Memory - a widespread biological system:

Memory can be divided into two main categories. There are memories as we usually understand them, that is episodic memories, and generally memories which have some sort of abstract or overt meaning to them (such as remembering what we had for breakfast, and how we felt after we ate it). But there is another type of memory function that we are not really aware of – learning responses based on environmental stimuli. Learning is in fact, a memory function.

The second type of memory function is known generally as stimulus-response learning. In stimulus-response learning, the more a given sensory stimulus is associated with a particular beneficial behaviour, the connections between the sensory input of the stimulus and the behavioural functions are strengthened. The structures involved in this type of memory are therefore the connections between sensory neurons and motor functions. This sort of learning has been demonstrated experimentally very many times. An example is monitoring the responses of rabbits to a puff of air directed at their eye. When the somatic sensors around the eye detect the blast of air, it elicits a reflex response of blinking. However, if the blast of air is preceded by an auditory input of a certain tone sounding, then the rabbit learns the association that a blast of air is to follow, and blinks in anticipation of that blast of air. This is shown by monitoring the onset of blinking with respect to the onset of the tone and the onset of the air blast. Without the tone, the rabbits blink after the air blast. With the tone, the rabbits blink before.

This process, often called classical conditioning, is not altogether cognitive – i.e., the rabbits do not hear the tone and necessarily “decide” to blink – it becomes an automatic response. This is down to the restructuring of neurons in the rabbit’s brain. In the brain, there are very many thousands of connections between neurons from many areas of the brain. The strengthening of these connections is the basis of learning and memory. To describe, I will focus on a basic description involving the rabbit blinking mentioned above. As I said, such a response relies on the strengthening of connections between sensory neurons and motor neurons. This process is described by the Hebb rule. The Hebb rule states that if a synapse repeatedly becomes active at about the same time the postsynaptic neuron fires, changes occur in the structure or chemistry of the synapse that serves to strengthen it. To simplify, let us assume that just one neuron detects the tone, one neuron detects the air blast, and one neuron controls blinking. The normal situation, without the tone, involves the neuron which detects the air blast firing, thus triggering the motor neuron controlling blinking. But with the tone present, the neuron detecting the tone fires as well, at around the same time the neuron detecting the air blast does. The synapse between the terminal button of the auditory neuron and the motor neuron strengthen to such a degree that the firing of the auditory neuron alone can cause the postsynaptic motor neuron to fire. The result is blinking as a response to the tone rather than the just the air blast. It is this basic system that allows for an enormous amount to be learnt as a response to environmental stimuli.

Investigators have found the mechanism that allows synapses to be strengthened. It is based on NMDA receptors located on the postsynaptic membrane. An increase of calcium in the postsynaptic membrane causes the strengthening of the synapse. NMDA receptors allow the influx of calcium only under certain conditions. The flow of calcium is usually blocked by a magnesium ion, but this ion is ejected when the postsynaptic membrane is depolarised. However, there needs to be another condition before calcium can flow and that is the activation of the presynaptic terminal button. This activation of the presynaptic terminal button causes it to release glutamate, a substance which the NMDA receptors are sensitive to. Only when these two conditions are satisfied can calcium enter the postsynaptic membrane. This calcium then activates calcium-dependant enzymes (such as PKC and CaM-KII), which in turn bring about greater release of glutamate from the postsynaptic terminal button. This is the mechanism that allows the strengthening of synapses so that learning can occur. This process is known as causing an increase in long-term potentiation, since it increases the magnitude of the EPSPs (excitatory postsynaptic potentials) in the postsynaptic cells over a long-term.

This process of learning can also be applied to operant conditioning. Operant conditioning (or instrumental conditioning) is more concerned with behavioural responses, which are more complex in nature than the simple reflex actions, such as the rabbits' blinking, which are involved in stimulus response learning. If behaviour in response to certain stimuli is beneficial, then that behaviour tends to be repeated. This is how animals learn from experience. This process is called reinforcement, and it has been found that it is associated with the release of certain neurotransmitters, dopamine being a particularly important one. These neurotransmitters can induce the synaptic plasticity mentioned above, by facilitating long-term potentiation. B. F. Skinner showed that animals could be made to repeatedly perform a certain task, such as pressing a lever, if that task was rewarded with a piece of food. The beneficial nature of a task corresponds to the release of dopamine agonists in certain areas of the brain, prominently the nucleus accumbens. To corroborate the reinforcing nature of dopamine, it has been demonstrated that laboratory animals, and humans, will repeatedly self-administer dopamine agonists, like amphetamine or cocaine. Further evidence for the importance of dopamine in operant conditioning is that injections of dopamine antagonists in laboratory animals results in no ability to learn from behaviour – presumably, it blocks the reinforcing effects of stimuli. Several areas have been identified as being locations of synaptic change in operant conditioning – the nucleus accumbens, the basal ganglia, the frontal cortex, and some suspect the prefrontal cortex is involved too.

Perceptual learning is an area that is much more greatly associated with memories. By this I mean those memories which are based on sensory information and we can “recall”, like “what did I have for breakfast”. Very simple perceptual learning can be done subcortically (such as the processes involved in the types of learning above), but for memories proper, we need to use many different areas in association with each other. Perceptual input causes changes in the neural circuitryof the appropriate sensory association cortex in accordance with the principles of Hebb law. These changes are the basis of memories. This has been inferred experimentally, for example by Mishkin (1982), by showing that lesions in the inferior temporal cortex (where the visual association cortex is located) disrupts an animal’s ability to remember what it has just seen. Such a lesion might mean that monkeys fail a task where it is presented with an image which is subsequently removed, then two images are presented for it to choose from, where touching the image that corresponds to the one it has just seen results in a reward.

In order for these changes in neurons and synapses to mean anything, there needs to be a system to combine them in a meaningful context. An example of such a system is the central nucleus of the amygdala. Here, associations between emotions and sensory stimuli happen. For instance, if in an experiment a rat is given a small electric shock via the floor of its cage just after a tone sounds, then the associations between the auditory inputs, somatosensory inputs, and the emotion of fear in relation to the tone, are made in the amygdala, thus producing a more complete memory than just a reflex action.

However, the main area of the brain involved in the recalling of memories is the hippocampus. Its function is to put events into context so that the correct information can be retrieved from other areas of the brain, thus recalling a memory. To do this it has many inputs from the sensory association cortices, the motor association cortex of the frontal lobe, the amygdala (which provides information on odours and dangerous stimuli as well as associating sensation with emotion), and also the neural circuits involved in classical conditioning. These inputs provide all the information needed to construct a complete memory given some kind of prompt for that memory (a question someone asks perhaps). The hippocampus can “know” an animal’s location in space through neurons that respond according to the relations among objects in the animal’s environment. It can therefore keep track of an animal’s location in space. This is the function that it is believed the hippocampus had originally, but evolution has enabled it to now do more. It has been shown that damage to the hippocampal formation does disrupt the ability to learn spatial relations. For instance, rats will not remember which arms of a radial arm maze they have already visited to look for food (however, they can learn which arms never contain food). Such rats also cannot learn the location of the hidden platform in Morris’ milk maze. Hippocampal damage disrupts ability to distinguish events in terms of time as well as space. This strengthens the idea of the hippocampus as providing a contextual basis for parts of memories to be associated together, as time and space are the qualities required for such a context. Research has shown that hippocampal damage disrupts the ability to distinguish context. For instance, going back to the blinking rabbits, Penick and Solomon (1991) found that rabbits with hippocampal damage would perform the whole air blasting task exactly the same when in a different location, whereas normal rabbits showed a change, in that they would lose some of the strength of the conditioned response. Apparently, the damaged rabbits could not remember the context in which the learning had taken place.

In humans, damage to the hippocampal system results in anterograde amnesia (the inability to remember any events after damage occurs, and also those events that had taken place a number of years before damage). Apparently, they are unable to reactivate the pattern of activity in the hippocampus which retrieves the memory. Such patients can still perform short-term memory functions, such as those needed to carry on a conversation, and ordinary stimulus-response learning appear unimpaired. For instance, HM, a very much studied anterograde amnesia sufferer, was found by Milner and Corkin to improve in mirror drawing tasks. Each time he attempted them he improved, even after a long spell between attempts. He had no memory of ever doing such a task before (he chalked his skill down to natural talent!), but clearly he had improved his motor skills in some way. What is clear is that anterograde amnesiacs have problems in relational learning, rather than such types of learning as stimulus-response learning. Thus, the hippocampus cannot be the only area involved in memory, though it is an important one. In fact, anterograde amnesia can be caused by damage to both medial temporal lobes, or a thiamine deficiency due to chronic alcoholism.

In fact, there must be other areas at work. Anterograde amnesiacs after all can still remember things some years prior to damage to the hippocampus. So if the hippocampus was the only structure used in the retrieval of memories, then such patients would remember nothing at all. What this other structure or mechanism is that allows retrieval of these very old memories, is still an unanswered question.

So, the hippocampus can be thought of as the centre of memory retrieval. But it requires inputs from very many areas of the brain, and in turn, these input areas themselves have many inputs. So from this perspective, the whole brain has some part to play in memory. For instance, it has been found that people who suffer damage in area V4 of the extrastirate cortex can not only no longer perceive colours, but they cannot remember what colours were like at all. However, simple perceptual learning, like stimulus-response learning, appears to be somewhat distinct from more complex memories, since hippocampal damage does not impair these functions. But these functions still go on in many areas of the brain – after all, we have to learn many things in many areas. The memory system is therefore very large and very complex, and as a result, may never be fully understood. Even those parts of memory function we do understand are still amazing and it is astounding that they can have such a sophisticated result as the recollection of events long since past.