Models of Memory

1
Models of Memory

• Psychology 3717

2
Introduction

• Really there are two types of models
• Models that look at specific phenomena
• ACT or TLC for example
• Models that look at general organization
• Atkinson and Shiffrin for example

3
But why model?

• Models organize data
• Models make explicit predictions
• Models can lead to application
• When you think about it, the first two are what
  science is all about, so models help us get
  beyond simple description

4
SAM

• Search of Associative memory
• Math model
• Basically looking at list learning results
• So study list
• Then recognition or recall

5
Assumptions

• Target items are viewed in relation to the memory
  representations of all other items learned
• This includes the words, the context etc
• To be learned items are associated with the
  context
• When words are presented they are rehearsed
• Words have a familiarity value
• Old v new decisions are based on familiarity vale
• Basically a signal detection approach

6
Memory strength

• Based on rehearsal during encoding
• Association between the stimulus and the
  representation of the word itself
• Association between the stimulus and the context
• Item retrieval based on prompt by the experimenter

7
Retrieval

• Retrieval depends on the joint contribution of
  the context, of all other items and the item
  itself
• So the strength is, basically, the sum of all the
  associative strengths in the list
• This explains why recognition is easier than
  recall

8
Explains a lot

• Longer presentations, better memory
• RI effects (context changes subtly)
• Serial position
• Encoding specificity
• Recognition failure of recall

9
Nice eh!

• Yeah, but.
• Makes a lot of assumptions
• Why arent all items recognized when one is?

10
Levels of Processing

• Craik and Lockhart
• Memory is NOT just this passive thing
• It is the result of encoding
• Perceptual analysis
• Pattern recognition
• Semantic elaboration

11
LOP is A OK

• Semantic processing produces better memory than
  perceptual processing
• Conceptually driven vs. Data driven
• Only true with explicit memory (though see
  Challis and Brodbeck, 1992)
• Deeper semantic processing, better memory
• Read generate effect

12
Levels

• Depth seems sort of vague though
• Hmm, when do you get better memory?
• Well when you have deeper processing
• How do you know you have deeper processing?
• Err umm cuz you have better memory
• Transfer appropriate processing may be a bit of a
  better concept

13
Memory systems approaches

• Tulving and episodic / semantic distinction
• One is explicit, one is implicit
• There is physiological evidence of a sort
• Tulving maintains that only humans have episodic
  memory
• For him, it involves consciousness because it
  must be self referential
• I dont agree really

14
Why would there be multiple systems?

• Sherry and Schacter, 1987
• When a problem shows up that cannot be solved
  with the present system a new one will be
  selected for
• So, our memory for facts could not deal with
  autobiographical stuff
• Or, birdsongs cannot be remembered with simple CS
  US associations

15
Conclusions

• Models are cool
• They can be tested because they will make
  explicit predictions
• They organize data
• They cannot have too many assumptions, the fewer
  the better