Effects of Video Modeling on the Acquisition, Maintenance, and Generalization of Playing with Imaginary Objects in Children with Autism Spectrum Disorder.

Three preschoolers with autism watched short clips of an adult pretending. The adult used imaginary objects: drinking from an empty cup or talking on a toy phone that was not there.

After each clip, the kids got the same toys and were told, "Your turn." The team counted how many pretend actions the children copied right away, one week later, and with new toys.

All three children quickly started using imaginary objects. The new play lasted after the videos stopped and showed up during free play with different toys.

Parents reported the kids also began to pretend at home without prompts.

Brodhead et al. (2019) and Lee et al. (2020) got the same result, but they used questions instead of videos. They asked, "What can we pretend this is?" and taught the kids to give many answers. Both methods worked, so you can pick the one that feels easier in your session.

Chee et al. (2017) used video self-modeling to teach functional play, like pushing a toy car. The new study moves the idea forward by showing that plain video modeling also works for imaginary play, not just real-world actions.

Agana et al. (2025) trained pretend actions with real objects first, then saw the actions move to toy versions. Ptomey et al. (2021) show the opposite path is possible: start with video, then watch the play jump to new toys and real life.

If a child with autism rarely pretends, a two-minute video clip can be your fastest prompt. Film yourself or a peer acting out the target scene, show it on a tablet, and hand over similar toys. You can loop the clip while you run other trials, making efficient use of time. Because the skill generalized in this study, you do not need to train every single toy set; three to four examples seem enough for the idea to spread. Try measuring play during free choice to see if the pretending shows up where you did not place any prompts.