Which Works Better for Dogs?

---

This winter, my training partner Amanda and I decided to teach our dogs “toy identification”. We lead a small weekly training class at my training center and one of the behaviors that the class expressed interest in was training their dogs to retrieve toys by name. Amanda and I selected an approach to teaching this behavior that is commonly referred to as Errorless Learning.

Here are the steps that we used with our students and with our own dogs (Note: This is NOT the only way these behaviors can be trained; it is simply the approach that we chose). A short video showing each step with Stanley is below:

• STEP 1: Select one or more toys that are of high value to the dog. For my dogs, this is any toy that squeaks or is fuzzy. So, I chose Pig, Squirrel and Chicken for my training with Stanley and Alice.

---

• STEP 2: Introduce the toy name (repeating several times), followed by “get it” (or touch if using a touch target as the indicator). Toys and their names are introduced singularly, sequentially. For example, I taught Stanley Chicken, then Squirrel, then Pig, each for several weeks, not overlapping.
• STEP 3: This is the first step in which a discriminating stimulus (another toy) is introduced. To minimize errors as much as possible, we introduced toys that were unnamed and were considered (by the dog, not the trainer) to be very low value. For Stanley and Alice, these were a set of retrieving toys that they would retrieve, but which were not their preferred toys. This sets up a situation in which the chance of the dog making the correct response is very high (named high-value toy with strong and recent +R history versus unnamed low-value toys with no recent +R history). The dog is asked to retrieve or indicated the named toy among the unnamed, low value toy(s). This step is repeated independently and sequentially for each named toy.
• STEP 4: This step introduces the concept of fading – introducing a competing stimulus (in this case, other high-value, named toys) at a very low level of intensity. The intent is again to minimize errors – choosing incorrectly – when asked to retrieve the named toy. The fading approach that we used was to place the other named toys under a basket, making them much less accessible. An alternate approach would be to place the competing toys a distance away, with the named toy that the trainer is requesting placed much closer.
• STEP 5: The final step introduces a competing named toy at the same intensity as the requested named toy. For example, Stanley may be asked to retrieve “Squirrel” while Chicken is available in the pile and Pig is still under the basket (and vice-versa). In all permutations, we drop back to having one or more of the named toys in the basket if repeated errors occur.
• VIDEO OF STANLEY’S TRAINING: A summary of these steps and video of Stanley’s progress is below.

---

Video: An Errorless Learning Approach to Toy ID with Stanley

---

Overall, Stanley has progressed consistently with this training approach and he is currently reliable with his three named toys for several reps per training session. So, one might ask whether or not this approach was more (or less) effective than a traditional trial-and-error learning approach in which Stanley was presented with the three toys and was reinforced only when he selected the correct toy (and ignored when he selected incorrectly).

---

First, let’s get a bit into the weeds regarding what errorless learning is and how it is related to traditional trial-and-error-learning.

What is Errorless Learning?

The term “errorless learning” refers to a training protocol in which the learner (in our case, the dog) is induced to demonstrate a desired behavior in situations that minimize opportunities to choose an incorrect response. This provides repeated opportunities to the trainer to positively reinforce all correct responses (in our example, click-treat) while reducing instances in which the dog chooses incorrectly and is not reinforced. Over time, other choices (technically referred to as discriminating stimuli) are gradually introduced at a low level of intensity – they are “faded in”. As the dog becomes proficient, the discriminating stimuli can be presented at increasing levels of intensity.

How does it Differ from Trial-and-Error Learning?

The traditional approach to trial-and-error learning would present two or more discriminating stimuli to the dog and +R correct choices, while ignoring incorrect choices. In our example of teaching toy identification, this might involve presenting several toys to the dog, naming one toy, and then repeatedly +R the dog for selecting that toy while ignoring incorrect selections. Errorless learning is simply a form of trial-and-error in which the trainer manipulates the situation in some way to increase the dog’s inclination to choose correctly (especially during early stages of learning).

Are There Demonstrated Benefits?

Errorless learning has been studied in many species, including human subjects. Although not all studies have shown measurable benefits, those that are reported include:

• Fewer errors when discriminating stimuli are introduced (in our case, this would be other toys)
• Increased rate of learning
• Reduced frustration due to making repeated errors (this is a big one for dog trainers)

Are there Errorless Learning Studies with Dogs?

Most of the research examining errorless learning has been conducted with pigeons, lab animals, or human subjects. However, two 2021 studies successfully used this approach (with faded-in stimuli) to train scent discrimination in working dogs (1,2). However, while they were successful, the researchers in those studies did not compare the errorless approach to a traditional trial-and-error technique. Recently, a group of researchers at the University of Adelaide School of Animal and Veterinary Sciences in Australia made this comparison (3).

The Study: Eight adult dogs were trained to discriminate between two colored circles – blue versus yellow (note: dogs can see blue and yellow hues). One group of four was trained using errorless learning by fading in the alternate color. The second group was trained using trial-and-error (both circles were available to the dogs at normal intensity; only selection of the correct circle was +R). Positive reinforcement was in the form of treats dispensed from an automatic treat dispenser (see paper for complete experiment design details). The researchers measured the number of correct and incorrect responses, number of non-responses, and frustration-related behaviors in all of the dogs.

Results: Although the sample group was small, the researchers did find several significant differences between dogs trained with errorless learning versus those trained with trial-and-error:

• Learning Success: Dogs who were trained with errorless learning had a higher number of correct responses overall and were more likely to successfully learn to discriminate between the two circles when compared with the dogs trained using trial-and-error.
• Learning Rate: Dogs trained with errorless learning also demonstrated faster rates of behavior acquisition – they needed fewer training sessions and reps to become successful.
• Frustration: They also showed fewer frustration-related behaviors when compared to dogs who were trained with trial-and-error. The trial-and-error dogs were also more likely to stop responding during a training session than were dogs trained with errorless learning.
• Individual Dogs: The researchers reported several learning response differences among dogs that may suggest individual preferences for training approaches. The small numbers in this study did not allow any conclusions about this observation, but suggest that differences in breed, age, and training experience may be important criteria to consider (and to study in future experiments).

Conclusions: The researchers conclude that this study provides some evidence, though from a small sample of dogs, that using an errorless learning approach that includes a faded-in stimulus may have benefits over traditional trial-and-error learning approaches. The benefits that are suggested by their study include more rapid behavior acquisition, reduced errors, and reduced levels of learning-related frustration in the dogs.

---

---

Take Away for Dog Folks

Errorless learning has been a buzzword among dog trainers for quite a while. The research with scent discrimination that used this approach (for COVID-detection dogs, no less) and now this newest study, provide us with some evidence that our attempts to reduce errors when teaching our dogs new behaviors are well-founded. As we have hoped, helping our dogs to learn by setting up situations in which they are highly likely to choose correctly can improve their learning efficiency, reduce risk of not responding (shutting down) and decrease frustration.

Personally, I have been very pleased with Stanley’ and Alices’ responses to our approach to errorless learning and to fading-in discriminating stimuli was we have taught them Toy ID.

So, for all of the trainers out there – Are you using errorless learning in your training program or with your own dogs? If so, what types of behaviors are you teaching with this approach? If you are fading in competing stimuli, how are you doing this? (I see lots of opportunities for creativity here!).

---

Comment below and share your thoughts about and experiences with Errorless Learning in Dog Training!

---

Cited Studies:

1. Essler, J.L., Kane, S.A., Nolan, P., Akaho, E.H., Berna, A.Z., DeAngelo, A., Berk, R.A., Kaynaroglu, P., Plymouth, V.L., Frank, I.D. and Weiss, S.R., 2021. Discrimination of SARS-CoV-2 infected patient samples by detection dogs: a proof of concept study. PLoS One, 16(4), p.e0250158.
2. Vesga, O., Agudelo, M., Valencia-Jaramillo, A.F., Mira-Montoya, A., Ossa-Ospina, F., Ocampo, E., Čiuoderis, K., Pérez, L., Cardona, A., Aguilar, Y. and Agudelo, Y., 2021. Highly sensitive scent-detection of COVID-19 patients in vivo by trained dogs. PloS one, 16(9), p.e0257474.
3. Handley, K., Hazel, S., Fountain, J. and Fernandez, E.J., 2023. Comparing trial-and-error to errorless learning procedures in training pet dogs a visual discrimination. Learning and Motivation, 84, p.101944.

---

Study with Us at The Science Dog Courses!

Completely on-line and on-demand courses and webinars for dog owners and pet professionals.

Two Masters Class Certificate Programs!

---