4.2 Core Study 2: Fagen et al.
(elephant learning)
Fagen, A., Acharya, N., & Kaufman, G. E. (2014).
Positive reinforcement training for a trunk wash in Nepal’s working
elephants: demonstrating alternatives to traditional elephant training
techniques. Journal of applied animal welfare science, 17(2): 83–97.
The Psychology Being Investigated
The term operant conditioning is used simply to mean learning from the
outcome of our behavior. For example, when we perform a behavior that
has a good consequence, we are more likely to repeat it. Psychologists
would state that this behavior has been reinforced.
Positive reinforcers might include food or praise. Another type of
reinforcement is negative reinforcement, which involves the removal of an
unpleasant stimulus after a desired behavior occurs, making it more likely
for the behavior to be repeated.
Key Words
• Operant conditioning: learning through the consequences of
our actions.
• Positive reinforcer: a reward for behavior that fulfills a
biological need known as a primary positive reinforcer. A stimulus that is
associated with primary reinforcers can also be learned and is known as a
secondary reinforcer.
Operant Conditioning Theory
Something unpleasant is removed or avoided in response to a stimulus.
One example of this could be going outside and feeling unpleasantly cold
(stimulus) and then putting on a coat (response). We quickly learn that
putting on more clothes keeps us warm (reinforcement).
Some reinforcers occur naturally. These include things which help us to
survive, such as water, food, shelter, or sleep. Alternatively, secondary
reinforcers have to be learned and are associated with primary reinforcers.
Money is one example, as it can help us to acquire food, water, shelter,
and so on.
The theory of operant conditioning can explain how humans and animals
develop complex behaviors. The behavior of humans and animals can be
shaped gradually. A good example of this is how children are rewarded
with praise as they learn to speak. At first, a child’s vocalizations might
sound only slightly like real words, but with encouragement, these words
become more accurate, and the child’s speech is very clear.
Also, simple individual actions can be combined in sequences or “behavior
chains.” Like riding a bicycle, the rider must first learn to pedal, then steer,
then balance on two wheels.
However, the theory does not try to explain underlying, unseen reasons for
behavior, such as thoughts or feelings. According to operant conditioning
theorists, the majority of behaviors are learned through trial and error.
Activity 4.2
Sometimes humans try to train animals to behave in certain ways.
1. Think of at least two different training methods that are used for
pets and/or domestic animals.
2. Copy and complete Table 4.2 to explain how each method
might work, how the trainer would know if it had worked, and which method
you think is more effective.
| Description of training technique | How it works/how we can see if it
works | Which method is likely to be more effective |
Captive Elephant Training
Traditionally, captive elephants are given “free” (unlimited) contact with
their handlers (known as mahouts), and elephant behavior is managed
using punishment (Figure 4.3). This form of operant conditioning relies on
unpleasant stimuli such as pain or fear to shape behavior. One type of
punishment used to shape elephant behavior is the pain inflicted by (gây
ra bởi) a bamboo stick.
However, because of concerns for captive animal welfare and keeper
safety, there is increased interest in reward-based training (positive
reinforcement) using “protected” contact (a barrier between elephant and
handler). In particular, positive reinforcement methods have been shown to
improve the psychological well-being of elephants (Desmond & Laule,
1991).
Key Words
• Captive animal welfare: psychologists working with animals
that are confined (hạn chế) or outside of their normal environment must
ensure the health and well-being of the animals. As part of this captive
management, an animal’s natural needs for accommodation, environment,
freedom of movement, food, water, and care should be appropriately met.
• Positive reinforcement: a form of operant conditioning. It
involves rewarding desirable behavior to encourage it to be repeated. For
example, praising a child for saying “please” and “thank you” positively
reinforces good manners.
Secondary Positive Reinforcement (SPR) Training
One type of positive reinforcement training is secondary positive
reinforcement (SPR) training. This method teaches animals to associate
specific sounds with food; specific noises like these are called “markers” or
“sound-markers.” Once the animal is conditioned to the sound-marker
relationships, the marker can be used to reward wanted behaviors. Using a
sound marker is better than rewarding with food on its own as the marker
can precisely indicate to the animal when it has correctly performed the
desired behavior, whereas there is a delay for preparing and delivering
food. To maintain the association between secondary and primary
reinforcer, food rewards are still given at intervals.
Key Word
• Secondary positive reinforcement (SPR) training: training in
which a secondary reinforcer such as a sound marker is used and then
followed with administration of a primary positive reinforcer (typically food).
Background
Diagnosing and treating illness in elephants is an important part of captive
animal welfare. In order to avoid using traditional methods such as
punishment, Fagen et al. investigated the use of SPR training to teach
elephants to reliably and voluntarily engage in a trunk washing procedure
designed to maintain their well-being.
SPR has been used successfully with several animal species, including
pandas, primates, and antelope. One advantage the method offers is the
ability to shape captive animals’ behavior to help improve their health and
well-being.
In this study, the focus of using SPR was as a method for detecting
tuberculosis in elephants. Tuberculosis is a serious respiratory disease
which is a significant concern in the captive elephant population, and can
be passed between animals and humans. The disease is best detected
through taking a sample from the elephant’s trunk through a “trunk wash”
method. However, getting elephants to trunk wash correctly can be
challenging, with many samples being insufficient for testing.
Aim
The aim of this study was to see whether free-contact, traditionally trained
elephants can be trained to participate in a trunk wash by using positive
reinforcement.
Method
Research Method and Design
This was a controlled observation involving a small group of elephants
living in captivity who were trained over a period of weeks. The researchers
watched the elephants’ behavior in response to a specific stimulus and
used a behavioral checklist to record the elephants’ responses as a
percentage pass. This means the study can also be described as a
structured observation.
Sample
This study included five female elephants: four juveniles and one adult, all
housed at the same elephant stable in Nepal. The juveniles were between
5 and 7 years old and had been born at the stable. The adult elephant was
estimated to have been in her 50s. The elephants were chosen from others
at the stable as they were docile, not currently pregnant or looking after a
calf, and their mahouts were willing to take part in the study. The elephants
were all traditionally trained and in free contact with their mahouts. None of
the elephants had previous experience of SPR.
The elephants spent most of the day grazing in the jungle under the control
of their mahouts. They spent the rest of the day leg-chained in a stable with
the freedom to move 6–8 feet (1.83–2.4 m) around the stake. The
elephants’ diet was fresh grasses, grain, and nutritional supplements, and
they were given access to water at a river during grazing hours.
Activity 4.3
Humans share a number of cognitive and behavioral traits with animals.
1. Explain the reasons why psychologists might use animals
instead of human beings in their research.
2. As discussed in Chapter 1, the British Psychological Society
(BPS) Guidelines for Psychologists Working with Animals (2020) include a
number of ethical guidelines on the topics of:
• replacement
• species
• number
• procedures
• Pain and distress
• Housing
• Reward
• Deprivation and aversive stimuli
Design an advertisement that could be sent to organizations who look
after captive elephants, in order to recruit animal participants for the study
by Fagen et al. The advertisement should focus on how the well-being of
the animals will be maintained throughout the research. Choose at least
two appropriate ethical guidelines from those listed above in your
advertisement.
Procedure
The method of training used in this study was SPR, with chopped banana
as the primary reinforcer and a short whistle blow as the secondary
reinforcer. Training was conducted during the indoor sessions (7:30 a.m. to
10:00 a.m. and 4:00 p.m. to 7:00 p.m.) by a trainer, with the mahout
present for safety.
The mahouts stood to the side and did not speak to or signal to the
elephants. No elephant went longer than two days without a training
session. Elephants could choose not to engage with a session by turning or
walking away from the trainer.
The purpose of training was to teach the elephants to perform a voluntary
trunk wash in several behavioral steps, actively moving their trunks in
response to a cue. After establishing the marker-reward relationships
(banana-whistle), elephants were taught using the following three methods:
• Capture: waiting for the animal to perform a behavior naturally,
then “capturing” it by marking it with a reward.
• Lure: for non-natural behaviors, an animal is “lured” into a
certain body position by placing a reward in a certain place.
• Shaping: after starting either capture or lure, rewards are then
only given for the behaviors that are “best,” i.e., incrementally closer to the
goal behavior.
Task
Using these training methods, elephants were trained to do the following
behavioral tasks separately (see Table 4.3 and Figure 4.4).
Training Details
There were three other tasks (targeting, trunk down, trunk out) that were
introduced, but partway through the training were discarded as they are not
essential for the performance of a trunk wash. After the elephant performed
the individual behavioral tasks, each task was then paired with a verbal
cue. The verbal cue was a one-syllable word with no meaning in English or
Nepali to avoid having any meaning to the elephants or mahouts.
Once all the behavioral tasks were established, the trainers put the
behaviors together in small sequences in a process known as behavior
chaining. This technique meant that performing the first behavior in a
sequence correctly earned the animal the chance to perform the second
behavior correctly and gain a reward. Separate behavioral tasks were at
first paired (starting with bucket and blow), then put together in longer
sequences until they formed the entire trunk wash procedure.
In addition, the trainer also introduced the use of a syringe to the trunk-here
position. This was done incrementally using a desensitization method.
Over a series of repetitions of the whole trunk wash sequence, the syringe
was gradually brought closer to the elephant’s trunk, then touching the
trunk, then inserted, then inserted with increasing amounts of fluid until the
elephant tolerated the full 60ml required for sample collection.
To avoid the elephants drinking the saline or water, they were offered
drinks before each training session. One elephant preferred drinking saline
and would drink the saline solution after rejecting the drinking water, so
was switched to a water solution for the behavioral task.
There was no time limit put on each stage of the training process; it was
determined by the success of the individual elephant, meaning that the
training plans varied according to the individual elephant’s needs.
Assistant’s Role and Results
An assistant recorded the length of each training session in minutes. The
assistant recorded the total number of times the elephant was given a cue
or “offer” for behavior. After Session 10, elephants were tested
approximately every five sessions on previously taught behaviors. A
passing score was 80% (i.e., eight or more correct out of ten offers/cues).
Once the 80% or higher score had been achieved on a sequence, this was
considered a “pass.” Once the whole behavioral sequence was “passed” at
a rate of 80% or higher, the training was considered complete.
Key Words
• Behavior chaining: a process that allows separately trained
behaviors to be performed in sequence in response to cues.
• Desensitization: to reduce and extinguish a response (e.g.,
fear) to stimuli (e.g., a syringe).
Research Methods
It was important in terms of validity that the mahouts did not assist with the
training. For this reason, they did not speak or signal to the elephants.
Verbal cues used in the study were also designed to be distinct from other
verbal cues the mahouts used with the elephants. This was an important
control in the study.
Results
It was found that the four juvenile elephants successfully learned the trunk
wash; however, the adult elephant did not (see Table 4.4). Elephants 2 and
4 never passed certain tasks in the sequence.
Results and Observations
Their steady test was passed, but they were able to pass their full trunk
wash tests. Elephant 5 did not pass her blow-into-bucket, desensitization to
syringe, or steady test. There are a number of factors which may have
influenced the behavior of elephant 5. She was older and likely had both
some visual impairment and trunk weakness. The elephant was reported
as being distracted and impatient during the last week of the study and had
a foot abscess during this period. Also, a calf from an adjacent stall
interrupted some of her sessions by entering the training area.
Another result was that some behavioral tasks were more difficult than
others. For example, the trunk-here task required more offers/cues than the
bucket or blow-into-bucket tasks.
The elephants gradually improved their performance over time (see Figure
4.5). The mean success rate went from 39% after 10 sessions of training to
89.3% after 35 sessions of training (mean percentage score never reached
100% as 90% was given as default to individual behaviors within a
sequence, once the sequence was “passed”).
Figure 4.5
Mean percentage correct among all elephants for all tasks during
each test
• The graph shows improvement in success rates from 36.00%
at session 10 to 89.33% at session 35.
• The percentage success steadily increased over time with key
milestones at 47.33%, 59.39%, 83.18%, and 86.55% across sessions.
Conclusion
The researchers concluded from this study that juvenile, free-contact,
traditionally trained elephants can be trained to participate in a trunk wash
using only SPR training techniques. Moreover, this training can be carried
out with the voluntary participation of the elephants, avoiding punishment,
to produce reliable results.
Strengths and Weaknesses
This study involved a small group of female elephants. Using a small
sample in a psychological study can make it difficult to generalize the
results to a larger population. As the elephants were captive animals from
one elephant stable, it would be hard to say that they are representative of
the general population.
The researchers mention that elephants were chosen for their
temperament (docile). Because of their large size, elephants can pose a
risk to themselves and humans, which is a major challenge for those
conducting research with them. Other practical challenges included trying
to control for distractions to the elephants. The presence of tourists and
other elephants during the training sessions might have affected the
elephants’ concentration or willingness to participate.
Controlled observations such as this can be easily replicated, using the
same observation schedule. This means the study can be tested for
reliability. However, the researchers state that the training sessions were
flexible to suit the mood and ability of each animal. If this study was
repeated with other elephants, it is likely that the experience of the trainer
and personalities/conditions in which elephants were kept (e.g., their stress
levels, previous experience of SPR training) could affect the results. This
creates an issue for reliability.
There were attempts to maintain the validity of the research; for example,
the mahouts were asked not to assist during the sessions.
Ethical Issues
When using animals as participants, different sorts of ethical issues must
be considered. In this instance, the elephants were reported to be well
treated and did not appear to have been physically harmed as a result of
the research. The elephants were able to graze freely for a large portion of
the day, were fed adequately, and were able to socialize with the other
elephants in the jungle. Investigating the effectiveness of positive
reinforcement training in elephants that are already captive rather than
capturing and using wild animals is a more ethical approach.
The study was carried out with the intention of developing a method for
producing usable trunk wash samples. The purpose of this was to help with
captive management of elephants and to enable tuberculosis to be quickly
diagnosed and treated, which is highly beneficial to the elephants. The
researchers actively avoided using punishment and instead focused on a
training method which minimized harm to the animals.
Summary
Fagen et al.’s study used a small group of five female elephants to show
that trunk washing can be learned through positive reinforcement. Through
a lengthy process of training and testing, researchers were able to
establish that most of the animals could perform a series of actions
designed to act as a health check. This was a controlled observation with a
small sample, meaning it had limited generalisability and could be difficult
to replicate. Using objective, quantitative measures of correct responses,
however, did show that juvenile elephants can learn trunk washing using
SPR training.
Questions
4. What was the purpose of teaching elephants the trunk washing
task?
Questions
5. The elephants’ mahouts were present in each session.
However, they didn’t look at the elephants or speak to them. Can you
explain why the mahouts were asked to behave this way?
6. What influence do you think individual differences between the
elephants had on the results of this study?
