Memory—what s in there The problem You have

Memory--what’s in there?

The problem You have a great deal of stuff in memory: How do you get the right thing out when you need it?

How do you find the right book in a big storehouse?

An obvious option is to number all the books. This is called an addressing system. Computer hard drives also use addressing systems.

Your mind clearly does not work this way. 1. More books means slower search 2. Slight errors mean possibly big mistakes It doesn’t take us longer to remember things as we learn more. When we make a mistake it’s typically a near miss.

Even more impressive. . . If you dropped a raw egg from the Eiffel tower, would it break?

How do you know that? You probably have not observed an egg being dropped from the Eiffel tower, so what happens is not in memory. How, then, do you answer the question? Your memory system provides relevant facts: Eiffel tower is tall Eggs dropped from a great height will break.

Theories of memory designed to account for. . . • Ready access to lots of information • Access to relevant information when crucial information is not stored

Network theories Node Input from environment Activation

Collins & Quillian Hierarchical Model Links Nodes 1. Relevant info becomes available (in some versions) 2. Accounts for property inheritance (“does a canary breathe? ”) 3. Accounts for identity questions (“is a canary an animal? ”)

Problems • Some of the predictions aren’t borne out (“is a gorilla a mammal? ” > “is a gorilla an animal? ”) • Properties don’t seem to be stored only once (e. g. , “flies” is a property that people strongly associate with robins, bluebirds, wrens, etc. )

Collins & Loftus Unlabeled link with weight Input from environment Congress Bush White House Node Congress “Bush” activation Activation spreads Congress Bush White House Time Spreading activation along unlabeled nodes.

Example

Same information • • Units State of activation. Output function. Pattern of connectivity=knowledge • A activation rule: integrate input • Learning rule.

Properties of networks • Allows retrieval of properties. • Allows contentaddressable storage (e. g. activate 40’s and Jet) • Accounts for typicality • Creates defaults • Resistant to faulty input

Distributed representation Local = one node, one concept Distributed = concepts distributed across multiple nodes

Architecture These models are frequently called Parallel Distributed Processing (or PDP) models

Advantages PDP models have all the advantages of local representation models, plus: 1) Graceful degradation 2) Also seeks to show memories are acquired, not just how they are organized after learning. 3) Automatically finds prototypes and exceptions to prototypes 4) Automatically generalizes

Is that everything in memory? Until 20 years ago, the answer would have been “yup. ” Today, most memory researchers believe that there a number of memory systems, at least five of them. We’ll start with how that idea got started. . .

Patient HM • Retrieval from LTM intact • STM intact • Ability to get material from STM to LTM (i. e. ability to learn) devastated

Pursuit rotor, mirror tracing HM is shown to learn a new motor skills.

Motor skill demo

More amnesics tested, more tasks learned

Conclusion from this work: It’s not just that there is memory and there are motor skills: perceptual stuff can be preserved. People start to think of amnesia as memory minus some process; this missing process impairs most memory, but certain tasks don’t depend on the missing process, so the tasks are learned.

What is the missing process? • Poor encoding • Fast forgetting • Poor retrieval

People give up Researchers gave up on the idea that there is a single memory system, which is missing some process, leading to the pattern of spared and impaired memory functions. Instead they concluded that there are multiple memory systems: each in a different part of the brain, learning different things.

Memory systems • Declarative memory--recall, recog.

Memory systems • Declarative memory--recall, recog. • Repetition Priming--gollin figures (like the elephant)

Other priming tasks: Stem completion priming List reading Window Reason Animal Toaster. . . Delay Filler task Stem completion Rea___ Acc___ Win___

Other priming tasks: Lexical decision priming Set display time Flash words on screen, (masked so iconic memory is foiled) at fastest time that yields 50% correct List reading Window Reason Animal Toaster. . . Delay Filler task Lexical decision test Reason Rollarp Justice

Memory systems • Declarative memory--recall, recog. • Repetition Priming--gollin figures (like the elephant) and others. • Motor skill learning--e. g. pursuit rotor

Memory systems • Declarative memory--recall, recog. • Repetition Priming--gollin figures (like the elephant) • Motor skill learning--e. g. pursuit rotor • Skeletal conditioning--e. g. , eyeblink conditioning

Classical conditioning--reminder If Unconditioned Stimulus--> Unconditioned Response (meat powder) (salivation) then pair Conditioned stimulus with the unconditioned stimulus (bell) (meat powder) then eventually Conditioned stimulus--> conditioned response (bell) (~salivation)

Skeletal conditioning=an overt response of motor system If Unconditioned Stimulus--> Unconditioned Response (air puff to cornea) (blinking) then pair Conditioned stimulus with the unconditioned stimulus (tone) (air puff to eye) then eventually Conditioned stimulus--> conditioned response (bell) (~blinking)

Emotional conditioning = the cs is an emotion If Unconditioned Stimulus--> Unconditioned Response (boat horn) (anxiety) then pair Conditioned stimulus with the unconditioned stimulus (picture of flower) (boat horn) then eventually Conditioned stimulus--> conditioned response (picture of flower) (~anxiety)

Memory systems • • • Declarative memory--recall, recog. Repetition Priming--gollin figures (like the elephant) Motor skill learning--e. g. pursuit rotor Skeletal conditioning--e. g. , eyeblink Emotional conditioning--emotional resp. to previously neutral stimulus VERY important source of evidence for separability is the brain basis of these types of learning

Declarative memory Repetition Priming Motor skill learning Skeletal conditioning Emotional conditioning hippocampus

Declarative memory Repetition Priming Motor skill learning Skeletal conditioning Emotional conditioning Visual cortex

Declarative memory Repetition Priming Motor skill learning Skeletal conditioning Emotional conditioning Basal ganglia

Declarative memory Repetition Priming Motor skill learning Skeletal conditioning Emotional conditioning Cerebellum

Declarative memory Repetition Priming Motor skill learning Skeletal conditioning Emotional conditioning Amygdala

Multiple memory systems This research has expanded our notion of memory. There appear to be multiple cognitive systems in the brain that have some plasticity.