concept into daily practice.