Neuronal mechanisms of WM

Question 1

neurons in PFC fire during WM - Fuster (1974)

Answer 1

Monkey neurophys studies suggested role for PFC in WM

Monkeys see piece of food in tray, shutter comes down and tray closed – when shutter opens, monkey has to remember where food located

Single neuron recordings from PFC showed elevated neuronal firing during delay period, e.g. when shutter down

Interp as showing that neurons in PFC hold representation of to-be-remembered stim (e.g. location of food)

see notes

Question 2

Goldman-Rakic (1987): standard model of WM

Answer 2

Sustained delay = Baddeley’s WM storage buffers

Sustained activation in PFC during delay period of WM task reflects neuronal WM ‘template’, temporary representation of to-be-remembered info

Question 3

ev from monkey neurophysiology - Funahashi et al. (1989)

Answer 3

Oculomotor delayed response task

see notes

Monkeys saw cue on left/right fixation and had to maintain eye gaze at centre for 3s and then make eye movement (saccade) in direction of cue – hold direction of cue in memory for 3s

Direction-specific firing of PFC neurons during delay period (between cue and response)
o Neuron fired strongly when cue pointed to upper left location

see notes

Showed direct neurophys correlate of WM template, temporary representation of spatial location indicated by cue

Question 4

supporting human neuropsych ev for a role of PFC in WM - Petrides and Milner (1982)

Answer 4

Self-ordered WM task given to patients w/ frontal and temp cortex lesions

Patients instructed to touch one picture per sheet paper and not touch same pic twice – 12 sheets and touched all 12 images

Required to remember from one sheet of paper to next which images touched previously

Patients w/ frontal lesions disproportionately impaired – deficit w/ WM

Again they interpreted this as evidence that the PFC holds a representation of to-be-remembered information over the short periods of time of a WM task.

see notes

Question 5

what v where in WM

Answer 5

Continuation of dorsal/ventral what/where distinction in visual processing into PFC

see notes

Question 6

ev from monkey neurophys for what/where dissociation in WM - Wilson et al. (1993)

Answer 6

see notes

Saw cue that instructed them to make eye movement in particular direction, then had to remember info before making response

Key change to task was that in addition to standard spatial cues (cue appearing in actual location to which monkeys had to make eye movement) also pattern cues, in which pattern appeared in centre of screen and pattern instructed monkeys to make eye movement in particular direction

Response in both trial types same but type of info to be remembered differed

In spatial trials, monkey had to remember spatial location of cue and in pattern trials, had to maintain identity of pattern

see notes

The graphs show the responses of two neurons during the delay period during each part of the trial.

B shows the response of a neuron in the inferior convexity – the lower, ventral part of the PFC

C shows the response of a neuron in the dorsolateral (upper, dorsal) part of the PFC.

As you can see, the neuron in the inferior ventral PFC shows higher activation to the patterns and lower activation to spatial cues

Whereas the neuron in the upper dorsal PFC shows higher activation to spatial cues and lower activation to patterns.

You can also see that each neuron preferred a particular pattern or a particular direction.

Elevated activity during pattern task – remembering info about what object was presented

Double dissociation between indv neurons in diff parts of PFC

Question 7

what v where in WM - supporting ev from human neuroimaging (PET) - Courtney et al. (1996)

Answer 7

Object WM task – remember identities of 3 faces
o Activation in ventral PFC – what task

Spatial WM task – remember the locations of 3 faces
o Activation in dorsal PFC – where task

see notes

possible confound - type of processing
o Subjects may use strats (e.g. chunking) to help perf of spatial task - requires manip of stored info in WM – may differ depending on task required
o Object WM tasks can be perf using simple maintenance of info (is this face familiar to me?) whereas spatial WM tasks are amenable to strats such as chunking
o E.g. in this task I can remember each of the locations and then I can mentally combine them into single image in which all of relevant locations highlighted
o Makes remembering locations easier and aids perf of task
o Also requires manip rather than simply maintenance of stored info
o Process of mental manip that drives activation in dorsal PFC

Question 8

neuropsych ev for a role of the PFC in WM - Petrides and Milner (1982)

Answer 8

What processes does this task require?
o Storage of previously touched item
o Suppression of previously touched item
o SA to novel item
o Planning/strat use – creating chunks in memory – which segments of the sheet have you already touched?
o Sustained attention – just had stroke 

Always trade-off in neuropsych

see notes

Question 9

ev against standard model - monkey neurophys - Rao et al. (1997)

Answer 9

Task required monkey to remember object and first make eye movement to correct object, then make eye movement to correct location

Over half of neurons showed both object and location selectivity
o Neurons in PFC adapt flexibly to represent task relevant info
o Most neurons have v. fixed responses

see notes

Question 10

ev against standard model - monkey neurophys - Lebedev et al. (2004)

Answer 10

see notes

Monkeys saw cue in location, and then cue moved

Cue either brightened in which case they made saccade to attended location, or it dimmed and they made a saccade to remembered location

Classified neurons as ‘attention’, ‘memory’/’hybrid’

Majority of neurons selective for attended location, not one held in memory

Suggests PFC neurons more imp for attentional selection than WM

Few showed combo of responses, consistent with flexible adaptation theory

see notes

Question 11

fMRI ev for a ventral/dorsal dissociation in PFC according to type of processing - D-Eposito et al. (1999)

Answer 11

see notes

Asked subjects to perf a WM task in which they had to either maintain info in WM by simply holding letter string in memory and then judge whether probe letter was part of memory set

Or in manip condition subjects had to rearrange letters into alphabetical order and judge whether probe letter was in specific location in letter string after rearranging

Activation in dorsal and ventral PFC during delay period of maintenance and manip trials

Activation higher in dorsal PFC during delay period of manip trials

Suggests PFC organised according to type of processing (dorsal = manip, ventral = maintenance) rather than type of stim maintained (spatial v nonspatial)

Question 12

.

Answer 12

Human and monkey studies converge to suggest ‘standard model’ incorrect – PFC not organised according to type of stim held in WM

PFC may be organised according to type of processing
o Dorsal: manip
o Ventral: maintenance

PFC may not even hold representations of stim held in WM
o Where is such info stored in brain?
o What is role of PFC in WM?

Question 13

multivoxel pattern analysis of fMRI data

Answer 13

see notes

Smooth data so groups of indv ‘voxels’ treated as clusters

Useful way of revealing where in brain process happening, but risks missing imp info that might be contained in indv voxel responses

Question 14

multivoxel pattern analysis

Answer 14

Takes adv of fine-grained patterns of activation in brain

Uses machine learning techniques to teach algorithm about pattern of neural activation associated with particular stim

Algorithm able to ‘decode’ what subject looking at simply by viewing pattern of brain activity

Can we decode what item subject holding in WM from pattern of activation in PFC?

Question 15

multivoxel pattern analysis - Linden et al. (2011)

Answer 15

Subjects perf task requiring them to hold several objects in WM

On each trial, required to decide whether single object part of memory set

4 categories of objects – faces, bodies, flowers, scenes – categories seem to produce reliable patterns of objects

Trained pattern classifier to learn patterns of activation for each category

Tested ability of pattern classifier to predict which category subject holding in WM on each trial

see notes

Question 16

brain regions in which the pattern classifier was able to correctly predict which category the subjects were holding in WM

Answer 16

see notes

Regions holding category related info exclusively in posterior part of brain

Visual processing regions, e.g. FFA and other regions of ventral visual stream – same regions activated when categories of object presented to subjects

Implication that same regions that enable us to process object when see it with eyes also involved in storing temp representations of objects in WM

Activation = success at pattern classification

Success appears at back of brain – not from PFC

Parallels with attention

What is role of PFC in WM, if not storage?

Question 17

can we decode info about items held in WM from patterns of activation in PF/parietal cortex? - Riggall and Postle (2012)

Answer 17

Scanned people performing WM task where had to memorise moving dot array, and during delay period cued to either remember direction/speed of dots

Then shown probe either match/mismatch

When cue appeared subjects had to follow rule (attend to speech/direction of dots in WM) and also hold in WM a specific stim (speed at which dots moving/direction)

see notes

Could decide which direction and how fast dots moving but only from visual cortex and temp cortex

PFC provided no info

Task instructions could be decoded from PFC and parietal regions

Consistent with Linden study – stim specific info stored in posterior, sensory specific cortex

Info about task rules stored in frontal and parietal cortex

Question 18

an alternative view to the standard model - WM as an emergent property - Postle (2006)

Answer 18

“If brain can represent it, brain can also demo WM for it … WM functions produced when attention directed to systems that have evolved to accomplish sensory-, representation-, or action-related functions. From this perspective, WM may be property that emerges from NS capable of representing many diff kinds info, and endowed w/ flexibly deployable attention”

o WM as attention to internal stim?

Question 19

fMRI ev for internal attention hyp of WM - Higo et al. (2011)

Answer 19

see notes

Ps maintained 2 objects in WM

Cued either to maintain both objects (non-SA condition) or single object (SA condition) in WM

Asked to decide whether any of objects in array matched object(s) holding in WM

see notes

Activation in PFC greater for selective condition than non

Connectivity analysis demo’d that activation in PFC modulated activation in diff occipitotemporal regions depending on which stim subject maintained in WM

Combined TMS/fMRI showed that disruption of PFC activity caused distal effects on occipitotemporal activation, thus establishing direction of causation

PFC seems to send attentional bias signal to sensory specific regions to enhance processing of task-relevant object during WM

Question 20

distributed neuronal architecture of WM

Answer 20

Lower level visual regions maintain temp representations of items held in WM (templates) and PFC/parietal regions hold representation of task rules for manip info (central exec)

see notes