3.2 - Fagen et al.: Elephant Learning
Operant conditioning
Operant conditioning is a learning process where behaviours are modified through consequences, such as rewards or punishments. This occurs because behaviours followed by positive outcomes become more likely, while those followed by negative outcomes become less likely.
Key processes in operant conditioning
- Shaping - This involves gradually modifying an animal's behaviour through successive approximations. Rewards are provided for actions that increasingly resemble the desired behaviour, building up to the final response over time.
- Behavioural chaining - Animals learn a sequence of behaviours by first mastering each one individually through shaping. Reinforcement is then given only when behaviours are performed in order, adding more steps until the full chain is complete.
Types of reinforcement
Reinforcement can be positive or negative and primary or secondary, each influencing behaviour in distinct ways. Punishment, by contrast, decreases unwanted behaviours.
| Type of reinforcement | Description | Example in elephant training |
|---|---|---|
| Positive reinforcement | Adding a desirable stimulus after a behaviour to increase its likelihood | Providing chopped banana when the elephant raises its trunk |
| Negative reinforcement | Removing an unpleasant stimulus after a behaviour to increase its likelihood | Stopping a prod with a stick once the elephant raises its trunk |
| Punishment | Adding an unpleasant stimulus after a behaviour to decrease its likelihood | Prodding the elephant with a stick for incorrect trunk positions |
| Primary reinforcement | Using a stimulus that directly satisfies a biological need | Offering chopped banana as a reward for trunk raising |
| Secondary reinforcement | Using a stimulus that has been paired with a primary reinforcer, without directly meeting a biological need | Blowing a whistle after pairing it with banana rewards through classical conditioning |
Background to the study
Traditional elephant training by mahouts (handlers who work with elephants) often relies on punishment and negative reinforcement, raising concerns about animal welfare and handler safety. In response, Fagen et al. explored alternative approaches using rewards to teach a specific routine called a 'trunk wash'. This procedure involves a chain of behaviours needed to test elephants for tuberculosis, a disease that affects both animals and humans.
The study focused on secondary positive reinforcement, where a neutral stimulus like a whistle is first associated with a primary reinforcer such as food. This association allows the whistle to act as a reward on its own, making training more efficient and less reliant on constant food provision.
Aim of the study
The aim was to determine if secondary positive reinforcement could effectively train elephants to perform a trunk wash voluntarily. This would provide a safer, more ethical alternative to traditional methods, improving welfare for both elephants and trainers.
Methodology and sample
Fagen et al. conducted a structured, non-participant observation, a research method where behaviours are systematically recorded without the researchers directly interacting or influencing the subjects during data collection.
Data was gathered using a behavioural checklist to assess success in completing trunk-wash behaviours. Variables included training time in minutes, number of verbal cues given, and success rates for each behaviour (defined as 80 per cent accuracy, meaning correct performance in at least eight out of ten attempts). Observations occurred every fifth session after the tenth to track progress.
Sample details
- Size and demographics - Five female elephants from a stable in Nepal: four juveniles aged 5-7 years, born at the stable, and one adult in her 50s.
- Sampling technique - Selected based on docile temperament (easy to handle), non-pregnant status, and availability of handlers.
Procedure
The elephants were trained in their stalls using secondary positive reinforcement, shaping, and behavioural chaining to learn the five-step trunk wash. Sessions lasted 2.5-3 hours, with standardised verbal cues (short, one-syllable words) to prompt behaviours.
The five trunk-wash behaviours
- Trunk here: Placing the trunk into the trainer's hand for saline insertion via syringe.
- Trunk up: Lifting and holding the trunk to allow fluid to flow.
- Trunk down: Lowering the trunk into a bucket.
- Blow into the bucket: Expelling fluid to collect a sample.
- Hold steady: Keeping the trunk still during the process.
Training techniques used
- Capturing - Rewarding naturally occurring desired behaviours with banana.
- Luring - Using banana positioning to guide the elephant towards the behaviour.
- Desensitisation - Gradually introducing the syringe (an aversive stimulus) closer to the trunk, rewarding tolerance each time.
- Counter conditioning - Pairing the syringe with positive rewards like banana to change the elephant's association from negative to positive.
First, the whistle was classically conditioned to pair with banana as a secondary reinforcer. Behaviours were then shaped individually and chained together, with rewards given immediately after correct responses.
Controls and ethical considerations
- Controls - Standardised procedures, such as consistent verbal cues and the same behavioural checklist with operational definitions (clear, measurable descriptions of behaviours), ensured reliability.
- Ethical issues - Elephants were chained but could walk away if unwilling, minimising distress. Positive reinforcement avoided aversive stimuli, reducing psychological harm to animals and physical risks to trainers.
Results
The study yielded quantitative data on training efficiency and success rates, highlighting differences between age groups.
Key findings
- Juveniles completed the full trunk wash in 25-35 sessions, averaging 12 minutes per session (total around 367 minutes).
- The adult elephant did not master the full sequence, failing 'blow' and 'hold steady', and never fully desensitised to the syringe.
- Two juveniles struggled with 'hold steady' in isolation but succeeded in the full chain.
- Overall success rates improved from 39 per cent to 89 per cent after 35 sessions.
- 'Trunk here/trunk in hand' was hardest (295 reward offers needed), while 'blow into bucket' was easiest (54 offers).
These results demonstrate the effectiveness of the training for younger elephants but reveal challenges with older ones.
Conclusions
Secondary positive reinforcement proved effective for training juvenile elephants to perform a trunk wash voluntarily and reliably. However, it was less successful with the adult, suggesting age-related limitations. This approach offers a humane alternative to traditional methods, enhancing welfare and safety.
Evaluation
The study's strengths and weaknesses can be assessed in terms of reliability, validity, and generalisability.
| Strengths | Weaknesses |
|---|---|
| Reliability - Clear operational definitions in the behavioural checklist ensured consistent judgements on success or failure, making results replicable. | Subjectivity - A single observer recorded data, potentially introducing bias if they hoped for positive outcomes. |
| Validity - Mahouts were instructed not to speak or signal, isolating secondary positive reinforcement as the cause of behavioural changes. | Generalisability - The adult's poor performance might stem from her specific health issues (e.g., vision problems), limiting applicability to other older elephants. |
Issues and debates
Individual and situational explanations
Individual factors, such as the adult elephant's age and potential physical health issues, may explain her inability to learn the full trunk wash. Situational factors could also play a role, including trainer biases towards her age or distractions from a young calf entering her stall.
Application to everyday life
The training method provides a practical, ethical way to conduct health checks on elephants, reducing tuberculosis risks for animals and humans. It can extend to other veterinary procedures, like blood sampling or X-rays, lowering stress and improving overall wellbeing in captive animals.
The use of animals in research
The small sample size aligns with ethical guidelines to minimise animal use. Distress was reduced through positive reinforcement and the option to disengage, prioritising welfare while gathering necessary data.