Lecture 2

• Categories are useful

  • help organize information into meaningful groups
  • = facilitate cognitive processes - perception, memory, decision-making
  • grouping similar items = efficiently process & understand complex information, make predictions, communicate effectively
• Exemplar theory 


  • when encounter new stimulus = compare to all specific examples or instances of category we have previously encountered
  • emphasizes importance of storing & remembering individual instances of a category
• Prototype theory 


  • when encounter a new stimulus = mentally compare it to abstract representation or "prototype" we have stored for category
  • prototype based on an average of all instances of category we have encountered in past
  • emphasizes role of generalized representations
  • = simplified & efficient way of categorizing new stimuli
• Category learning in pigeons (Herrnstein, Loveland, & Cable, 1976)

  • thousands of pictures used in training
  • successful discrimination of novel pictures at test
  • learned categories of water & people - even specific person
  • also tropical fish
• Styles of paintings - Watanabe, Sakamoto, Wakita 1995

  • pigeons' ability to discriminate paintings
  • = able to distinguish between artists' works = demonstrating level of visual discrimination previously thought to be exclusive to humans
• Discrimination among multiple categories

  (Bhatt, Wasserman, Reynolds, & Knauss, 1988)

  • 4 categories learned
  • Novel pictures classified successfully
  • Performance maintained when training pictures never repeated
• Larger sets make for better categories (Bhatt, 1988)

  larger sets = poorer learning for training items but better performance on test items
• Categorisation is capuchins is successful, but…

  (D’Amato & van Sant, 1988)

  • Successful discrimination of people from non-people

  some errors were unusual

  • Photographs classified as people when looked nothing like people
  • In training set, only people slides contained red
  • Non-people test slides containing red misclassified
• Categorisation determined by features (Huber & Lenz 1993)

  • 4 stimulus dimensions used = each dimension = 3 possible values
  • Pigeons trained to peck at faces with sum greater than 0 - behavior reinforced
  • rate of pecking response determined by sum of features in stimuli
  • aimed to assess how pigeons categorized & responded to stimuli based on sum of features
• Aust & Huber 2006

  • Perhaps non-human animals don’t know what pictures stand for - just respond based on low-level features, but…
  • pigeons trained to recognize pictures of objects through visual categorization task
  • = suggest pigeons possess representational insight = able to generalize learning to novel stimuli
  • highlights cognitive abilities of pigeons in visual recognition tasks - perceptual & categorization capabilities.
• Aust & Huber 2010 - experience

  • investigates pigeons' recognition of human body parts in pictures
  • trained to discriminate handless or headless humans & nonhumans
  • Some had prior exposure to real human heads - others didn't
  • without prior experience of heads failed to show preference to pictures of missing parts over arbiturary skin patches
  • Suggests evidence of representational insight in pigeons
  • Highlights influence of real-world experience on visual categorization tasks
• Concrete categories

  • groups of objects defined by physical attributes, properties, characteristics that can be directly perceived through senses
  • all pictures of people have common concrete attributes
  • are identifiable physical features present
  • simple associative learning can explain acquisition of concrete categories
• Abstract categories

  • groups of concepts defined by shared characteristics or properties not directly observable through senses
  • based on subjective interpretation
  • 'son' - based on relationship
  • evidence that non-humans can learn relational categories = mixed
• categorising perceptual features

  • human infants tend to categorize objects based on perceptual features rather than abstract categories = more conceptual in nature
  • evidence suggests non-human animals also rely on perceptual features for categorization
  • perhaps non-human animal struggle to learn realtional categories?
• Same/ different learning in pigeons

  • rewarded for pecking at flanker matching sample image
  • group 2 - 2 sample pictures - fast learning
  • only 4 combinations for g2 = may have remembered correct response for each configuration
  • novel pictures = no evidence of same/ different rule
  • group 152-samples = 18 months to learn - generalised to novel pictures
• Matching to sample

  • Pigeons find task very difficult
  • other species able to master more easily
• Irene Pepperberg's parrot Alex

  easily respond verbally to questions about what was same/ different
• Why do pigeons find same/ different judgments so difficult?

  • pigeons can perceive sameness, but it lacks salience? - doesn't stand out?
  • Pigeons might prefer to use individual stimuli - rather than make abstract judgments
  • Training with lots of instances should help
• Effects of varying length of training

  • size of training set strongly influenced transfer to novel items - more training higher transfer correct
  • same effect seen of rhesus monkeys & capuchins - but performance generally better
• Abstract rules - opposites

  • abstract rules - dictate how make decisions
  • how fill in gaps of pattern - responses to new based on similarity or opposites rule
  • Humans: rule-based - decisions cos of abstract rules or principles
  • Rats & pigeons: similarity-based - of stimuli
  • Humans under cognitive load: similarity-based
• Second-order relationships in chimpanzees 


  • young chimpanzees, matching problem - pairs of objects attached to a board - select pair of objects whose relationship matched relationship of sample
  • Despite extensive training - over 1000 trials = chimpanzees failed to demonstrate ability to perceive second-order relationships.
  • Children under 5 also struggle with similar tasks -suggests complexity of problem
• Testing second-order relationships in chimpanzees - Further research

  indicated chimpanzees could distinguish between relationships when presented with different pairs of objects after playing with one pair
• What does it mean?

  • lots of evidence that non-humans animals can form categories based on physical features
  • can be stored using concrete code
  • moderate amount of evidence of learning about relationships - mostly from matching experiments
• sameness might have a concrete component 


  • may not be entirely abstract or purely conceptual
  • idea of similarity might be influenced by concrete elements or characteristics.