the ‘consensus’ view of memory

+a ‘self’ module

+

the constructive memory framework (CMF)= The best functional Account of memory there is…

The End (of lecture 7…)

Episodic Memory Mechanisms

ConsolidationMechanisms

AttentionalControl

Encoding Storage Retrieval

AttentionalControl

SemanticRecords

PerceptualRecords

Binding

ContextSemanticRecords

PerceptualRecords

Binding

Context

CMF adheres to the consensus view

ENCODING

Forming multiple individual records of attended information (see also Moscovitch’s model)

Associating (binding) each individual record with the current ‘spatiotemporal context’

Keeping each set of bound records distinct from all the others already in memory (pattern separation)

CONSOLIDATION

Abstraction of semantic gist

Formation of multiple retrieval pathways

‘Offline’ playback mechanisms during sleep and quiet states

CMF adheres to the consensus view

CMF adheres to the consensus view

RETRIEVAL

Retrieval ‘focus’

Access to the records of attended information via a retrieval cue (by hippocampal pattern completion)

Inhibition of irrelevant information

Re-activation of episodic content (held in the neocortex)

Monitoring/evaluating retrieval products (prefrontally mediated)

CMF Neuroanatomy• The hippocampal formation

‘Indexing’ of episodes: exactly how is unknown Necessary both for encoding and retrieval Damage leads to dense retrograde and anterograde amnesia

• The frontal lobes Strategic control over memory: exactly how is again unknown! Damage leads to confabulations, delusions, heightened false

memory, source amnesia

• The entire ‘association’ neocortex Representation of experienced content Damage should lead to loss of specific content of prior episodes

View of Long-Term Memory

ENCODING RETRIEVAL

Seeing Word

Hearing Word

MTL

MTL

MTL

MTLBut can we get beyond this cartoon state of knowledge, for example adding in the

CMF?

the End (of lecture 7…)

How does episodic memory actually work in vivo, and in real-time?

Maps of the MindMemory and Cognition Lecture 8

Electrophysiological

Haemodynamic

Cognitive Neuroscience Methods…

of seeing inside the box of tricks…

Psychophysiology

• Aim is to develop mind reading technologies

• We are most interested in the PPY of Perception and Cognition. In other words, Cognitive Neuroscience

• Can we tell what a person is thinking or experiencing just by looking at their brain activity?

Phrenology Was Odd…

• There is no known mechanism that would sculpt the contours of the skull according to underlying brain shape – i.e. there is no correlation between local contours of the

skull and the underlying size or shape of the brain

• Their psychological ‘model’ was based on common sense constructs of personality

– I.e. Looking in the wrong place for the wrong thing!

But not entirely wrong…

• The idea of functional localisation has survived, but in a different form

– Localisation does not respect character traits, like honesty, peevishness

– Localisation may respect, for example, sensory modality, ‘cognitive systems’ (e.g. LTM), along with other psychological mechanisms yet to be elucidated

Acceptable ‘modern’ principles of functional neuroanatomy

• Functional Segregation Discrete cognitive functions are localised to specific

parts/circuits of the brain (complex tasks are ‘divided and conquered’)

• Functional Integration Coordinated interactions between functionally specialised

areas (e.g. during retrieval from episodic memory, reading, perceptual binding etc)

Where We At?

• We want to read a person’s mind from the activity of their brain

• Their mind is composed of lots of interacting cognitive processes

• Each distinct process is carried out by networks of brain regions, each region is probably performing specific functions, but they all work together

• So we need a device or a technique that can detect changes in brain activity specific to any cognitive process

So What Do We Need?

• In an experiment we (think we) engage different functions in different conditions. For every condition we

– Detect rapid changes in neuronal activity (requires a temporal resolution of milliseconds, 1/100ths of a second)

– Locate activity within brain structures that are engaged (may require an anatomical (spatial) resolution of millimeters or better)

• Currently no such technique exists. Instead we rely on converging data from many techniques

Electrophysiological Techniques EEG

non-invasive recordings from an array of scalp electrodes

Averaging EEG produces ERPs

• Portions of the EEG time-locked to an event are averaged together, extracting the neural signature for the ‘event’.

10uV+

-

TIME (sec)0 21

DOG

AIR

SHOE

AVERAGE

What do ERP waveforms tell us?

CONDITION A

CONDITION B

0 1 2

TIME (seconds)

5uV+

-

ONSET OF EVENT

INFORMATION ABOUT THE NEURAL BASIS OF PROCESSING IS PROVIDED BY THE DIFFERENCE IN ACTIVITY

Functional Inferences Based Upon Electrophysiology

Timing Upper limit on time it takes for neural

processing to differ Time course of a process (onset,

duration, offset)

Level at which a process is engaged

Engagement of multiple processes at different times or in different conditions

Early Topography

Late Topography

Another Electrophysiological Technique

Electrophysiological Techniques

Principle advantages non-invasive high temporal resolution direct reflection of neuronal activity easy to produce event-related potentials by

selective averaging of EEG epochs. topographic mapping Cheap (for EEG but not MEG)

The Brain’s Plumbing

Haemodynamic Techniques

Oxygen and glucose are supplied by the blood as ‘fuel’ (energy) for the brain The brain does not store fuel, so Blood supply changes as needs arise Changes are regionally specific - following the local dynamics of neuronal activity

within a region

Haemodynamic techniques localise brain activity by detecting these regional changes in cerebral blood supply

Positron Emission Tomography (PET)

Samples the entire brain volume homogeneously

Has an effective anatomical resolution of about 10mm or so in group studies

An ‘indirect’ measure of neuronal activityDue to radiation dose, only a limited number of

scans can be taken from each subject

Magnetic Resonance Imaging (MRI)

Put head into a strong magnetic field Water protons align themselves with respect to the field alignment is then perturbed by radio-frequency pulses non-invasive and fast (few seconds) protons ‘relax’ back into alignment, giving off a signal relaxation signals can reveal

tissue type physiological state (e.g. blood oxygenation) 3D position in the magnetic field

Our starting point …• Electrophysiological and Haemodynamic techniques

Have different temporal and spatial resolutions Measure different physiological signals Constrain experimental design and functional inferences in

different ways May provide complementary information when functional

maps from each technique can be formally co-registered

ERP PET

Consensus View of Long-Term Memory

ENCODING RETRIEVAL

Seeing Word

Hearing Word

MTL

MTL

MTL

MTLBut can we get beyond this cartoon state of knowledge, for example adding in the

CMF?

Focussed Search

Retrieve / Inhibit Monitor

Stimuli

0Time - 1 3 4 65 7 92 8

Retrieval success!!Retrieval failure

cueonset

Ecphory/inhibition

MonitoringRetrieval Perception/attention

Patterncompletion/

Binding

‘selective attention’

Stimuli

Time 0.1 0.2 0.40.3 0.5 0.70 0.6

CMF{retrieval}

Stimuli

Time 0.1 0.2 0.40.3 0.5 0.70 0.6

Ecphory?

Monitoring?

Implicit Memory?

Familiarity?