Animal Learning – Reinforcement, Schedules, and Behaviour-Change Techniques

Learning Objectives

• Introduce core learning-theory terminology and concepts for newcomers.

• Encourage experienced learners to re-examine familiar material from fresh angles (e.g.
  considering ethology and emotion in addition to classic Pavlovian or Skinnerian views).

• Equip you to:

  • Explain why animals alter behaviour after experience.

  • Plan scientifically sound training programmes (hands-on methodology is addressed later and in assessments).

Principles of Reinforcement

• Reinforcement ≈ a change in the animal’s affective (emotional) state that makes the behaviour it follows more likely.

• Punishment ≈ a shift in affect that makes the preceding behaviour less likely.

• Key vocabulary:

  • Primary (unconditioned) reinforcer – innately valuable (e.g. food, water, warmth, social contact, pain relief).

  • Secondary (conditioned) reinforcer – gains value through pairing with a primary one (e.g. clicker sound, “good dog”).

• Ethical & welfare angle:

  • Emphasise positive reinforcement to enhance welfare; minimise aversives.

  • Acknowledge that any procedure that manipulates affect has moral implications.

Reinforcement Schedules

• Schedules = rules describing when each behavioural response will be reinforced.

Continuous Reinforcement (CRF)

• Every occurrence of the target behaviour earns a reinforcer.

• Advantages:

  • Rapid acquisition.

  • Clear information about contingency.

• Disadvantages:

  • $\text{Rapid extinction}$ once reinforcement stops (animal notices immediately).

Partial / Intermittent Reinforcement

• Only some responses are reinforced.

• General pattern:

  • $\text{Slower acquisition}$ compared with CRF.

  • Far greater resistance to extinction; animal keeps “checking” if payoff will reappear.

• Four canonical sub-types (Skinner, 1938):

  • Fixed Ratio (FR)

  • Reinforce after a set number of responses, e.g. FR-5 ⇒ every 5th response.

  • Generates high, steady rates; brief post-reinforcement pause.

  • Performance often described by a “stair-step” cumulative record.

  • Variable Ratio (VR)

  • Reinforce after an unpredictable number; centred on a mean, e.g. VR-20.

  • Produces the highest response rates; very hard to extinguish (gambles, slot-machines).

  • Fixed Interval (FI)

  • First response after a fixed time interval earns reinforcement, e.g. FI-30 s.

  • Typical scalloped pattern: slow just after reward, accelerating as interval ends.

  • Variable Interval (VI)

  • First response after a variable, unpredictable interval; mean value defined.

  • Produces slow, steady responding; common in natural foraging.

Differential Reinforcement (DR)

• Values of rewards scale with quality of performance.

• Example protocol:

  • Below-average response → no reward.

  • Average → low-value treat.

  • Above-average → medium treat.

  • Exceptional → high-value treat (e.g. steak cube).

• Enhances precision and encourages continual improvement without aversives.

Shaping, Capturing & Luring

• Shaping = reinforcing successive approximations until the full behaviour emerges.

  • More humane than force or flooding; leverages animal’s natural variability.

  • Tips for effectiveness:

  • Define the final objective behaviour precisely.

  • Break into small, achievable criteria; raise the criterion after 2–3 consecutive successes.

  • Keep rate of reinforcement high; avoid frustration.

  • If the animal stalls, back up to last successful step ("gradient of difficulty").

• Capturing = waiting for the animal to spontaneously perform the behaviour, then marking & reinforcing.

• Luring = using a visible reinforcer (or target) to guide the movement; fade lure promptly to prevent dependency.

• Chinning (mentioned) often refers to gently guiding via manual contact (commonly in farm species training).

• Extinction = withholding reinforcement so behaviour frequency diminishes; may provoke an extinction burst (temporary spike in intensity or variability).

Chaining

• A chain = ordered sequence where each behaviour’s cue is the completion of the previous behaviour.

• Two assembly strategies:

  • Forward Chaining

  • Teach behaviour A → reinforce.

  • Add behaviour B after A → reinforce, etc.

  • Natural for sequences that mirror real-world order.

  • Backward Chaining

  • Start with last behaviour (closest to primary reinforcer) — ensures every trial ends with success.

  • Add preceding links progressively (Z → reinforce; Y then Z → reinforce; X then Y then Z…).

  • Favoured in complex routines (e.g. guide-dog opening fridge → bring soda → sit politely).

• Practical advice:

  • Task analysis – break the routine into clear, trainable components.

  • Decide whether to reinforce only after the full chain or maintain occasional within-chain rewards.

• Widely utilised in advanced contexts: assistance-dog tasks, marine-mammal shows, equestrian liberty acts.

Discrimination & Generalisation

• Discrimination = responding differently to stimuli that vary along a critical dimension.

  • Example: dog sits on “sit” cue but not on “down”.

  • Rooted in adaptive ecology – recognising food vs. toxin, friend vs. foe.

• Generalisation = treating diverse stimuli with shared features as functionally similar.

  • Example: horse accepts various trailer types after trailer-loading training.

  • Promotes efficiency: no need to relearn every novel instance.

• Trainers shape the stimulus control spectrum via deliberate exposure, differential reinforcement, and context variation.

Factors Affecting Learning

• Motivational State

  • Hunger, thirst, social drive, pain relief, play motivation.

  • Under- or over-motivation leads to sub-optimal performance.

• Biological Predispositions (“preparedness”)

  • Species-specific tendencies facilitate or hinder certain associations (e.g. rats readily link taste with nausea but not with electric shock location).

• Environment

  • Distractors (noise, smells, movement) may compete for attention.

  • Anxiety / stress impairs cognitive processing; aim for $\text{Optimal Arousal Zone}$ (Yerkes–Dodson curve).

• Age

  • Example: foals handled early show reduced neophobia later.

  • Senescent dogs often exhibit declines in working memory and learning speed; adjust session length & complexity.

Connections to Classical Conditioning & Emotion

• Operant contingencies frequently overlap with classical (Pavlovian) conditioning.

  • A clicker becomes a secondary reinforcer through classical pairing with food.

• Emotional states (fear, relief, anticipation) function as motivating operations; they enhance or diminish reinforcing power.

• Integrating ethological context prevents contra-free loading (animals evolving to "work" for outcomes) conflicts.

Practical & Ethical Implications

• Favour least-intrusive, minimally aversive (LIMA) principle.

• Partial schedules & differential reinforcement reduce reliance on punishment while preserving precision.

• Recognise and respect agency: allow the animal choice & control wherever feasible.

Numerical / Statistical References

• Notation recap:

  • $\text{FR-}n$ = reinforce each $n^{\text{th}}$ response.

  • $\text{VR-}n$ = mean of $n$ responses, variance unspecified.

  • $\text{FI-}t$ = first response after $t$ seconds.

  • $\text{VI-}t$ = variable interval averaging $t$ seconds.

Study Hints

• Diagram each schedule’s cumulative-response curve to visualise differences.

• Practise constructing shaping plans: list at least 6–8 approximation steps.

• Observe real-life examples (e.g. slot machines = VR, checking phone notifications = VI).

Summary

• Reinforcement alters probability of operant behaviour; proper timing & contingency strength are critical.

• Shaping + thoughtfully chosen schedules = humane, efficient training.

• Always integrate motivation, sensory biases, and emotional wellbeing for successful learning and ethical practice.