Neuroanatomy of Memory - Structure and Function

Memory in networks 1 & 2

(square brackets [ ] indicate references which may be found at the end).

Networks 1 & 2 process sensorimotor and spatial memory. [1] These memories are designed to enable basic tasks, like controlling vital functions, locomotion, facial expression, place, person and object recognition, balance, hand-eye coordination and so on.

Memory in N1 and N2 is largely unconscious. Associations develop as a result of concrete experience as congruous concepts based on our hard-wired core concepts, and they don’t take up conscious awareness simply because they don’t need to. Sensorimotor/spatial skills like walking, climbing, eating, dancing and swimming can all be learned entirely without conscious thought, and so can territory-mapping (knowing what sort of thing happens where in familiar territory.)

This unconscious memory is sometimes called “muscle memory” or “skill memory”, because it 'feels' as though our bodies or hands just 'know how to do it', and it’s because of different individual experiences in making this memory that we all have individual ways of moving, walking, talking and thinking. Body language, the way you smile, shake hands, bow or wave, your posture, the sound of your voice, are all dependent on habits –the habits of your sensorimotor and spatial memory.

These and many other aspects of our behavior are expressed so automatically they may seem unchangeable, or believed to be innate. But we should not jump to such conclusions, which forget the roles of plasticity and epigenetics, and the quality and frequency of input that was necessary for setting up those habits in the first place and maintaining them. Unconscious memory and unconscious processing can both be brought into conscious awareness and improved, and we'll look at how that happens later.

sensorimotor memory

Each core concept relates to different kinds of input, and each network recognizes its 'own' kind of input. Network 1 deals with input that associates with its core concept; matter.

Definitions: By 'sensorimotor memory', we mean how you remember and recognize what stuff is because of its material properties; for example because of what it feels like, tastes like, what things are hot or cold, sharp or blunt, hard or soft, smooth or hairy, sweet or sour. Their common abstracted property is that they are material and physical; they concern immediately experienced properties of matter. What we mean by 'concrete' is the 'hardware' of reality as opposed to abstract software (such as math or language systems or thought processes, which are still real but have no material presence).

This sort of memory goes a long way towards enabling us to identify what many things are, but by no means all the way. Remember, each network stores only parts of memories, and each uploads its bits when N3 requests them via association and re-members them (in permanent storage memories are literally dismembered -their concepts remain separate until they are associated again). Thus we can forget one or more aspects of a memory if one or more association links fail; this is one reason why memory is not totally reilable and also why individual aspects of memories can be wiped without losing the rest.

Memories are stored in N1 as physical response patterns of its cells to all the textures, tastes and tactile experiences we have ever encountered in our lives.

Spatial memory

Definitions: By 'spatial memory', we mean the memories of contexts, places, spaces, simple behaviors (movements) and backgrounds, and how to recognize and perform basic behaviors. The way you find your way around your house or local neighborhood relies on spatial memory, and so does knowing what happens where.

Network 2 recognizes and memorizes contextual spatial concepts. Their common abstracted concrete property is space. Behavior and events are motion in space, and places and objects are things in space. Together these form the context of all our experience.

In the database here are selections of cellular movement patterns, this time not associated with material objects but associated with our proximal responses in different contexts, places and behaviors. Network 2 memorizes audio as well as visuals, allowing imagination to compare cellular response patterns of incoming sights and sounds against previously experienced responses in the memory database.

This is NOT the same as procedural memory (although a lot of studies will label it thus). The difference is, procedural skills and memory require co-ordination between abstract mental and concrete fine motor control. For example, pressing the keys of a piano is a sensorimotor/spatial skill, but playing the piano with a band or an orchestra, or writing a song and working it out on piano -that's procedural and it needs N4. Procedures are timed complex behaviors that apply rules to bring order out of chaos. Operating a stone age hand axe requires N2, but building a hut, operating a complex machine or performing in synchrony with others requires N4.

Memory in N1&2

Structure and function

cerebellum

The part of the brain responsible for most long term sensorimotor and spatial memory in N1&2 is the cerebellum. The cerebellum is in the bottom back part of the brain and looks a bit like a lump of cauliflower. It contributes to sensorimotor and spatial memory in our control of movement in sensorimotor tasks that use the body as a tool, such as walking, throwing and catching, cleaning our teeth, riding a bike, dancing, swimming, driving a car, and all the many basic movements that we must make every day.

The cerebellum is also involved in posture and balance (both of which are modified by experience and contribute to spatial memory), body movements, balance, depth perception and finding our way around in the dark. (In darkness, hippocampal place cells fire at a lower rate, so N3 has to rely more on the cerebellar neurons involved in self-motion cues to remember where it's been and work out where it's going).[7]

Damage to this area prevents easy learning of sensorimotor and spatial skills and through associated research it has more recently been linked to a role in automating the unconscious process used when learning such skills [10].

Studies have shown reduced connections from the cerebellum to the primary motor area with practice as skills become automatic, it is presumed because of a decreased need for error correction from the cerebellum. The cerebellum passes the memories it processes to the basal ganglia in N3, which is currently the most likely candidate for long term storage of automatic unconscious memory.

Occipital lobe

The occipital cortex stores long term sensorimotor & spatial memory. The occipital lobe is located in the rearmost part of the skull directly above the cerebellum.This lobe is shared by several networks, as the main function of the occipital lobe is that of vision.

Retinal sensors send signals through the optic tract to the primary visual cortex, where it is organized and sent down one of two possible pathways; the ventral (N1) stream is responsible for object representation and recognition and is also commonly known as the "what" stream. The dorsal (N2) stream is responsible for guiding our movements in behaviors and recognizing where objects are in space, commonly known as the "where" or "how" stream [12].

Two other parts of N 1 & 2 are important for the formation of memory. They are the brain stem and the hypothalamus. They are important because they modulate our hormonal state, affecting the weighting of memories by sending feedback to the amygdala in N3. The brain stem, which flanks the cerebellum all the way up into the centre of the brain, produces most of the neurotransmitters we need for many processes, including dopamine, serotonin and norepinephrine. We looked at the hypothalamus in Tutorial 2, so you'll remember that it's the bridge (user interface) between the brain and the body for hormonal control and chemical feedback.

Memory in N3

N3 stores long term eidetic memory. By 'eidetic' memory we mean memories of whole events or episodes in graphic format.

In networks 1 & 2 if a cell is firing, the pattern of its internal behavior will be recorded. That's memory on a cellular level. Memory is automatic (unconscious) in rear networks so basic processes can do all the work. By ‘basic’ we mean that cells can respond automatically to relevant signals to carry out the cellular processes triggered by those signals with very little intermediary software; processes like recording their own responses in memory, comparing their responses to others in the database, picking out and responding to emergency patterns, tagging frequent patterns for permanent memory, and forwarding data to network 3.

Together, the components of N3 form one of the two main processing hubs in the brain. Network 3 provides the bridge between unconscious and conscious memory, essentially because it provides a bridge (user interface) between mind and brain.

This is where code is given meaning; where graphic core concepts held in memory are used to associate and translate information between mechanical physical motion and abstract internal thoughts and ideas.

Core concepts are universal. Recent research (Oct, 2011) has shown that different individuals' brains use the same, common neural code to recognize complex visual images.

The study demonstrates that objects are similarly represented across different brains, allowing for reliable classification of one person's brain activity based on another's. [72]

All individuals use a common code for visual recognition, making it possible to identify specific patterns of brain activity for a wide range of visual images that are the same in all brains.

As a result of their research, the team showed that a pattern of brain activity in one individual can be decoded by finding the picture or movie that evoked the same pattern in other individuals.[72]

N3 is also where memory is weighted according to importance and tagged for eventual location, should it become long term. N3 holds permament memories of odors, pheromones, emotional weightings, and associations of modes of input that are indicative of dangers or benefits.

Q: How many modes of input do you think are indicative of danger? Try to work this out for yourself.

Clues:

...Don't automatically jump into thinking of a long list of 'dangerous things' to humans; they are only the details. Go back to basics.

...Think about the common properties of all things dangerous to biology, and abstract those common properties.

...ALL things dangerous to biology either cause some kind of sensory overload or some kind of sensory deprivation...so...

...Think about how many causes of problems for intelligence there are. We don't mean the symptoms of problems; we mean the underlying causes.

...If you're thinking, 'Anxiety!' or 'Incongruity!', you are getting there; they certainly do pull us out of the green zone, but think about what causes both of these. Keep thinking backwards to the source of the problem...

...Think about what it always comes down to; in the 'when things go wrong' section...

...so, how many modes of input are indicative of danger?...

A: The answer is two. (If you don't understand the answer, RTFQ. How do you think you got distracted from remembering the question?)

As far as N3 is concerned, there are only two modes of input that are indicative of danger: not enough input or wrong input.

Basically any event that takes us out of the 'green zone' for growth & development does so because of one or both of these factors; not enough of the right input, or too much of the wrong input. Once you understand this, looking after your mind becomes a lot easier. You only ever have two things to watch out for, and their warning signals are boredom and confusion.

If input density falls below or rises above certain values, outside the healthy growth/development zone, protection mode is engaged because the unconscious mind imagines emergency responses may be necessary. That's all N3 'knows'; that it MUST enable emergency responses via engaging protection mode whenever there is wrong input or lack of input.

Hyperdensity of input (wrong input) indicates sensory overload, and hypodensity of input (no input) indicates sensory deprivation. Biology knows that either is deadly unless addressed. So all associations with such input modes are stored locally for expediency of response.

All memories stored in N3 are kept in eidetic (graphic imagery) format, and all other memories are translated into this format when passing through N3 and being associated with meaning.

Remember your tiger? Black and orange stripes are automatically unconsciously associated with a danger signal. When your ancestors were walking through the jungle, they did not pick up black and orange striped things. They had learned that to do so could result in a fatal sensory overload called being eaten, so those who got 'that alarmed feeling' when they saw signa of one were much more likely to survive.

Consequently, here for their descendents in the 21st century, N3 still has a graphic of black and orange stripes with a big “NO!!!” associated with it, closely followed by the associated concept “run your ass off”; and the weighting signal to flood your body with hormones that can coordinate brain and body for extreme responses in less than a couple of seconds.

The graphic concept of black and orange stripes is an archetype; that is to say, am image that has an effect on the unconscious mind and a simultaneous complementary effect on the physical body, long before its recognized by the conscious mind. We'll be studying archetypes more closely in tutorial 10.

structure and function

the parietal lobe

Sits on top of the occipital lobe, at the rear of the brain.

The parietal lobe assists with verbal and visual short term memory and damage to the supramarginal gyrus causes short term memory loss [17].

The parietal lobe and medial temporal lobes may well be the repository for RAM -a bundle of random memories that we've either used recently (and so might need again soon) and new memories that we haven't had time to fully process yet, all thrown together in temporary storage. We can hack new memories much more easily while they're still in RAM (before they are defragged and permanently stored). Activity in the posterior parietal cortex is tightly correlated with the limited amount of scene information that can be stored in visual short-term memory [18].

The lateral intraparietal area is also involved in visual categorisation (a process previously though to be exclusive to N6). [79]

medial temporal lobes

Functional MRI (fMRI) can identify distinct activations within the MTL. The medial temporal lobes house the outside arms of the hippo, which are essential for memory function - particularly the transference from short to long term memory and control of sensorimotor/spatial memory and behavior. Damage to this area typically results in anterograde amnesia.

There is evidence that older memories do not rely on the medial temporal lobes, whereas more recent long-term memories do. In consolidation, interactions between the medial temporal lobes/hippo and various lateral cortical regions are thought to store permament memories outside the medial temporal lobes by slowly forming direct links between the neuronal representations of coincidental input [9]

Both temporal lobes process aspects of audio and visual input. Long term auditory memory has been found to be stored in the primary auditory cortex (no surprise there then), contralateral to the ear of distal stimulus presentation. [14]

Adjacent areas in the superior, posterior and lateral parts of the temporal lobes (TL) are involved in higher-level auditory processing. In humans this includes speech. The logical meanings and grammar of speech are processed in the left TL and the metaphoric meanings and sound of words are processed in the right TL. This is why, with left temporal stroke damage, patients who cannot speak can still remember how to sing and swear.

A wealth of mirror neurons resides in this area [80] and it is clear the MTLs play a large part in image generation, empathy and modeling.

thalamus

The thalamus is the central relay station for information processing, the 'information highway' at the very center of N3. Subjects with thalamic damage demonstrate deficits on measures of attention, psychomotor speed and unconscious visuomotor sequence learning. [2]

Anterograde amnesia (loss of abiity to make new memories) can result from damage to the thalamus and the surrounding networks; new information is processed normally but never gets encoded and stored, since the connections between hippocampus and cortex (through the thalamus) are disrupted.

Two pioneers [3] in research on the thalamus examined the close two-way relationships between thalamus and cerebral cortex and looked at the distinctive functions of the links between the thalamus and the rest of the brain. Countering the dominant approach —which does not recognize that all neocortical areas receive important inputs from the thalamus and send outputs to lower motor centers—they argue for a reappraisal of the way we think about the cortex and its interactions with the rest of the brain.

We can only heartily agree. Researchers looking at the functions of Nitric Oxide discovered how the thalamus functions to 'boot up' the brain and provide for central processing and control of all impulses going to and from the cortex. They describe its function as an operating system, but from the description it actually seems closer to the functions of a kernel and an essential part of the brain's CPU. [4]

basal ganglia ('BGs')

We'll tell you two things about the basal ganglia: One, they're not basal, and two, they're not ganglia.

Confused? So were we! But it's true; the BGs are not at the base of the brain, they are mostly medial/frontal in position, and they're really not ganglia; they're nuclei.

Perhaps you are beginning to understand why neuroscience students get confused! Unfortunately, the term 'basal ganglia' is also another one interpreted in the mainstream in various ways. Some researchers include virtually the whole of N3 in the basal ganglia, some include parts of N3, N6 and N2. It looks like way back somewhen, someone found a selection of little lumpy brain areas that nobody really understood (at that time) very well, and lumped them all together and called that area 'the basal ganglia', which is probably pretty much what it looked like down an ordinary microscope. Some call the same areas the telencephalon.

The BGs are in fact tiny nuclei, most of which cluster around the back end of the forebrain, with several protuberances wrapping around the hippo/amy/thalamus:

The main components of the umbrella term “basal ganglia” are the striatum (also called neostriatum) composed of caudate and putamen, globus pallidus and (oddly) substantia nigra and the subthalamic nucleus. [13]

The BGs are responsible for so many tasks in processing and memory that we won't be covering them here but in advanced tutorials. For now, just get used to where they are and remember that all of them send input to N3.

The amygdala’s role in memory

The amy's main role in memory is modulating by weighting the memories that are being constructed in the hippo & thalamus.

Studies spanning several decades have shown us the amy’s role in the emotional ‘weighting’ of memory. It gets feedback about what’s going on both directly from the body and via the hypothalamus (that sits very close to it in the brain) and it also responds to neurotransmitters in the rest of the brain, so at any given time the amy can sample overall neurochemistry and associate it with input. If overall chemistry is anxiety, all input seems more dangerous than it really is.

Its role in memory involves deciding what is relevant to your wellbeing and what doesn't matter. It decides how much events matter by looking at both input density and feedback from whatever sort of hormonal and neurochemical responses they cause.

The amy puts an ‘importance weighting’ on events that are beneficial or harmful,[29] by attaching a record of the pattern of body chemistry associated with the event (which we experience as emotion) via feedback coming in from the hypothalamus and other areas [16].

The resolution of a whole memory depends on its bit density, and each known concept is one bit of meaning. High-density memories contain more information, including that of the associated emotions. The process of weighting strengthens the memories being made during an event by increasing their 'bit density' (or resolution, if you like). Strongly weighted memories are more easily retrieved, last longer, and their details are clearer and more vivid.

You'll remember we have two amygdalae. Current research is showing that the one on the left (usually, but not always) deals with detecting of, and weighting memories for, beneficial events, and the one on the right (usually) looks for and weights deleterious events. (There is some variety in this in the same way that some people are right- or left-handed, and it’s useful to remember that when you start exploring NMS and TMS.)

The amy is not the only part of N3 that is capable of altering weighting density, but it is the main area for weighting danger/benefit associations and it is the amy that often gets hacked in treating PTSD, as certain drugs can remove the weighting 'tag' whilst leaving other aspects of a memory intact [8].

The hippocampus’ role in memory

The hippo is really the star of the memory show. Interacting with other parts of N3, it enables all long term memory. It makes and maintains the inner model (cognitive maps) that all input and memories are compared against for processing, it performs all main memory processes including the making of new memories in learning, encoding (turning short term memories into long term ones), and consolidation (moving them from short term storage to their permanent destination), and it facilitates recall & reconsolidation.[63] Overall, what it does best (with the help of the thalamus) is association. Wow. That’s a hell of a job description.

The hippocampus’ right side is more oriented towards responding to N2's spatial aspects, whereas the left side is associated with N1's concrete material information. Also, there is evidence that experience in building extensive mental maps in the inner model can increase the volume of one’s hippocampus [15].

It's not possible to form new eidetic (graphic episodic) memories without the hippo.

Glial cells; the cells that support neurons, also modulate their activity in learning and memory.[55]

Memory in N4 & N5

Procedural memory (sometimes confused with spatial memory or implicit memory)

By procedural memory we mean memories of abstract complex behaviors as procedures, and memories of skills involving 'feedforward' synthesis; such as programming a computer, writing a poem, composing a symphony or putting together enough data and working out the maths to find the proof for a theory. Procedural memory is knowing how as opposed to knowing what. It remembers complex abstract procedural concepts, such as, composing, engineering, cooperating, constructing, and understanding things like hyperbolic geometry.

Declarative memory (sometimes called semantic memory or explicit memory)

By declarative memory we mean the ability to store and recall information that can be 'declared' (spoken or written). To remember how to perform an experiment we need procedural memory, but to explain that experiment to someone else or to make a written record, we need declarative memory. Declarative memories are memories of abstract facts about things, and memories of skills involving feedback analysis; such as logic, quantity surveying, hacking complex security systems, assessment, administration, resource management or reductionistic experimentation.

Structure & function

The dorsolateral striatum is associated with the acquisition of procedural habits and is the main neuronal cell nucleus linked to procedural memory.

Two parallel information processing pathways diverge from the striatum, both acting in opposition to each other in the control of movement, and they allow for association with other needed functional structures [19]. One pathway is direct while the other is indirect and both pathways work together to allow for a functional neural feedback loop.

Many looping circuits connect back at the striatum from other areas of the brain; including those from the medial temporal cortex and the ventral striatum (N3).

Permanent storage of procedural memories occurs in the right frontal cortex (N4). The main looping circuit involved in the motor skill part of procedural memory is usually called the cortex-basal ganglia-thalamus-cortex loop [20].

Damage to the right side (N4) cortex causes memory problems with identifying geometric shapes, patterns, figures and faces, general perception and problem solving skills [22].

Permanent storage of declarative memories occurs in the left ventrolateral cortex (VLPFC)(N5). Many studies indicate that broad or abstract concepts, such as nouns like "mammal", are represented towards the front (rostral) regions of the cortex. In contrast, more specific or tangible concepts, such as nouns like "rat", are represented caudally (towards the rear) in N1. [21]

Damage to the left side (N5) cortex can lead to language memory discrepancies, i.e. difficulty in properly recalling letters, numbers and words, and causes general language problems [22]. Recent TMS studies that have targeted disruption of a VLPFC further indicate that this region is necessary for intact semantic processing of stimuli [23]

Memory in N6

The prefrontal cortex serves our ability to remember what's going on in the here and now; in order to coordinate behavior, plan, strategise, predict and interact.

N6 helps us select out memories that are most relevant to each given occasion. It can coordinate various types of information into a coherent whole. For example, the knowledge of where information came from must be put together with the information itself, into a single memory representation (this is called source monitoring.) In poor memory such information can become separated, such as when we recall something but cannot remember where or when it took place; (this is referred to as a source monitoring error) [24]. The frontal lobes are also involved in the ability to remember what we need to do in the future and when we need to do it; this is sometimes called prospective memory and it uses both working memory and the RAM in N3 [25].

Working memory

structure & function

Research identifies the frontal cortex and anterior cingulate gyrus (N6) are essential for working memory, but these work in conjunction with the parietal cortex and the basal ganglia (N3).

There is evidence from brain-imaging studies that prefrontal cortex shows sustained activity during the delay period of visual working memory tasks, indicating that this cortex maintains online representations of stimuli after they are removed from view. There is also evidence for domain specificity within frontal cortex based on the type of information, with object working memory mediated by more ventral (towards the front) frontal regions and spatial working memory mediated by more dorsal (towards the back) frontal regions [26].

There are three separate prefrontal regions and processes associated with working memory in humans.

(1) a `phonological loop' for maintaining auditory & verbal information; (2) a `visuospatial sketch pad' for maintaining information about the visual properties of objects and about spatial locations; and (3) a `central executive' for attentional control and for coordinating the manipulation and use of information held in both the phonological loop and the sketch pad.

Thus, the concept of working memory encompasses both the informational content of our consciousness (knowledge as ability and knowledge as information) and the wilful use and manipulation of that information.

Three regions have been identified in studies as being associated with monitoring and/or manipulation; Brodmann areas 6/8 in the Premotor cortex and Supplementary Motor Cortex (includes Frontal eye fields); the Anterior Cingulate, and Brodmann areas 9/46 in the Dorsolateral Prefrontal Cortex. The left PFC (adjacent to N5) dominates for analysis-based object representations, and the right (adjacent to N4) for synthesis-based (image-based) object representations. [27]

The regions activated in object working-memory tasks seem similar to the regions of prefrontal cortex activated in long-term memory tasks, since working memory can also temporarily hold items from LT memory this is not surprising.