Lecture 7 Flashcards
1
Q
What is categorisation?
A
- When a group of objects gets the same response
- Concepts are when a group of objects form a class the members of which are reacted to similarly
- Category/concept are overlapping terms e.g concept of chair was thought to be dependent on language and therefore uniquely human
- Distinction between concrete categories; where objects are classified on the basis of their physical characteristics AND abstract categories: objects are classified on basis of some property other than physical attributes
2
Q
What was a study looking at categorisation was not necessarily human?
A
- Pigeons and categorisation
- TV screen on a pecking channel: looked at trees vs not trees
- Lots of unique photos = some with trees and others that are not trees but looked like trees e.g celery
- Rewarded when pecked screen that was tree = they succeed with pictures of oakleaves vs other leaves, other people, works of art, words/nonwords, classical music, phonemes etc.
- Categorisation has turned up at every level of the animal kingdom where it has been competently sought
3
Q
What are the explanations for categorisation?
A
- Rote learning: learn lots of examples = based on memory of rewarded stimuli BUT cannot explain how novel instances are recognised
- Feature learning: trees have branches, leaves and trunk: most things in category have common features = can generalise to novel situations that have said features
4
Q
What is the evidence for feature learning?
A
- When do animals get it wrong
- Pictures with humans vs without humans in it e.g oval thing with pink in it, can be human and watermelon = can put something novel into the category that doesn’t really belong
- Pigeons taught discrimination between male and female faces, then started messing with features e.g pixelate and blur and remove colour to look where pigeons go wrong to find out what features they learnt = colour removal = incorrect discrimination
- They learn about a restrictive amount of features that required colour = doing feature learning
5
Q
What was another study looking at feature learning? (No config but elemental)
A
- Trained pigeons to discriminate between pictures of humans and pictures without
- Then tested novel situations
- Then scrambling the picture by pixelation and rejumbling
- Breaks up configuration but leaves features intact
- Pigeons can still discriminate when pretty jumbled (not all the way)
- Elemental theories of associative learning
6
Q
What might be another explanation for categorisation?
A
- We can remember a chair = if we learn based on feature, we should not be good at remembering certain chairs (examples) = if we formally test this = exemplar = even better at identifying particular examples, even though we recognise novel situations
- We learn categories by learning examples
- Exemplar theory: configural theory: every example and consequences is stored in memory
- Problems include correct classification of novel instances = in configural theory, we can respond to novel instances through similarity = what is learnt is examples, and we can generalise based on similarity
- Problem 2: Scrambling is not catastrophically disruptive for configural theory = does not align with the configuration = it should affect performance in config theory but only to the degree it interrupts similarity
- Some scrambling is catastrophically disruptive
7
Q
What was the experiment looking at disruptive scrambling?
A
- Looking at cartoon people and cartoon animals
- Moved features around so configuration changes to look like something else e.g pigeon had legs moved up to where arms would be on human etc. = all elements present but moved around = new configuration
- Moving creates a config that is too different that what was learnt about = features are not enough
8
Q
What are the conclusions of feature/exemplar theories?
A
- Thought to be language based but it isnt
- Evidence consistent with both, but both things happen = just depends of circumstance = lots of stimuli = featural mechanisms, limited = exemplar
- Both matter but depends on what and how it is being learnt
- Based on Rescorla Wagner AND configural theory
9
Q
What are abstract models?
A
- Concrete categorisation must go beyond physical features - it’s not as seen above
- e.g concrete thought is an actual thought of someone from bhutan but abstract is that idea
- Relationships are one way to test if animals can understand abstract thought e.g motherhood cannot be something seen = beyond physical
10
Q
What was an experiment about animals and relationships?
A
- Relationship of same vs different: matching a picture to a sample = rewarded to responding to the thing that is the same to the sample
- Sameness is not a feature of either object, it is not physical = sameness is in the relationship between them = not a property of either stimuli
- Called a same-different task = 4 ways to set up exp if you use two stimuli = examples they learnt about = now adapt to novel stimuli = animals are disrupted = initially thought animals cannot learn about abstract
11
Q
What did Katz and Wright do?
A
- Pigeons do not have training of same/different like humans = do not have the concept so they trained pigeons via trials before testing with novel stimuli
- When pigeons tested on stimuli trained on: doesn’t matter on how many stimuli 8-1024 = perform well
- When tested on novel stimuli: 8 training = 50% correct, as this increases to 1024 = 90% correct
- With monkeys = same general pattern of results but it occurs a lot faster = fewer examples of novel stimuli needed
- Exp performed on a lot of species: monkeys, crows, dolphin and budgerigars are successful, but poor transfer with new stimuli and pigeons = suggests split in animal kingdom of which can perform abstract thought
12
Q
What is the alternative explanation?
A
- Normal way of matching sample task confounds question of same-different with familiar-unfamiliar = thing that is most familiar is rewarded = physical cue is recency = two cues being measured
- Other relationships were measured e.g inside/outside = pigeons looked at stimuli with dots either inside/outside a weird shape = pigeons did eventually get above chance = but so many examples seen = generalising
- Same height vs different height bars = pigeons learnt it but after every combo was presenting
13
Q
What was the study of Alex the Parrot?
A
- Trained to speak e.g names of colour, shapes, questions, has vocab of 200 words
- Presented alex with a blue plastic cup and blue paper square and ask what is different and same = above chance
- Can also answer bigger/smaller = relationships = abstract thoughts
- Question of language training or stupid pigeons
- Repeated the bar height study on the parrot (wrong species) and it was worse than the pigeon
14
Q
What are second order relationships?
A
- Looked at same/different but no familiarity e.g sample is two blue bars, correct matched response is two red bars etc = animals have to pick relationship
- Tested on Chimp = trained on symbolic language = Sarah can perform second-order task
- Chimps who were not language trained could not do this = language training gives animals abilities they did not have before
- Animals were then given discrimination learning to make the selection slightly easier and then the chimps could differentiate = with training can perform relationship task without language
- Monkeys could learn discrimination BUT couldn’t do the same/different matching task = couldn’t generalise = suggests some species can do it if given training
- Reinvestigated with ducklings using imprinting = exposed to things that were the same/different and asked is mum same/different and they could do it
15
Q
What are other things to animals to determine abstract thoughts?
A
- Analogical Reasoning: cat to kitten as cow is to?
- Sarah, language trained chimp can do this with training
- Syntax: words do not information in them, the information from word order describe the relationship
- Dolphins taught symbolic language that has syntax = dolphins can understand what they are told and will use syntax to disambiguate instructions = beyond concrete thought
