L20 - Relational learning beyond infancy

Question 1

What is the natural partitions hypothesis? Gentner (1982)

Answer 1

• We parse the world into objects and relations among the objects
• Objects are perceptually cohesive: Relations are perhaps harder, dynamic/unstable, indefinite number of options
• Hypothesis: children will form categories of objects before they form categories of relations between objects universally
• This means early vocabularies will be filled with simple object nouns, like, dog & chair, before prepositions of, verbs throw, and even relational nouns like guest
  
  • Category is about the role the object is playing rather than intrinsic qualities
• In some languages, early predominance of nouns over verbs is found in English, Italian, Japanese, Korean, Mandarin

Question 2

How does analogical comparison show how relations are learned?

Answer 2

• e.g.
  
  • Rutherford model of atom vs atom is like the solar system
  • Not about intrinsic qualities comparison - objects are different but analogies are the same
    
    • Sun is more massive than the planets which causes the planets to revolve around the sun.
    • Nucleus is more massive than electrons which causes the electrons to revolve around the nucleus

Question 3

How do mundane comparisons highlight relations?

Answer 3

• Markman & Gentner (1993)
  
  • Ask what goes with the woman from the top picture in the bottom picture
    
    • Most people say the other woman
    • But if compare the two scenes first, pick the squirrel (both receiving food)
    • Even simple comparison of simple scenes, shifts focus to relational structure.
      
      • Similar to more “sophisticated” analogical reasoning

Question 3

What is the relational shift?

Answer 3

• Relational shift in “the career of similarity”
  • In any given context, might start out representing domain but as you gain experience you shift focus towards common relationships when common relationships exist
  • Consequences for word learning
    • Relational words are learned later (Gentner & Boroditsky)
    • Relational words are often assumed to be featural at first
      • e.g Uncle = beardy guy → Uncle = parents brother

Question 4

What did Christie & Gentner, 2010, find about comparison and the relational shift?

Answer 4

• Triad word extension task for a novel relational word
  
  • 3 year olds
  • 4 year olds
  • Three conditions: 1 solo example vs. 2 examples presented in sequence vs. Comparison of 2 examples
• Findings
  
  • Go from almost no relational matches in solo condition to an massive increase
  • At 3 years, not really representing relationship but once you add two and compare them, they can confirm relationship
• When add sequential condition (limited as children don’t have as strong WM capacity)
  
  • Showed similar pattern to infant relational learning.
    
    • Doesn’t matter the age when learning an entirely novel relation → comparison helps

Question 5

How does language foster relational thinking? How can this impact homesigners?

Answer 5

• Giving two things a common label is a powerful invitation to compare them, abstract their commonalities
  
  • Understand their relational structure
• Reify relational systems
• Recall that relations are perceptually ill defined. Systematic relational labelling is critically for having stable relational representations.o Homesigners (not exposed to sign language) without spatial language have poor spatial skills.o Builds on the seedlings of our non-linguistic relational abilities.
• Study (Gentner et al., 2013)
  
  • Looking at community of Turkish homesigners
    
    • Homesign does not have systematic preposition use
    • Spatial relation words only really comes up out interaction with others
    • Task: Spatial relationship task
      
      • Bookshelves with cards on them, on the back one of the cards is a prize. Prize on same place in other set of shelves. Need to communicate where the prize is to one another
    • When object competitor, over 50% correct
    • Without a system of spatial relation in words, an easy spatial relationship task is not achievable when there is an object competitor (in deaf children)
    • Language amplifies pre-existing capacities

Question 6

What is abstract causal reasoning?

Answer 6

• Learning novel spatial relations seem to need a lot of support.
• On the other hand..
  
  • Alison Gopnik (and colleagues) show abstract causal learning during a similar age range
  • That is, when children are attempting to figure out what causes what else in the world

Question 7

What study supports abstract causal reasoning?

Answer 7

• Lucas et al. (2014)
  
  • 4 year-olds and adults learned about 2 different “blicket” machines
  • The first machine would either work with one specific block, or only with a specific combination of blocks
  • Not just any 1 or 2, but specific 1 or 2
  • Then the 2nd machine (with a new set of blocks) was presented such that it would be ambiguous between working from 1 specific block, or from 2 specific blocks
  • Results
    
    • Adults assumed 1 block for 2nd machine regardless of how the first machine worked
    • 4 year-olds interpreted the 2nd in line with the first
    • Adults were biased to single-object explanations due to prior experience
    • So, kids at least, seem ready to interpret the relations between a new sets of object in terms of just learned causal relations
    • Kids are more open minded as they haven’t had the experience of how other things work

Question 8

What did Carey propose about the development of number?

Answer 8

• Carey posits a core knowledge system for number
• However, core knowledge is often qualitatively different than mature conceptual knowledge
• Language plays a key role making this transition
• Two core knowledge systems of number1: Precise number of small sets
  • can distinguish 2 dots from 3 dots
  2: Analog magnitude scale
  • can distinguish large quantities from others, vaguely 350 dots from 500 dots, not 350 dots from 352 dots
  • To learn maths, clearly something needs to change
• 2 year olds can often count to 10 and beyond but this doesn’t mean they know what those words mean
• About 4 year olds they understand that each word has a role in number system
• Systematic structure in language can be the basis of an analogy for systematic conceptual structure
• Evidence that Piraha tribe in Brazil have no numerical language and never leave core knowledge state of numerical cognition
  
  • See work by Peter Gordon & Mike Frank
  • Sometimes language may support abstract concept learning, but people may still learn these concepts without language. Other abstract concepts seem perhaps unlearnable without language

-