Dr. Joe Layng earned a Ph.D. in Behavioral Science (Biopsychology) from The University of Chicago, where he collaborated on research, working with pigeons, investigating the adduction of untrained complex symbolic behavior from simpler behavioral components.
He has over 35 years of experience in the experimental and applied learning sciences with a particular focus on the design of instruction. His experiences working with both children and animals give him a unique perspective on how we can combine science and observation, or practical experience, to develop teaching programs that are user-friendly and effective.
As I mentioned in my report on Dr. Reid’s presentation, one of the themes of this conference was understanding how we can teach our animals to learn and maintain chains and sequences of behavior. Dr. Reid got us started by looking at behavior chains and how they lead to the development of behavioral skills. Dr. Layng took us a little deeper into the subject by looking at the differences between different types of sequences and why this matters.
As a matter of fact, he started right off by saying that one way to get around chaining theory is to recognize that sequences are not always chains. When I heard him say this, I had to think for a moment because I had been taught that chains and sequences were different ways of combining behaviors so that the animal performed several behaviors for one reinforcer at the end. You were either teaching a chain or a sequence.
But, it turns out, that in the human instructional literature, the term “sequences” refers to the superset (larger category) and chains are one type of sequence. So, we should not be talking about chains vs. sequences, but about the different types of sequences. It may seem like a minor shift in terminology but I think it matters because it changes our understanding of what controls each unit in the sequence. And, as we will see later, this connects nicely to the question of whether your goal is for the animal to achieve autonomy (or not) when teaching a behavior chain.
As part of this discussion, Dr. Layng did mention that there are 50-75 years of understanding of human behavior that do not exist in the scientific literature. This is because the people who are studying education are usually separate and isolated from the people doing research in other related fields (behavior analysis, psychology, etc.).
This chart shows where chains are in a repertoire hierarchy that ranges from simple responses to complex cognitive processes. Compared to humans, animals have a more limited repertoire (marked in red) and are not capable of everything listed in this chart.
- These are the basic building blocks of any sequence.
- There are a variety of linkages.
- A true link – you cannot do the second step unless you do the first step.
- There are other types of sequences including kinesthetic repertoires and conjunctive chains. (He didn’t discuss these)
The main determining factors are the requirements for reinforcement (does it have to be done in a specific order?) and whether there is a change in stimulus conditions that the individual uses as information about when to go on to the next step.
- Tandem units: Step two can occur without step one occurring, but will only be reinforced if a set sequence – step one, step two, etc. – is followed (when primarily verbal, serial learning).
- Conjunctive units: Any step may be performed in any order, but all steps must be performed.
- Chained unit: Step two can only occur if step one has occurred.
He had the following graphic to show the differences between the three types of sequences:
- You are presented with a corral with separate “enter” and “exit” gates.
- You are only reinforced if you go in the “enter” gate and out the “exit” gate.
- There’s nothing in the environment that prevents you from doing other combinations – enter and exit through the “enter” gate, enter and exit through the “exit” gate, enter through the “exit” gate and exit through the “enter” gate – but none of these will earn reinforcement.
- You are presented with two corrals.
- Each corral has one gate, which is used for both entering and exiting.
- To earn reinforcement, you must enter and exit each corral, but the order doesn’t matter. You can go in and out the one on the left first, and then go in and out the one on the right, or vice versa.
- You are presented with two corrals that are connected.
- Each corral has only one gate – either an “enter” or an “exit” gate, but the exit gate is closed.
- The only possible way to do the sequence is enter through the “enter” gate,” which allows you access to the second corral and the “exit” gate – which opens for you when you approach from inside the corral.
- There’s only ONE way to do the behavior.
My takeaway from these examples was that the “chain vs. sequence” distinction is more accurately described as the difference between a tandem sequence and a chain sequence.
Example of a Tandem Sequence: Piano playing (assuming you are following music)
- You can play notes in any sequence. There is no physical limitation on which notes you can play. You could play ADE even if the music calls for ABC.
- The completion of one note may occasion the next, but is not required to play the next. It’s possible to play the wrong note, but the emphasis here is on the sequence (and the quality of each note).
- Moving on through the music is a criterial issue – Criteria follow a fixed response sequence. You play the notes in the correct order because that’s the requirement for reinforcement.
- Component sequences can be practiced to fluency. You can take out one section and practice independently of the rest.
- Beginning anywhere in the sequence is important. Can you pick up in the middle? recover after an error?
- A side note: good musicians learn to recognize and match sequences – that’s what makes for flow between musicians and in the different parts of a song.
- You can read more about motor performance analysis in the article by Francis Mechner (www.mechnerfoundation.org).
Example of a Conjunctive Sequence: Setting the table
- A series of separate steps – put out the plates, glasses, silverware, napkins, etc.
- Each step can be done at any time. Some might be more efficient like putting the napkin out before the silverware, so you don’t have to lift the silverware to place the napkin, but it’s not a requirement that the steps are done in that order.
Example of a Conjunctive Sequence: Sweeping the room
- You can start anywhere.
- You have a list of behaviors you need to complete – sweep corners, sides, center, etc. in order to earn reinforcement.
- But, the behaviors can be done in any order. You can start with the corners and sweep to the center, or start at the back and sweep toward the front, or …
Example of a complex chain sequence: Suiting up in biohazard gear
- Putting on protective gear must be done in a certain order to ensure that there is no exposed skin or gaps where contamination could enter
- In order for it to be a chain, there must be a requirement before moving on to the next step. You cannot do the booties unless you have put on the bunny suit because you have to tape the top of the booties to the outside of the bunny suit.
- The chain can be viewed as a series of stimulus – response pairs.
- When building a complex chain, look at what is already in the individual’s repertoire and then teach the components that need to be added.
- You can pull out and practice some parts of the chain (to learn the skills involved) but you can’t practice by starting anywhere in the chain and continuing to the end (there will be some components that rely on previous steps that you skipped).
Want to complicate it further?
Joe ended with another example of serial learning – Algorithms. An algorithm is a rigid sequence of steps which enables a performer to get to the answer for any problem drawn from the domain to which the algorithm applies.
Examples of Algorithms: Long division, artificial intelligence
Note from Katie:
This is all pretty interesting, but what does it mean in practical terms? And how does it apply to animal training?
These are the questions that immediately came to my mind after listening to Joe’s presentation. I also found that I still had some confusion over the differences between chain and tandem sequences, particularly as they applied to animal training. Most of Joe’s examples were from human instruction, which is relevant for those of us who are teaching other people to train, but I wondered how this information was useful when our learners are animals. For example, we often teach animals to do a series of behaviors in a specific order. Are these tandem sequences or chain sequences?
I decided to ask Joe during one of the breaks. One of the great things about ASAT is that there are many opportunities to talk with the speakers, so it was easy to find him and have a chat. Actually, I had several conversations with him over the course of the weekend and a few days later, he sent me some additional reading material. Based on that, here is my understanding of the difference between tandem sequences and chain sequences, as it applies to animal training and cued behaviors.
The biggest difference is that, in a chain sequence, the individual has to be taught how to recognize when she is ready to go to the next step. In most cases, her action (the behavior) in the current step produces a stimulus change that makes it possible for her to continue to the next step. Additionally, if he goes to the next step too soon, he won’t be able to do it correctly. The following example was included in one of the references that Joe sent me.
Example of a chain sequence: Sharpening a pencil
The article described the process of teaching a young child to sharpen a pencil. Most of us would write out the individual steps and leave it at that. But, having a list of steps does not always guarantee success. Anyone who has every tried to follow factory instructions to assemble or install something can confirm this!
Instead, each step is broken down so that it specifically addresses what the child must learn in order to do each step correctly. This includes the physical actions and the changes in the stimulus conditions that indicate that he is doing the step correctly and when it’s time to move on to the next step. This can be described as a stimulus – response – consequence “unit” that tells the learner when to start, what physical action to do, and when they are done. For example,
- broken pencil -> walk and look -> find sharpener
- sharpener at hand -> push in pencil -> won’t go further
- won’t go further -> turn handle – > grinding sound
- grinding sound -> …
This is a chain because each consequence becomes the stimulus for the next step, and in order to proceed, each step must be done correctly. When teaching a chain, the teacher must teach the child how to do the physical actions associated with each response – or at least verify the child can do them – and then she has to teach the child to recognize the stimulus change (the consequence) that tells him when he’s ready for the next step.
Let’s compare this to a tandem sequence:
In a tandem sequence, the individual only needs to learn the behaviors and the order. Think about tandem sequences like reciting the alphabet or a poem. You learn these sequences through association: A is followed by B, which is followed by C, etc.. You don’t have to wait for a change to occur after you say “A,” before you say “B.” You just follow the steps in order until you get to the end.
If we take this back to animals, I can think of several different scenarios where I might ask my horse to do a series of behaviors. I asked Joe about a few of these, to see which were chain sequences and which were tandem sequences. I was curious where cues (given at a specific time by the trainer) fit into the picture. I asked Joe what it would be called if I taught my horse to do three different behaviors and asked her to do all of them (as a sequence), but under different conditions.
Example: The horse has to walk, trot and canter before I reinforce her.
- This is a sequence because she must do all three behaviors before I reinforce her.
- It’s not a conjunctive sequence because I care about the order.
- It’s a chain sequence if she has to wait for me to cue her before proceeding to the next step. She is looking for a stimulus change (my cue) that tells her when to move on to the next step.
- It’s a tandem sequence if I teach her to do the behaviors in order, but she does not have to wait for a cue from me. In this case, she knows the pattern and moves through it on her own, without there being specific criteria that tell her when to move to the next step.
- It’s a tandem sequences if I teach her to do the behaviors in order, even if I am consistent about when I ask for the behaviors (trot at the corner, canter at the next corner). These “location cues” may function as cues within the sequence, but they are not true cues because I have not taught her that she should always trot (or canter) every time she goes through a specific corner.
- You can have a long sequence that contains both chain and tandem sequences.
I picked this specific example because when I ride, I often ask for many behaviors before I click and reinforce and it’s never been clear to me if I am creating chains or sequences (the distinction I had previously made). I had been taught that in a chain, the cue reinforces the previous behavior and while I sometimes want to take advantage of that, there are places in the learning process where I throw that idea out the window and just concentrate on teaching my horse to do behavior A, then behavior B, then behavior C – even to the point where I let her practice the sequence without a lot of input from me. Once she knows the general pattern, then I can work on fine-tuning and teaching her when and how to wait for my cues. After talking to Joe, I understand a little more about when and why it might be appropriate to think in terms of chain sequences vs. tandem sequences and the benefits of each.
A few final thoughts:
- You may find that recognizing different types of sequences changes how you teach sequences of behavior.
- The learner needs more information when learning a chain sequence then when learning a tandem sequence because she needs information about what should happen at each step and when to move on.
- After reading the references he sent me, it seems to me that chains lead to a higher level of “quality control” because each step depends on the previous one and the transitions between steps can be more tightly controlled.
- A better understanding of chain sequences makes it easier to troubleshoot them. If the individual skips a step, or stays too long on a step, perhaps it’s not just about strengthening the individual behaviors. Instead, it may be necessary to clarify what stimulus conditions indicate when it’s time to move on.
- Dr. Reid talked about how behavior chains lead to behavioral skills. I can think of situations in which I might start off teaching a behavior chain and then allow the horse to modify and adjust it as she learned, even using new “cues” (practice cues?) that were more relevant to her. Is this what it means to transition from a behavior chain to using behavioral skills?
- Where does autonomy fit in? Is the process different if you are learning a chain sequence or a tandem sequence?
While I try to be accurate in my note taking, there may be some errors either in my understanding or my presentation of the information from this talk. If you have questions, feel free to contact me or leave a comment.
Thanks to Dr. Layng for allowing me to share his presentation and for sending me additional resources. Thanks to the ORCA students and to the organizers of the Art and Science of Animal Training Conference for all their hard work on putting on this great event.