Chapter 14. Autism Enrichment Section: Is It Really Imitation?

1. We’ve said imitation is a crucial skill for kids to learn—a crucial prerequisite to learning most or at least may of the complex skills you and I have learned, everything from tying our shoes to talking, so crucial that almost every autism program starts off teaching imitation.  
   
And here’s the mistake many, if not most, of the autism programs make, including ours until we realized how dumb we were being; unfortunately, it took us many semi-futile years to realize that.
We have a toy truck and we give the kid one. We roll the truck and shape the kid’s rolling the truck.
We have a toy baby doll and we give the kid one. We hug the dolly and shape the kid’s hugging the dolly.
And we think we’re teaching imitation, but most often we’re not. Instead, we’re just teaching truck rollin’ and dolly huggin’. Cute, but not our goal of teaching imitation.
We’ve got the kid huggin’ and rollin’; we give the kid his truck; and before we can grab our truck and start modeling truck rolling, the kid’s already rolling his. Cool? Not really. We’re not really trying to teach truck rolling; we’re trying to teach imitation. Same thing happens with hugging the dolly; he beats us to the hug.

Well, when he was learning to hug and roll, wasn’t he imitating?
Not necessarily; he might never have looked at what we were doing. Maybe we just shaped up the huggin’ and rollin’, and the kid was ignoring what we were modeling with the doll and truck.
But he didn’t hug the truck or roll the doll, at least not often. Doesn’t that mean something?  
Yes, it means the truck was the S^D for rolling and the dolly was the S^D for hugging. But we need more than a simple discrimination between truck and dolly. We need to establish a conditional discrimination: The truck is an S^D for rolling, conditional on you, the model, rolling your truck. If you’re sitting there just holding your truck and he’s rolling his truck, that’s not imitation.

Sounds Bad in Theory, But Is It Really Bad in Fact?

1. So as part of her MA project, Breanne Hartley tested 10 kids who’d previously mastered our standard imitation-training procedure with 9 different toys, each using a different behavior, like rolling the car and hugging the dolly.  
2. Breanne found that most of the kids could still do the manipulations with toys they’d previously learned, though they sometimes manipulated the toys, before she’d modeled the manipulation. In other words, they weren’t really imitating, at least not any longer.
3. Then Breanne did a novel manipulation with each toy, like talking on the car, as if it were a phone and putting the car on top of her head. One kid did all the novel manipulations, some of the kids did a few, but two did none, though they all did the original manipulations well. So the traditional imitation training wasn’t a complete disaster, but it wasn’t nearly good enough.

The Solution

1. And for her dissertation, Breanne used a greatly improved training procedure, designed to teach generalized imitation. And she used this improved training procedure with all 6 of her kids: She made sure the kid was looking at her while she was modeling the manipulation. And she didn’t give the kid the toy to imitate her manipulation, until after she’d modeled the appropriate response. All this at least increased the chances the kid was paying attention to her modeling; much better than the procedures we’d all been using before.
2. The three in Breanne’s control group were taught six different manipulations, each with a different toy. All three of these guys learned all six manipulations, but we can  be sure whether they were learning to imitate or just to manipulate; in other words, we can not be sure Breanne’s modeling of the manipulation was actually exerting stimulus control of the kids manipulating.
3. And the three in Breanne’s experimental group were also taught six different manipulations, but this time two different manipulations with each of three toys. And again, all three of these guys learned all six manipulations, and now we can be sure they were learning to imitate and not just manipulate. Great!
4. But we want more than that they our kids can imitate six different well-trained, manipulations. We want our kids to be able to imitate novel manipulations; that’s our goal.
5. So then Breanne tested all six kids by modeling novel manipulations. The three experimental kids, who’d had to attend to Breanne’s modeling, showed considerably more generalized imitation of novel manipulations than did the control kids, who did not really have to pay attention to her modeling in order to get their reinforcers. In other words, if you want to more reliably achieve your real goal of helping your kids acquire the skill of generalized manipulative imitation, you need to make sure you’re teaching imitation and not just manipulation. However, even if you make the common mistake of failing to use at least two different manipulations of each toy, some of your kids may still learn a little generalized imitation.
6. But not even the experimental kids had acquired as much generalized imitation as they’d need, by training only six manipulations. So after this dissertation research, Breanne was able to make sure that three of the kids received more manipulative imitation training by the regular tutors in the classroom. And the result was that all three kids mastered the skill of generalized manipulative imitation. Not only that, it took training on only two to six more manipulations to achieve this excellent generalized manipulative imitation.

More Good News

1. And that’s not all the good news. Without explicit training, these three kids had also mastered physical imitation, you know, like clapping your hands, rubbing your belly, etc. Now that’s not too surprising, because manipulative imitation is essentially physical imitation with something in your hand, like putting a phone to your ear involves the physical imitation of raising your hand to your ear, and the only real manipulative part of it is that you picked up a phone before you made that hand-to-ear move.
2. And still more good news. Without explicit training, these three kids had also mastered vocal imitation with about half the sounds Breanne tested them on; you know, like mmm, ooo, eee. Not the Gettysburg Address, but that’s where we start; and it’s a real big deal. And that generalized vocal imitation really did surprise me—still not sure why it works that way, but happy it does. (If you can figure out why we’re so fortunate as to get generalized vocal imitation for free, please let us know.)

The Gina Green Dumbass Award

1. Remember from Chapter 13, the Gina Green Principle: You need at least two S^Ds in order to teach a simultaneous discrimination. And the error is: You train touch red with a red circle. Then you train touch green with a green circle. And then you put out both the red and green circle and are shocked to find the kid’s not paying any attention to what you’re saying; he’s just touching circles, whatever color. More technically, red and green are exerting no stimulus control over which circle he’s touching. Remember, for a stimulus to be an S^D you must also have an S^Delta . So if you want the word red to be an S^D for his touching red, then you need an S^Delta for touching red, like the word green. Otherwise, the kid may not pay any attention to the specific words you’re saying, more technically, the specific words won’t start exerting stimulus control. He’ll just touch red.
2. Same with imitation training. If you want the kid to imitate what you’re modeling, if you want the specific manipulative behavior you’re modeling to exert stimulus control over the kids manipulative behavior, if you want your rolling the truck to be an S^D for his rolling the truck, then you need an S^Delta for truck rolling, like your putting the truck on your head.
3. So why doesn’t the traditional imitation training doesn’t work well? Because there are no S^Delta manipulation for the various manipulations we model.
4. Why does Breanne’s imitation training worked so much better? Because there is an S^Deltas for each of various manipulations we model. She modeled two different manipulations with each toy; so every modeled manipulation with a given toy also had an S^Delta.