Session 1, breakfast in location 1:
I don’t think you’ll see the fireworks-induced slight concern in Game’s body language – but I certainly see it before and after this session. Thursdays are not our best days.
Session 2, dinner in location 2:
Today, Game gets an entire hot dog from my hand after the click. Yep, I know, she’s a vacuum! It takes her only a second to inhale that hot dog! This was certainly higher value than the usual hot dog piece: you can see that in the way she hesitates and looks at me after eating it. Maybe there’ll be another one?
Is it a behavior chain?
I generall stick with an approach for about a week. By then, if I’ve worked on it daily, I want to see measurable progress. In my first location (inside), I’ve succeeded in turning the visual/olfactory cue of the pile of kibble into a cue to return to me. In my second location (outside), I’ve seen what I’ve interpreted as glimpses of progress – but nothing tangible. And it’s been a week outside. It’s time to start looking at other possibilities. And there is a BIG one: behavior chains.
Two things can happen in set-ups like this: you can either get a cue transfer (which I’ve been aiming for: new cue (pile of kibble) followed by old cue (recall) should eventually turn the new cue into a cue for what used to be cued by the old cue. This is what I planned on, and what happened inside the house.
However – there’s another possibility: I could have been building a behavior chain of run around the corner and up to the kibble in order to get recalled, in order to get released to the kibble (or get to eat hot dogs and then be released to the kibble). It is entirely possible that Game has developed the superstition that running up to the pile of kibble is what she needs to do in order to set the entire progress in motion – as if there were a big kibble-pile-shaped button she needed to push: my recall cue marks the button push, which unlocks the reinforcement galore: hot dogs and kibble, praise and freedom to go look for the intermittent cat after.
We consciously build behavior chains all the time, for example in trial prep when we backchain towards a final big reinforcer that will be given at the end of a (far more complex) routine. If we set things up just right, the dog will perform the entire routine in order to get that final reinforcer – even a dog who is not intrinsically motivated to do the behaviors that come before the reinforcer. If I want to build a behavior chain, I will first work on all the parts independently to get fluency, and then chain them together like beads on a string, connecting them to a final large reward.
At this point, I truly wonder whether this is what’s going on with Game and the kibble pile outside. It would make sense, and here’s why. I have successfully turned environmental stimuli into cues in the past – among them critters, sheep, and all kinds of other animals, both in my own dogs and in student dogs. There’s one crucial difference though: in these scenarios, the reinforcer (chasing the critter; herding) is generally unavailable until after my release cue. There is little reinforcement history for chasing off cue. It’s under stimulus control because that’s how I’ve set up from the start. With students who already have a strong reinforcement history around freely accessing a particular stimulus, I will often implement a strict management plan as we work on turning that stimulus into a cue: no more free access to, say, chasing squirrels. I’ll want to convince the dog that chasing squirrels will become available – but only after either a cued behavior or a voluntary check-in, depending on the specific training goal. The management piece (no free access to that same reinforcer) is something I stress while we work on the training plan. The longer a history the dog has of freely accessing their reinforcer, the longer we’ll have to manage and train because we have to overcome all that history, and convince our learner that the reinforcer isn’t available “for free” anymore.
On the other hand, when building a behavior chain, I will work on all the links in the chain independently, getting them fluent and giving each its reinforcement history before chaining them together (for example, if you are working up to an obedience routine, you will train each behavior and give it a strong reinforcement history before chaining them together into an entire run).
Game is practicing one link in what could be a chain every day: freely accessing food out and about.
I managed to get the cue transfer just fine in the house with Game. She has a history of not being able to access food in the house when I cook or eat, and of not finding random food unless there is a cue to look for it. This history may have set her up for success: in the context of the house, there is no reinforcement history for random free food.
Outside, it’s a different story. As I mentioned earlier, I let Game scavenge to her heart’s content when we’re out and about. It’s a dog thing she loves doing, and she rarely gets sick. So I don’t worry about it, and let her enjoy the things she finds (unless it’s human poop, because it grosses me out if my dog is going to stick their nose into my face later that day). Everything else – go for it. When she is on leash, I will tell her what a lucky girl she is when she finds something, and wait until she has finished eating. I’ll simply adapt her daily food ration when she eats out a lot. When I used to live in the center of Guanajuato, a guesstimated 10% to 30% of her caloric intake was scavenged on any given day. The rest was provided by me.
There are times when I test if my recall or leave it cues or marker cues are still sharp, and then I’ll interrupt her from eating or call her back right before. But that would be an exception rather than the norm.
My idea for this particular experiment was to only use a single environment outdoors, and always use the same food (kibble of a brand she doesn’t find out in the world). I thought I could get a cue transfer result in this location even though I let her scanvenge elsewhere.
Why have I not implemented a management strategy?
It’s just a deal breaker to manage the scavenging while I teach the kibble pile as a recall cue. She enjoys scavenging, and I do not want to take it from her. I am only teaching the kibble pile cue transfer to show you all how I would do it, not because I actually need it. If I needed this behavior, it would be a different story, and there would be strict management.
But since I don’t: in any other context, she has been continued to be allowed and encouraged to freely access whatever food she finds. The food she finds is also usually higher value than the kibble she’s currently on. So in a word – she has a VERY strong and long reionforcement history of freely accessing high-value food. And I may simply have been wrong about the fact that I could still teach a cue transfer in this one outdoors situation where she doesn’t usually find food.
I’m going to give it one more day (tomorrow), with entire hot dogs from my hand, and unless I see a result by tomorrow night, I’ll move on to a different strategy.
What would a cue transfer result ideally look like?
The result I want to see looks like the video below: Alicia and Dylan are currently working on recalls, following the protocol laid out in Calling All Dogs (an FDSA class I teach twice a year). They are at the stage where the dog is off leash, but the distraction is being protected by a barrier. Alicia uses a wire crate to keep the distraction safe. Dylan does not have a strong reinforcement history of freely accessing the kinds of distractions Alicia is working up to. In this case, the distraction is a bowl in the crate. Because we have practiced recalls away from the bowl (and other things) so often in different locations, first on a long line, and then with the crate, Dylan predicts Alicia’s recall here: he approaches the crate, notices the distraction, remembers that it reliably predicts a recall, and returns to Alicia before she has a chance to call him!
Alicia has not been aiming for this behavior – she has just been aiming for recalls off distractions. It happened anyways because that’s how cue transfers work when our reward is high value and our training and management are being consistent! That’s what I expected to see in Game by now. I am starting to suspect that we are not going to get that cue transfer with the kibble pile – but patience, grasshopper! I’ll give it one more day.
2 thoughts on “Distractions as cues, day 12: what if I’ve built a behavior chain?”
I fully agree with you:
The longer a behavior excists (scavenge),
the harder to change in.
Yes – it has a strong reinforcement history, and so do all the other links in the chain! It’s very interesting to observe!