Teaching Kids to Play
They define play as being internally motivated, voluntary, spontaneous and flexible, involving attention to the process of playing rather than to any end product, actively engaging and controlled by the player, involving at least some freedom from the constraints of reality, and being safe and enjoyable. They ask the question: can a spontaneous, internally motivated, flexible kind of behavior really be taught by systematic training?
First, they note the differences between their definition and the operative definitions used in behaviorist studies they reviewed. 'Appropriate play' was most often defined as playing with an object in the way that someone else intended children to play with it (eg toy cars were intended to be treated as cars and made to do car-like things). Another definition was playing within the way most neurotypical children would play with it. They noted that both of these definitions seem to preclude a flexible generation of new ways of playing, which they felt was inherent to what play is.
They also refer to the suggested inverse relationship between stereotyped behaviors and play, which at least one study they reviewed explained by stating that 'stereotypical behavious have a play function and can be replaced by play behaviour'. They pointed out that although stereotyped behaviors aren't flexible, they meet all the other criteria for play - they are internally motivated, spontaneous, done for the experience rather than as a means to an end, and appear to be enjoyable. In fact, stereotyped behaviors seem closer to play than what the reviewed researchers are attempting to train into the children, even though they look more abnormal. They criticize these teaching methods as being motivated by the cosmetic potential of the taught play behaviors, rather than the true meaning and purpose of play.
A few studies they reviewed did attempt to teach generative and creative play. One, for example, trained children to self-monitor their own activities (play with a robot, conversation and drawing for the three children) and attempt to produce varied responses. All three produced more variable responses. Since two were being trained in 'play' situations (playing with a robot and drawing), those two could potentially be considered as having developed more varied play. However, it appears that the children continued to self-prompt rather than internalizing this behavior, since the boy learning to have more varied conversations, when unable to think of anything new to say, would just say 'a new one'.
Overall, a few studies they found did make claims of developing dispositional play rather than just topographically normal-looking trained play. All of these used at least some non-ABA features, such as giving the child more control over the process and informing their goals by developmental research into play. One never used any explicit reinforcement, depending on the intrinsic reinforcement of play to motivate the children.
They also point out that proponents of ABA often defend against criticism that it teaches rigid, rote behavior by claiming that this behavior undergoes a shift over time - it starts out rote, but becomes more fluent and spontaneous as it gets consolidated. None of the studies reviewed directly examined this claim, not even one study that claimed the two children had become 'indistinguishable from their peers' as a result of intensive ABA. If those children had actually become indistinguishable from neurotypical peers, then their behavior must have shifted from rote trained behavior to more fluent spontaneous behavior, yet no description of this process was provided.
Personally, I think a lesson can be taken from the similarity of stimming (stereotyped behaviors) to play. Stimming differs from typical play primarily in that it's typically not as varied and that it's usually not interactive. Although not much research has addressed this question, in my experience, both of those can be changed fairly easily in an autistic person without any attempt to teach normal-seeming play.
Firstly, the way to make stimming interactive is quite simple - join the child in their stimming. In fact, I've often seen attempts by autistic kids to prompt an adult to join them in stimming. One such attempt (completely overlooked by the adult recipient) can be seen in the Autism Every Day video - a boy is sitting with his mother at a table while she is being interviewed, he is flicking his fingers in front of his eyes, and then he spontaneously moves his hands to flick his fingers in front of his mother's eyes. To me, this is a clear attempt by that child to show his mother the interesting visual stimulus of flicking fingers in front of her eyes, and invite her to interact with him. If it had been me in that situation, I would have immediately flicked my fingers in front of his eyes, and made a social game out of it. Instead, she simply pushes his hands aside, failing to notice that a child with a diagnosis characterized by reduced social initiations had just made a social initiation towards her.
Even if the child doesn't overtly invite you into their stimming, most autistic kids I've met won't actively object to you joining them - as long as you don't mess things up. One boy, for example, got a big smile on his face when, while he was staring at a fan and acting excited, I went behind the fan and waved my hand so he could see it through the fan.
When it comes to encouraging more variability, making stimming interactive can naturally encourage this, because you'll show them other ways to do it. But I also question to what extent stimming really isn't variable. In Amanda Baggs' video In My Language, she first flaps and rocks while staring at a bright window in a darkened room, then starts brushing something rough (I can't make out what it is) against a wall, then strokes a computer keyboard, then flicks a necklace around against the light, then stares down a dangling slinky, then - well, you get the point. Each behavior is sustained for a significant amount of time, but overall, there is quite a wide variety of things Amanda Baggs does in that video. There are many autistic individuals who, like Amanda Baggs, show a wide variety of stimming behaviors. This could reflect more severe autism, or it could reflect more creativity in their stimming. Research hasn't studied this question much.