L19 - Abstract relational learning in infancy Flashcards
1
Q
What are the principles of constructivist infant cognition?
A
- Innate domain general information processing system that detects low-level featural information, such as colour, motion
- Higher-level units formed from relationships among these.
- Higher-level units formed from these units. Learning is hierarchical & constructive
- Infants tend to use highest level units to interpret their environment
- If system gets overloaded, revert to lower-level of processing while incorporating new information
2
Q
What is relational learning in infancy?
A
- From 3 months of age, infants, under the right conditions, can abstract same/different relations and generalize across objects
- They can get overloaded! - It is fragile
- For 3-9 months olds, making objects salient, and not part of abstraction learning, interferes with generalization
- Too many pairs for 3-month olds during learning also overwhelms
- If you match training of 3 month olds to other animals, 3 month olds come out ahead
3
Q
What did Ferry, Hespos and Gentner (2015) find about same/different relational understanding in infancy?
A
- 7-9 months
- E1 = One pair of objects during familiarization, which they don’t generalise
- Have no novelty preference - aren’t generalising to the new sets of objects - E2 - Four pairs of objects during familiarisation
- Before they started the experiment, they were given objects to play with, some of them were in the habituation and some that weren’t
- Familiar objects part of habituation
- Familiar objects not included in habituation
- Completly novel objects
- Looking time decreases over habituation (as expected),
- Show novelty preference at test to novel relation
- No discrimination between novel relation and familiar relation for the objects they played with in the waiting room but weren’t part of the habituation.
- When they have an object they are interested in, it overloads and you just look at the object
4
Q
What did Anderson, Change, Hespos & gentner (2018) find about relational learning in infancy?
A
- 3 month olds
- E1 - 6 pairs of objects during familiarization, they don’t generalise
- E2 - 2 pairs of objects, they do generalise
Same pattern as previous study - show relational discrimination when in habituation and novel but not objects they played with that weren’t part of the habituation
5
Q
How do we develop the perception of causality of events? (Leslie 1984)
A
We “see” causality given certain spatial and temporal relationships
- As common moving surfaces are inputs to core object knowledge, certain spatial-temporal motion relationships are input to core causal knowledge
- When there is spatial contact with no delay, it looks causal - if that isn’t there we don’t infer cause and effect
- Will 6 month-old infants see events as collections of spatial & temporal features, or will they use those features to see them as causal or non-causal?
- Direct launching = Without delay
- Launching without collision = Gap between the two
- Last one = spatial gap and delay
- Two conditions
- One habituations to the direct launching and at test gets delayed without collision
- One habituates to delated reaction and dishabituates to launched without collision
- G1 - Habituates to causal event, tested with non-causal event with both temporal delay and spatial gaps
- First trial of habituation, long looking times
- Last habituation trial, short looking times
- Test trial has long looking times - they discriminate between the causal and non-causal event
- G2 - Habituate to a non-causal event (with temporal delay) and then at test see a different non-causal event (spatial gap)
- First habituation trial, long looking times
- Last habituation trial, short looking times
- Test trial is not significantly longer than the last habituation
trial. They generalized their habituation across non- causal events. They do not appear to discriminate between them, despite their differing spatial and temporal features
- Across the two conditions, it is clear the infants are responding in terms of causal status.
- Group 1 dishabituates to anon-causal event after habituating to a causal one.
- Group 2 generalizes their habituation from one non-causal event to another.
- Leslie interprets results as evidence for innate causal module
6
Q
What was the response from constructivism to Leslie’s suggestion?
A
- Higher-level units formed lower level units. Learning is hierarchical & constructive
- Infants tend to use highest level units to interpret their environment
- If system gets overloaded, revert to lower-level of processing while incorporating new information
- Innate causal module: given certain temporal and spatial cues, infants perceive events as causal
- Constructivism: Causal perception emerges from lower-level features. Infants may be able to perceive causality under some conditions, but may become overloaded in others, and perceive events only in terms of spatial and temporal features.
7
Q
What did Oakes & Cohen (1990) find that supports the constructivist response to Leslies suggestion?
A
- Test trial data in log - Looking time for 10 month olds habituating to direct launching causal event
- Familiar causal event shows a low looking time
- Both novel non-causal events (temporal delay & spatial gap) show long looking times. They discriminate between causal and non-causal events
- Test trial data for 10 month olds habituating to one of the non-causal events
- Long looking times for novel causal event
- Equally short looking times for the familiar non-causal they habituated to and the novel non-causal. They generalized their habituation across spatial and temporal features because of the shared non-causal classification
- 6 month olds habituating to the causal events, unlike the 10 month olds do not show a novelty preference to the novel non-causal events at test
- Perhaps because the same complex objects are involved, and they can’t get passed a focus on them
- Familiar test trial is actually even longer than the two novel test trials
- For the 6 month-olds who habituated to a non-causal event, they looked no longer to the novel causal than the novel non causal, suggesting they are not treating the events in terms of their causal classification
8
Q
How do leslies and constructivism research compare?
A
- Leslie (1984) used the same simple objects on every trial 6 months old treated events as causal or non-causal
- Oakes & Cohen (1990) used more complex objects. Same complex objects on every trial 6 month olds responded based on spatial-temporal features 10 month olds treated events as causal or non-causal
- Cohen & Oakes (1993) used complex objects that varied on every trial
- Will they be overloaded and revert to spatial-temporal responding?
9
Q
What did Cohen & Oakes 1993 research suggest?
A
- Complex objects that varied on every trial. Same kind of launching events as Oakes & Cohen (1990)
- Habituate to temporal delay
- Shows equal novelty preference to causal and spatial gap
- Habituate to spatial gap
- Shows equal novelty preference to causal and temporal delay
10
Q
How does an innate modular account explain the progressions?
A
- Oakes & Cohen (1990)
- 6 month-olds perceive events with simple objects causally, but not events with complex objects 10 month olds perceive events with complex objects causally, when all event exemplars had the same few objects
- Cohen & Oakes (1993)
- 10 Month-olds fall back to the simpler mode of perceiving events as mere collections of spatial and temporal features when the complex objects vary across examples of events
11
Q
What are the principles of constructivist infant cognition?
A
- Infants tend to use highest level units to interpret their environment
- If system gets overloaded, revert to lower-level of processing while incorporating new information
- Feature correlations vs. just features
- Cause and effect vs. space/time
- Same/different vs. an object focus
