Lecture 7: Models of attention and memory

Question 1

Children can respond to joint attention through subtle cues in an adult’s body and inflection. What are some non-verbal cues for joint attention?

Answer 1

Prosody
Gaze
Gesture

Question 2

What is motherese? What is another name for it?

Answer 2

Motherese or Infant Directed Speech (IDS) is the way people, usually parents, talk to children.

Question 3

What are some characteristics of motherese?

Answer 3

• More repetitive than normal speech
• Simplified grammar
• Exaggerated intonation

Question 4

What are the two types of memory the ACT* model distinguishes between? Wha type of knowledge do they contain? What do they involve?

Answer 4

Procedural memory
○ Implicit knowledge
○ Involves skills

Declarative memory
○ Explicit knowledge
○ Involves facts

Question 5

What does the 1968 Atkinson & Shriffin “multi-store model of memory” consists in? What is the problem with this model?

Answer 5

Pathway: You are exposed to stimuli from the environment and you apply your attention mechanism to parts of this sensory input. This is stored in your short-term store, which per definition has a limited capacity. In order to transfer the information from short to long-term store, rehearsal is needed.

Rehearsal does not always lead to learning (try to remember university logo that you have been exposed to many times). In addition, learning can happen unintentionally, it doesn’t necessarily need rehearsal.

Question 6

What’s the difference between the ACT* and the ACT-R model?

Answer 6

ACT-R is the upgraded and more detailed version of ACT. It was not only meant to describe memory (like the ACT model did), but also other aspects of human-like decision-making, abstraction and generalization.

Question 7

What does the ‘U-shape learning’ for children learning the English past tense look like? Which model uses this part of language acquisition?

Answer 7

1. They hear adults using the past tense
2. They copy it directly, without understanding; low error rate
3. They notice patterns, and overgeneralise (everything -ed); high error rate
4. The wrong past tense verbs get corrected by adults, the correct ones don’t
5. They now know the exceptions to the rules; low error rate again

ACT-R implements the ‘U-shaped leaning’.

Question 8

What does the connectionist model consist of?

Answer 8

An input layer, hidden layers, an output layer with multiple nodes.
Each node is a simple computation.
There are weighted connections between the nodes. Those weights can change based on feedback, and this is how learning happens in this model

Question 9

What is known as “Metamemory”?

Answer 9

We know that we used to know something but can’t retrieve the memory at the moment