Thursday 27 March 2014

What is Working Memory? How does it relate to general intelligence? Speculations...

*

NOTE ADDED 15 MAY - I HAVE COME TO THINK THE BELOW CONCEPT OF WORKING MEMORY IS WRONG, OR AT LEAST USELESS AND MISLEADING. INDEED I WONDER WHETHER WM MAY BE A FUNDAMENTALLY FLAWED CONCEPT?... I SHALL PROBABLY DELETE THIS POST SOMETIME, BUT FOR THE PRESENT SHALL LEAVE IT - WITH THIS WARNING.

Some time ago I was very interested by 'Working Memory' - trying to understand how to conceptualize it, what its structural brain basis might be - and also to measure it in this study:

http://www.hedweb.com/bgcharlton/tina-fry.html

The concept of Working Memory played a very large role in my book Psychiatry and the Human Condition - published in 2000.

http://www.hedweb.com/bgcharlton/psychhuman.html

Here is an excerpt from Psychiatry and the Human Condition describing how I visualized WM c 1999 - other discussions can be found by word-searching the phrase 'Working Memory'.

*

Working memory (WM) is the site of awareness, located in the prefrontal lobe of the cerebral cortex. WM functions as an integration zone of the brain, where representations from different systems converge and where several items of thought to which we are attending can simultaneously be sustained and manipulated. When we deliberately grapple with a problem and try to think it through - this process is happening in working memory, when we are aware of something it is in working memory, when we wish to attend to a specific stimulus, we represent it is WM...


Awareness comprises attention and working memory (WM). To be aware of an perception it must be selectively attended to, and the representation of that entity must be kept active and held in the brain for a length of time adequate to allow other cognitive representations to interact with it and in a place where other cognitive representations can be projected. Working memory is such a place, a place where information converges and is kept active for longer than usual periods. Hence working memory is the anatomical site of awareness.

The nature of working memory can be understood using concepts derived from cognitive neuroscience. Working memory is a three-dimensional space filled with neurons that can activate in patterns. Cognition is conceptualized as the processing of information in the form of topographically-organized (3-dimensional) patterns of neural activity called representations - because each specific pattern ‘represents’ a perceptual input. So that seeing a particular shape produces a pattern of cell activation on the retina, and this shape is reproduced, summarized, transformed, combined etc in patterns of cell activation in the visual system of the brain - and each pattern of brain cell activation in each visual region retains a formal relationship to the original retinal activation.

Representations are the units of thinking. In the visual system there may be representations of the colour, movement and shading of an object, each of these having been constructed from information extracted from the original pattern of cell activation in the retina (using many built-in and learned assumptions about the nature of the visual world). The propagation and combination of representations is the process of cognition.

Cognitive representations in most parts of the brain typically stay active and persist for a time scale of the order of several tens of milliseconds. But in working memory cognitive representations may be maintained over a much longer time scale - perhaps hundreds or thousands of milliseconds - and probably by the action of specialized ‘delay’ neurons which maintain firing over longer periods. So WM is a 3-D space which contains patterns of nerve firing that are sustained long enough that they can interact with other 'incoming' patterns. This sustaining of cognitive representations means that working memory is also a ‘convergence’ region which brings together and integrates highly processed data from several separate information streams.
Any animal that is able selectively to attend-to and sustain cognitive representations could be said to possess a WM and to be 'aware' - although the content of that awareness and the length of time it can be sustained may be simple and short. The capacity of WM will certainly vary between species, and the structures that perform the function of WM will vary substantially according to the design of the central nervous system. In other words working memory is a function which is performed by structures that have arisen by convergent evolution, WM is not homologous between all animals that possess it - presumably the large and effective WM of an octopus is performed by quite different brain structures from WM in a sheep dog, structure that have no common ancestor and evolved down a quite a different path. The mechanism and connectivity of the human WM allows cognitive representations from different perceptual modalities or from different attended parts of the environment to be kept active simultaneously, to interact, and to undergo integration in order that appropriate whole-organism behavioral responses may be produced.
Working memory is reciprocally-linked to long term memory (LTM), such that representations formed in WM can be stored in LTM as patterns of enhanced or impaired transmission between nerve cells (the mechanism by which this occurs is uncertain but probably involves a structure called the hippocampus). So temporary patterns of active nerves are converted to much more lasting patterns of easier or harder transmission between nerves. The patterns in LTM may be later recalled and re-evoked in WM for further cycles of processing and elaboration.
This is how complex thinking gets done - a certain ,maximum number of representations can interact in WM in the time available (maybe a couple of seconds). So there is a limit to what can be done in WM during the span of activation of its representations. To do more requires storing the intermediate steps in reasoning. The products of an interaction in WM can be summarized (‘chunked’) and ‘posted’ to LTM where they wait until they are need again. When recalled and reactivated these complex packaged representations from LTM can undergo further cycles of interaction and modification, each building up the complexity of representations and of conceptual thought.
WM is therefore conceptualized as a site for integration of attended perceptual information deriving from a range of sensory inputs. Awareness seems to be used to select and integrate relevant inputs from a complex environment to enable animals to choose between a large repertoire of behavioral responses. There is a selective pressure to evolve WM in any animal capable of complex behavioural responses to a complexly variable environment. So the cognitive representations in WM in non-conscious animals are derived from external sensory inputs (eg. vision, hearing, smell, taste and touch).
The critical point for this current argument is that non-conscious animals may be aware of their surroundings, but they lack the capacity to be aware of their own body states. Awareness of outer environment is common, but awareness of inner body states is unique to conscious animals.
**
Working Memory Revisited in light of IQ
But I now need to go back and revisit my old understanding of Working Memory in light of my more recent understanding of general intelligence - because when I wrote Psychiatry and the Human Condition I knew essentially nothing about IQ. That such a thing can happen has at least two causes - the first is my own obtuse ignorance, the second that when I did try to tackle intelligence I was put-off by the fact (and it is a fact) that nearly-all psychometricians (including many of the best and most famous) are non-biological and non-evolutionary in their basic mode of thinking.
And this is true even when psychologically-trained psychometricians were writing about biology and evolution - it was (and is) obvious that they fundamentally hadn't a clue! This is a matter of training, especially early training - and that fact that traditionally psychology was taught in isolation from biology, hence evolutionary theory - and from medicine - in a weird No Man's Land of proliferating ad hoc theories and slavish devotion to arbitrary methods and statistics.
*

Working Memory versus Intelligence 
My understanding is that Working Memory and Intelligence are conceptually different, serve somewhat different functions, and are dissociable - such that a person may have higher than average intelligence and lower than average WM - or vice versa.
Intelligence is, roughly, a measure of the speed of processing (which may be roughly equivalent to efficient connectivity) while WM is, roughly, the size of the active 'workspace' -  presumably constrained by the anatomical size of the effective working memory zone and the fact that the relevant nerve cells can only be activated for a timescale of a few seconds.
So, Working Memory might be visualized as the 3D size of the space in which processing occurs - that which is processed may be visualized as the interaction of complex 3D shapes which represent the content of thought (ideas, perceptions, emotions etc) - and intelligence is the speed with which all this happens.
High intelligence means that more interactions can occur within a given size (and duration) of Working Memory; while a larger WM means that for a given level of intelligence, more things can be thought-about simultaneously.
*
(Something to flag-up: In a computer analogy, intelligence might be equivalent to the speed of a microprocessor in terms of the efficiency and complexity of its circuitry; while working memory might be equivalent to a microprocessor's Cache Memory. So intelligence is how quickly the microprocessor can do operations; while WM represents the amount of information which can be included in active processing possible at a given time. But I will leave critique and development of that analogy to those who know more about computers than I do - which is nearly-everybody.)

*

So, the highest level of thinking - such as creative genius - would seem to need both a high intelligence and also a large capacity Working Memory; since the WM would allow a person to hold many things simultaneously in-mind, including emotional evaluations - while high intelligence would allow these things to interact complexly within the short time-frame of WM.
*

The combination of WM and Intelligence can be regarded as processing power, and it can be seen that various combinations can lead to the same overall power. 

For example, to use some simple numbers to give an idea of ratios - 

An IQ of 3 and a WM of 2 - compared with an IQ of 2 and a WM of 3:

The WM number represents the complexity of content, while the IQ number represents the number of iterations of processing. 

*

So, when the WM is half as much again (3 compared with 2) then that means more possible combinations between the items being processed; and when when the IQ is increased by half - then there are 150% more iterations of processing.  

A WM of 3 and IQ of 2 might be 2 iterations of 3 --

2 X 3 = 6 as a number representing power.


By contrast a lower WM of 2 and IQ of 3 has one and a half time the number of iterations - therefore three iterations instead of two, but with a lower WM number to represent less complex content --

2 X 2 X 2 = 8 as a number representing power

*

So, in percentage terms, IQ (general intelligence) seemingly has a greater influence on processing power, because it allows more iterations of processing.

However, the quality of thinking of a WM 3/ IQ 2 will be different from a WM 2/ IQ 3 - I would guess that relatively higher intelligence would lead to a more linear, narrow and logically-extrapolative style of thinking; while a relatively higher WM would I think have a more associative style - better at multifaceted judgment.

All speculative stuff! But let's see where it leads...

*