Location of memory in the brain (The operation of memory (a single neuron can learn))

We guess the brain is connected with behavior. When we are studying brain functions, we analyze behaviors that are connected to memory, whether recent, past or innate. This involves all cases, even if we examine brain activity at rest. We cannot avoid consideration of the memory problem, even if we take an interest in any other function. It has become evident that many nervous centers are involved in the phenomenon of memory, no single and specific memory center exists, and various parts of the brain participate in the representation of any event [1255, 845]. Various centers perceive signals, create the aware image, record new and call to mind past information and generate output reactions, by which the investigator may judge that a memory system functions properly. Learning-induced alterations in neuronal electrical activity usually coincided with formation of material traces (c-Fos immunoreactivity, transcription, etc.) in the same area. The long-term changes in neuronal firing produced by instrumental learning occur in the same areas where there were activated during acquisition of the same task. The neurons participating in learning are not without fail those neurons which store the memory. Although neuronal correlates of learning are observed in countless brain neurons, recollection of information is a much more local process. Only a small number of neurons, when compared with those initially active in learning, form and retain task-specific firing patterns in well-trained animals [1203]. This corresponds to the discoveries of I.P. Pavlov [947] of phase generalization and specialization of conditional reflexes. Concordance in time of the activity in a local brain area with the memory process does not tell us a lot. On the other hand, if stimulation of a brain area causes recollection of images [958], this does not mean that memory is kept right at this place. Stimulation may excite another brain area. This is, evidently, a very complex problem, but there is one characteristic of memory, that allows one to hope that the puzzle will be solved: memory retains material traces and they can be found. Essentially, memory traces are kept in the healthy brain, in opposition to traces of trauma, insult, infection, etc.

Where does brain store its memory and how are memories are stored within brain? We do not know how images are kept in memory. Specific images may be kept in local places and then there is a one-to-one correspondence between memory elements and spatial coordinates in the brain. In such a case an image could recall when this particular place is activated. As we will demonstrate, this is not the best idea for memory mechanism, since memory is not a local process. Therefore, before discussion of "how" memory is stored in the brain, let us consider "where" the brain stores its memory. At first glance, the question "where" is much more simple than the question "how". Yet it turns out to be impossible to find memory without finding out its mechanism. Memory activates motivational and emotional centers and another question one may ask is: where do these aspects of memory lie in the brain? Also, can perception and recollection of images out of memory be indifferent? Habituation testifies against indifference. If something is not recorded, it can not be remembered. Emotional load of an image promotes its storage.

There is a distinction to be made between safety of conscious and unconscious memories. If, as was shown, reproduction of a previously elaborated system of reflexes in rats worsens with an increase in neocortex mass that has been ablated after 180-200 pairings of CS+ and US (a 3-linked chain of conditional reflexes with an interval between CS+ and US of a few seconds), the reflex system does not depend at all on the mass of the cortex removed after 350-400 pairings (phase of automation). Neocortex does not participate in automatic behavior [1250]. As we already have mentioned, reproduction of more complex behavior is more sensitive to deterioration. "Complex behavior" in this context means a participation of awareness in the control of behavior. Automatic behavior may also be complex (at the stage of information acquirement), but during the automation phase it is launched as a whole performance. Aware memory is, probably, more vulnerable, especially aware memory recall. It is sensitive to impairment to many places of brain but it has a tendency to recover, and recovery time depends on the severity of damage. In reality, patients with amnesia can show selective impairment in conscious memory in comparison with preserved skill learning. Enhanced sensitivity of a complex or aware behavior to damage is, evidently, a general rule. So, genetic loss in mice of the protein subunit responsible for fast inactivation of K+ channels results in an alteration of the excitability and damage of complex learning, while simple learning is not altered [468]. Complex behaviors need, to a large extent, to exploit a diverse past experience within awareness.

Difference between localization of earlier and recent memories has been also described. Using functional magnetic resonance imaging in healthy older adults, it was shown that medial temporal lobe lesions typically produce retrograde amnesia characterized by the disproportional loss of recently acquired memories; the hippocampus participates in processes of memory lasting a few years, while entorinal cortex is associated with memory extending up to 20 years [511].

The precise role of specific brain structures in storing/retrieving remote memories currently remains unclear [657]. If centers of memory exist, the search for them is complicated because of the existence of the powerful apparatus of self-recovery. It has become apparent that no single memory center exists, and many parts of the nervous system participate in the representation of any single event. Extirpations of brain tissue or other influences to neural centers have not revealed the center of memory and give rise to the hypothesis that the ability to memorize is an inalienable property of any neuron. K. Lashley formulated the law of mass action, according to which the extent of a memory defect in rats may be correlated with the size of the cortical area removed, but not with its specific location [705]. That is to say the cortex (and may be the brain) operates as a whole. Later this result was confirmed in rats [1250], in chickens [1040] and even in the mollusk Aplysia. The gill withdrawal reflex of Aplysia is generally depicted as a simple behavior mediated by a simple neural circuit in a simple organism. However, the reflexive withdrawal of the gill and other mantle organs is anything but simple. Even here, memory may not be localized to specific loci, but rather may be an emergent property of physiological mechanisms distributed throughout the entire circuitry [769, 278].

High neural centers, such as the cortex and hippocampus, play a colossal role in mammalian behavior. They have a screen-like structure and represent the world as a point-to-point correspondence, for instance, body surface, vision field, etc. This representation is continuous and closest points of the world are represented in adjacent points of brain. Of course, this is not ideal correspondence, since a closed object in three-dimensional space (as, for instance, our own body) is impossible to represent at the open two-dimensional surface of the cortex or hippocampus. Nevertheless, this does not prevent the creation of an inner model of the outer world and additionally to correlate events with a determined point in time. After cortex removal, an animal is a severe invalid. Observing such an animal, one may detect a deficit of almost any behavioral function with the exception of the inborn behaviors and the simplest forms of learning. These invalids behave much more primitively, than a simple vertebrate, fishes or reptilians, which do not have a cortex at all. Interestingly, learning after brain injury slows down, but the vegetative component of learning almost does not differ from the norm [103]. Therefore, how may we agree that the temporal lobe or hippocampus is the centers of learning or memory? Evidently, higher neural centers in mammalians have some high functions and the absence of these functions prevents the realization of primitive functions. One such function is, evidently, the possibility to integrate past events into a whole, intact and ordered representation of the world that relates to aware perception. Can awareness be present, even in a primitive form, in an isolated part of the neural system or in one neural cell itself? We will return to this problem later. Although a memory center is absent, maybe particular elements of memory, as images, are stored locally (grandmother neurons)? This suggestion looks doubtful, too. Usually, brain injury impedes the most complex forms of learning, memory recall, and weak, recent and unessential memory. Behavior is recovered during the weeks or months after injury, so evidently, memory existed and exists in the remaining regions of the brain. Long-term memory, evidently, never disappears, only its reproduction is worsened. It is difficult to imagine that a memory initially is recorded in one place and later is over-recorded in other places, or, say, is scattered through brain tissue.

If a function is once recovered after a temporary down-turn of some brain area (for instance, cortical representation of the paw), repeated down-turn of the same or any other brain area does not hinder the same function. Therefore, even localized functions have the potential to become non-localized. As for memory, even if a special form of memory is localized in a certain part of the brain, it can be delocalized after deletion of this zone. This means that even before removal, the memory traces were spread within the brain but only a limited volume was actually responsible for this trace. Lesions of different brain areas show specificity, but shortly after injury non-specific impairments of memory also surface. Damage of various brain areas induces temporary impairment in long-term memory. This suggests absence of the localization of specific memory element in specific brain zones [1316]. It is not without possibility that temporary impairment of behavior after brain damage arises not because of breakage of the neuronal structure, but as a result of influence of the wound itself. Brain may feel its wound affecting its chemical background. Recovery from mild head injury is displayed typically 1-3 month post-injury. Temporary boundaries of available memory after traumatic amnesia are gradually narrowed. Memory is partially affected during injury of neural system. Wonderful, these harms finally become too little to be a problem.

Recordings of neuronal activity during learning have shown that a large amount of brain neurons change their behavior. The frontal cortex, hip-pocampal system, basal ganglia, hypothalamus and other areas, each of which has neurons whose activity undergoes systematic evolution during learning [1006, 662, 1348]. Even during simple classical conditioning of the gill-withdrawal reflex in the simple nervous system of Aplysia hundreds, if not thousands, of neurons are active during learning [476]. Nevertheless, various neurons do not promote learning in the same degree. When, for example, monkeys were trained to learn new visuomotor rotations for only one target in space and neuronal activity was recorded in the primary motor cortex before, during and after learning, specific elevations of the firing rate of neu-ronal activity were only observed in a subpopulation of cells with directional properties corresponding to the locally learned rotation; that is, cells only showed plasticity if their preferred direction was near the training, while non-responding cells did not participate [951].

Learning may be artificially directed to a restricted population of neurons. Capability for learning is displayed in restricted parts of the neural system, when they are artificially divided from the rest of brain: in the peripheral neural system (without head) [376, 345, 1342, 1351], in a brain slice [734, 949], in a tissue culture [670, 1121, 371] and in an isolated neuron [1157, 348, 1419, 1301]. Neuronal analogs of learning are observed after transference of CS+, US, or instrumental reaction in the neuron’s vicinity [616, 710]. These primitive forms of behavior are not concerned with awareness, or at least we do not have evidence that any degree of awareness is present, say, in isolated brain slices (an example of tissue culture). We also do not know if diverse past experience concerning different habits may be combined into whole representation within isolated parts of brain, but this question is related to the problem of awareness.

When we are speaking of memory that is spread within the brain, this means that each image is spread. A specific image is never harmed; memory damage concern to more general categories: impairment related, say, to recent or unessential events, and so on. We cannot remember the scrap of a given image. Yet, this pertains only to memory. Partial loss of function happens during perception of actual events. A patient with damage to the visual cortex may see only half of a visual scene [263]. Output of motor reactions after brain injury also may be damaged partially. Therefore, if each image is spread within the brain, how may we resolve the "binding" problem that is a necessity to unite a whole body of information relating to a given image within a single neuron? An image cannot be spread over brain as a photo. Each feature of the image must be spread and each image must be projected through a great number of neurons. This would be possible if each neuron remembers little by little about each and every event. It would be sufficient, if a neuron would remember only the necessity to generate or fail to generate a reaction in response to a given stimulus. The outer limit of images which a neuron theoretically may discriminate is 2n , where N is the number of synapses converging to a given neuron. This colossal number to a marked degree exceeds the number of images which we meet during our life (222, if we remember an image each minute). In order to have the capability to recognize 222 images, a neuron must maintain specific chemical processes in 22 different synapses and all these processes ought to have the capability for mutual interactions. However, even this biochemical system is overcomplicated. Later we will give a theoretic description for these processes.

Thus, if the center of memory exists, it possesses some very perplexing properties. When memory grows old, it is rewritten from the place to the place, while a special "library" preserves important data. Besides, skill memory is kept in files of fast access and when a species is perfected it opens new memory centers, especially exposed to spoiling. This picture looks strange and unbelievable. Additionally, memory centers seem to have a common urgency system, which temporarily turns off all memory centers when damage threatens any one of them. In other words, memory may be spread within the brain, but access to memory depends on many circumstances. These data mean that one cannot identify neural centers with the elements of learning machinery. It is impossible to hope to find centers for short and long-term memory, apparatus for recording and reproduction, timer pulsing, a catalog of addresses, a coincidence detector, etc. All these functions are concentrated in each place of the neural system. Probably a small part of neural tissue is capable to learning and memory and, maybe, a morphologic unit of brain a neuron – serves as its functional unit. It important to understand what is the upper limit of neuron function.

Nonetheless, in neurobiology there are examples of learning forms that appear to be localized in narrow areas and are incapable of compensation after removal of these areas. These are the songs of birds and classical eyeblink conditioning. The latter example has undergone extensive analysis. Classical eyeblink conditioning (about 100-200 pairing) typically involves paired presentations of a tone or light as a conditioned stimulus (CS+) and a periorbital shock or air puff as an unconditioned stimulus (US) and represents unconscious skill learning. The essential neural circuitry for acquisition and performance of classical eyeblink conditioning is found disposed in the cerebellum and related brain stem structures (especially the interpositus nucleus). The local nature of eyeblink conditioning is displayed in morphological alterations that accompany its manifestation; the number of excitatory synapse within the interpositus nucleus increases after training by 1.6 folds [640]. At the same time as this system appears to be highly localized, many other brain areas beside cerebellum are recruited during eyeblink conditioning. Learning-related activity has been found, for instance, in the auditory cortex [1441, 1356], hippocampus [865] and in a variety of other structures including the neostriatum, thalamus and trigeminal nucleus [1175]. Comprehensive analysis reveals that even though learning-related plasticity occurs in many regions of the brain, maintenance of plasticity is critically dependent on processes that occur in the cerebellum.

Although the cerebellum appears to be absolutely necessary for establishing the basic eyeblink conditioning in all training situations, the cerebellum alone does not encode all features of eyeblink conditioning. Other brain structures and circuits are critically involved in encoding various aspects of classical eyeblink conditioning and it is likely that plasticity in at least some of these areas may be established independently of the cerebellum. While interpositus nucleus lesions abolished eyeblink conditioning established with a tone CS+ and air puff US (a somatic response), the same lesions had no effect on conditioned changes in heart-rate (an autonomic component of the response). This means that when eyeblink conditioning cannot be recollected after lesion of the interpositus nucleus, safety of the memory is not disturbed.

Eyeblink conditioning is a simple form of automatic skill learning, characterizing by a low specificity in respect to CS+ and CS~, which was never paired with the US. Only after 200 presentations of CS+ and US and 200 presentations of single CS~, conditioned motor responses in neurons of the cerebellar interpositus nucleus began to be slightly different (CS+ – 85%, CS~ – 60%). Neuronal response to the CS+ exceeded response to the CS~ with latency at 200-275 ms (75-0 ms before US). Responses with shorter latencies were not different [865, 411, 1175]. Interestingly, another well known higher localized specific response, to faces of famous persons, also had long latency, between 300 and 600 mS after stimulus onset [1005]. It is necessary to denote that specific response to the CS+, typically has short latency, from 7 to 100 mS and usually from 10 to 30 ms [1292, 1003, 1351, 352, 1037, 1356, 951, 8]. Evaluating the biological significance of the signals shows that the sensory part of responses is specifically altered. By the way, during eyeblink conditioning to a click (as a CS+), specific response to a click in the auditory cortex of the awake cat shows latency as short as 8-12 ms [1356]. A secondary (1216 ms) temporal component of response to the click was not specific. The executive parts of responses also may change, but they are observed already after decision-making. Thus, later components are generated after completion of decision-making. Therefore, a highly localized eyeblink conditioning response in the cerebellar nucleus is, evidently, an executing component of the reflex and it, certainly, is not connected with the memory.

As in the case of eyeblink conditioning multiple brain areas are involved also in other forms of primitive behavior: in conditioning of siphon withdrawal in Aplysia [769], vestibuloocular reflex gain [429], two-neuron spinal reflex [1351], etc. Simple behavior changes do not happen due to single alterations at single areas. Not only brain, as a whole, but neural centers are multitasking structures and participate in many behaviors. On the one hand, a given behavior, under artificial conditions may be controlled by a local site of brain and on the other hand, various brain areas participate in given behavior and various brain centers participate in various behaviors. An intact brain may be important for fine coordination.

We will demonstrate further that the chemical nature of memory has a firm basis. Conversely, an importance of spatial distribution of synaptic terminals is impossible to reject. Usually neurons receive heterogeneous chemical information, but they can recognize synaptic patterns which consist of inputs transmitting the same substances. The only difference between patterns in this case is a spatial distribution of the inputs.

Next post:

Previous post: