Chapter 18Learning and Memory

Let me begin by telling a little story.When I was a graduate student we hadto take an exam that Cornell does in

an interesting way. They put you in a swivel-chair surrounded by your committeecomposed of 4-5 faculty members. You arespun around, and a question comes from thedirection you face when you stop. One of thequestions I was asked was, “What will bethe most important accomplishment in thefield of neuroscience in the next 10 years?”Knowing that knowledge usually advancesin small steps, I said that we wouldgradually know more about most aspects ofthe field, but I didn’t expect any majorbreakthroughs. To a man, the committeeagreed that in 10 years we would have acomplete solution to the problem of learningand memory. The word complete was theirterm, presumably meaning that we wouldknow everything there was to know aboutlearning and memory. Well, at the 10-yearmark, I sent them all a letter saying, “I toldyou so!” Their predicted event didn’thappen. However, we have made someprogress; that progress is what this chapter isabout.

DefinitionsWhat is learning? According to Eric

Kandel (2000) “Learning is the process bywhich we acquire knowledge about theworld.” While this definition is erudite, itdoesn’t help us much in knowing what tostudy. Another definition (Kimble, 1961),"Learning refers to a more or less permanentchange in behavior which occurs as a resultof practice," is a little better. It tells us thatlearning is more or less permanent; it won’talways be there, but often will. It also tellsus that this is something that happens

because we practice–repeat something overand over. A further definition says,“[Learning is] either a case of differentialstrengthening of one from a number ofresponses evoked by a situation of need, orthe formation of receptor-effectorconnections de novo; the first occurstypically in simple selective learning and thesecond in conditioned reflex learning" (Hull,1943). It is the strengthening of existingresponses and the formation of newresponses to existing stimuli that make thisdefinition unique. So what is learning? Itisn’t clear that we have an inclusivedefinition. It appears that learning is thestrengthening of existing responses orformation of new responses to existingstimuli that occurs because of practice orrepetition. How much practice? Sometimes asingle practice session is sufficient as inavoidance of painful or noxious stimuli.Sometimes a lot of practice is necessary asin learning to drive a car.

What then is memory? Againaccording to Kandel (2000), ". . . memory isthe process by which that knowledge of theworld is encoded, stored, and laterretrieved." By this definition, memory is nota thing; it’s a process. Interesting! In anotherdefinition, "Memory is a phase of learning . .. learning has three stages: 1. acquiring,wherein one masters a new activity . . . ormemorizes verbal material . . . 2. retainingthe new acquisition for a period of time; and3. remembering, which enables one toreproduce the learned act or memorizedmaterial. In a narrower sense learningmerely means acquiring skill . . ." (Sargent& Stafford, 1965). From these definitions,we see that memory has to do with keeping“knowledge” someplace and then retrieving

it when it is needed. What we don’t see hereis that the “knowledge” doesn’t have tocome into consciousness. I have twocars–one with an automatic transmission,one with a stick shift. I don’t have to bringinto consciousness the process for shiftinggears when I get into the car that requires meto do that–I just do it!

Types of MemoryThere are actually two basic kinds of

learning and memory. One is declarative orexplicit; the other is non-declarative orimplicit. Knowledge of facts–what we knowabout places, things and people–and themeaning of these facts is explicit memory.These things must be recalled intoconsciousness to be used. Patients who havebilateral medial temporal lobe lesions havean inability to learn and remember items offactual knowledge. They can’t rememberpeople that they met the day before. Theycan’t remember what they did the daybefore. Some people will further parcelexplicit memories as episodic (we rememberevents) or semantic (we remember facts). AsKandel (2000) points out, in either case thecontent of all explicit memories can beexpressed by declarative statements such as“I was here yesterday” (episodic) and “Thehippocampus has something to do withmemory” (semantic).

Implicit memory involvesinformation about how to performsomething; it’s recalled unconsciously. Weuse implicit memory in trained, reflexivemotor or perceptual skills. I know how todrive my car; I know how to get to work.The same people with bilateral medialtemporal lobe lesions can learn simplereflexive skills–they habituate and aresensitized, they can be classically andoperantly conditioned (see later). They canlearn certain perceptual tasks. For example,they can recall a word learned previously

when given only the first few letters of theword. At the same time, they deny everhaving learned the word previously. Implicitmemory is often further parceled asassociative and non-associative. There aretwo well-known types of non-associativelearning: habituation and sensitization.Habituation is a decrease in response to abenign stimulus when the stimulus ispresented repeatedly. A dog will be arousedwhen a strange tone is played. If the tone isplayed over and over, the dog willeventually no longer be aroused by the tone.We say that it has habituated. This kind oflearning makes sense; it is not efficient foran organism to go on responding to astimulus that has no meaning. The otherform of non-associative learning,sensitization, is an enhanced response tomany different stimuli after experiencing anintense or noxious one. For example, ananimal responds more vigorously to a toneof lesser intensity once a painfully loud tonehas been played. Here we say that the animalis sensitized.

These two forms of learning alsointeract. Once a response has beenhabituated, it can be restored bysensitization. In this case, we say that theanimal is dishabituated. As an example: ahabituated startle response to a noise can berestored by strongly pinching the skin.

In non-associative learning, it is notnecessary that the animal learns to associatethe stimuli involved (thus the name). Forexample, the dishabituated animal does notlearn to associate the noise with the pinch.As we shall see shortly, this is the hallmarkof associative learning. Not all forms of non-associative learning are as simple ashabituation and sensitization. For example,we learn language by imitation of peoplewho already speak. This involves noassociation of stimuli and is clearly morecomplicated than habituation.

In associative learning, we “learn”that two stimuli are associated with eachother or that a response is associated with agiven event or has a given consequence.Perhaps important in clinical considerations,a person can also learn that an outcome isnot associated with a response. So a personmay learn that what happens to him is notrelated to what he does. Two sorts ofassociative learning have been well studied:classical conditioning and operantconditioning. Classical conditioning is welldemonstrated by Pavlov’s famousexperiment in which he presented meatpowder to a dog, causing it to salivate. Herepeated the presentation, and each time thedog salivated. If he repeatedly rang a belljust before presenting the meat powder (theywere paired), the animal came to associatethe bell with the presentation of the meatpowder, and it would begin to salivate whenthe bell was rung. In fact, for a while itwould salivate if the bell was rung but nomeat powder was presented (they wereunpaired). After a while, the bell stoppedpredicting the presentation of meat powderfor the dog, and it ceased salivating when itwas rung. (This process is called extinction.)It should be noted that for classicalconditioning to occur the ringing of the bellmust precede the presentation of the meatpowder, often by a certain critical interval oftime (of the order of 0.5 sec). One way tolook at classical conditioning is to think ofthe bell as becoming a signal that the meatpowder is about to be presented.

In Pavlov’s paradigm, the meatpowder normally elicits salivation withoutexperimenter intervention (it is innate orperhaps previously strongly learned), and itis called the unconditioned stimulus (US).The response is called the unconditionedresponse (UR). The bell comes to elicitsalivation only after it is repeatedly pairedwith meat powder; so it’s called the

conditioned stimulus (CS). The response toit (again salivation) is called the conditionedresponse (CR). The UR and the CR areusually similar but often not identical in typeor strength.

Initially investigators thought thatclassical conditioning involved simplylearning that two stimuli werecontiguous–that they occurred close togetherin time, one after the other. Now we thinkthat what the animals learn iscontingencies–that existence of somethingdepends upon existence of something else.With this in mind, it is possible to see thatsimply learning that two stimuli werecontiguous could often lead to behaviors thatwere maladaptive, with animals associatingenvironmental events that had no realrelationship. On the other hand, theexistence of superstitious behaviors, even inhumans, suggests that this does occur.

It is tempting to think of extinctionas an example of forgetting, but alas it is not.The difference is that something new islearned during the process of extinction–theanimal learns that the CS is no longer asignal that the US is about to occur, rather itis a signal that the US will not occur.

In operant conditioning (sometimescalled trial-and-error learning), a person oranimal learns that it gets a reward if it doessomething. So, a pigeon learns that it getsfood if it pecks at a certain key, but not if itpecks at another. A rat learns that it canavoid getting an electric shock if it presses abar at a certain time. Presumably what theanimal learns is that one of its manybehaviors (pecking or bar pressing) isfollowed by food. It is constitutional inanimals to repeat behaviors that lead topositive reinforcement (something pleasantor the absence of something unpleasant) andavoid behaviors that lead to punishment ornegative reinforcement.

Figure 18-2 - Block diagram of the supposed flow of informationbetween areas associated with learning and memory.

Figure 18-1. The relative positions of parts of thelimbic system involved in learning and memory.

Neuroscience of Learning andMemory

A great deal has been written aboutthe kinds and properties of learning.What has been said here is probablyenough for the purposes of this chapter.If you want to know more, you canconsult any good textbook on learningor the psychology of learning. We wantto know about what is going on in thebrain when a person or animal learnssomething, stores what has beenlearned and later retrieves it for use inbehavior.

Explicit MemoryIn overview, experiments on learning

can be interpreted to say that explicitmemory is first acquired through one ormore of the three polymodal associationareas of the cerebral cortex, namelyprefrontal, limbic and parieto-occipital-temporal. Then, the information istransferred to parahippocampal andperirhinal cortices, entorhinal cortex dentategyrus, hippocampus, subiculum and back toentorhinal, parahippocampal and perirhinalcortex. The locations of these areas relativeto one another are shown in Fig. 18-1,whereas a block diagram of the connections

is shown in Fig. 18-2.

Different forms of learning areaffected differentially by lesions in differentlocations. Damage to parahippocampal,perirhinal and entorhinal cortices producesgreater deficits in memory storage for objectrecognition than does hippocampal damage.Right hippocampal damage produces greaterdeficits in memory for spatial representation,whereas left hippocampal damage producesgreater deficits in memory for words, objectsor people. In either case, the deficits are information of new, long-term memory; oldmemories are spared.

Current thought is that thehippocampal system does the initial steps inlong-term memory storage–different partsbeing more important for different kinds ofmemory. The results of hippocampalmachinations–presumably memories–aretransferred to the association cortex forstorage.

There is no general semantic(factual) memory store; rather memories of asingle event can be stored in multiplelocations. This make sense when it isrecalled that a single memory has multiplefacets–each event contains sounds, smells,tastes, somatosensory experiences, visual

Figure 18-3 - A reverber-ating circuit: Neuron Aexcites B and vice versa.

images and so forth. Long-term storage ofepisodic (event) memories seems to occur inprefrontal association cortex.

So, each new explicit memory isformed by four sequential processes:

Encoding-information for eachmemory is assembled from thedifferent sensory systems andtranslated into whatever formnecessary to be remembered. This ispresumably the domain of theassociation cortices and perhapsother areas.Consolidation-converting theencoded information into a form thatcan be permanently stored. Thehippocampal and surrounding areasapparently accomplish this.Storage-the actual deposition of thememories into the final restingplaces–this is though to be inassociation cortex.Retrieval-memories are of little use ifthey cannot be read out for later use.Less is known about this process.

Implicit MemoryImplicit memories are stored

differently depending upon how they areacquired. “Fear conditioning” (training thatinvolves use of fearful stimuli) involves theamygdala. Operant conditioning involves thestriatum and cerebellum. For example, eyeblink conditioning is disrupted by lesions ofthe dentate and interpositus nuclei of thecerebellum. Classical conditioning,sensitization and habituation involve thesensory and motor systems involved inproducing the motor responses beingconditioned. Perhaps surprisingly, certainsimple reflexes mediated by the spinal cordcan be classically conditioned even after thecord has been surgically isolated from thebrain. So, it appears that all regions of thenervous system may be capable of memory

storage.

Processes of Learning

Given the definitions for learningand memory, what sort of mechanismswould we expect to find in the nervoussystem? One early thought was that neuronsin “memory” pathways were arranged inreverberating circuits. In such a circuit, oneneuron excites another and the other excitesthe one such that, once the circuit isactivated, action potentials run aroundcontinuously. An example of this kind ofarrangement is shown in Fig. 18-3. Here areshown only 2 neurons in the circuit but anynumber may be included. If this kind ofarrangement accounts for memory, then anyevent that temporarily stopped activity in thecircuit should disrupt memory.

Unfortunately forsupporters of theidea,electroconvulsiveshock, whichtemporarily stopsor resets allelectrical activityin the nervoussystem producesonly a significant,transitory loss ofrecent memory, but

no loss of older memories. Some years ago, the psychologist

Donald Hebb (Hebb, DO (1949) TheOrganization of Behavior: ANeuropsychologi-cal Theory. New York:John Wiley) mulled this problem and cameup with a principle that has become knownas Hebb’s rule. Briefly, the principle is“When an axon of cell A . . . excites cell Band repeatedly or persistently takes part infiring it, some growth process or metabolicchange takes place in one or both cells so

Figure 18-4 - Simplified neural circuits involved inthe habituation process in Aplysia. There are about 24sensory neurons in the siphon; these are glutaminer-gic. They synapse on 6 motor neurons that innervatethe gill and various interneurons as shown. Thecontrol condition is shown on the left, the habituatedcondition on the right.

Figure 18-5 - Sensitization is produced by applying anoxious stimulus to the tail of the Aplysia’s tail,activated sensory neuron 2. This, in turn activates afacilitating interneuron that enhances transmission inthe pathway from the siphon to the motor neuron.

that A’s efficiency as one of the cells firingB is increased.” As we shall see, currentthought is an extension of Hebb’s rule.

HabituationWhat happens in the nervous system

to produce habituation? Experimentsperformed in Aplysia californica, the seaslug, were designed to address this problem. Their results are shown schematically inFig.18-4. If the siphon of the animal isstimulated mechanically the animalwithdraws the gill, presumably forprotection. That action is known to occurbecause the stimulus activates receptors inthe siphon, which activates, directly orindirectly through an interneuron, themotoneuron that withdraws the gill. This is asimple reflex circuit. All of this is shown onthe left side of the figure. With repeated

activation, the stimulus leads to a decreasein the number of dopamine-containingvesicles that release their contents onto themotoneuron. There appears to be no changein the sensitivity of postsynaptic NMDA ornon-NMDA receptors. As yet, we don’tknow why the dopamine release decreases. Itis presumed that habituation in vertebrates,including man, occurs by a similar process.

SensitizationIn sensitization, a stimulus to one

pathway enhances reflex strength in another.An example, again taken from experimentsin Aplysia, is shown in Fig. 18-5. Again,stimulation of the siphon leads the animal towithdraw the gill by activating sensoryneuron 1, which in turn activates a

motoneuron. If the tail of the animal isstimulated just before the siphon is, then thewithdrawal of the gill is quicker and moreforceful. The mechanism of this appears toinvolve serotoninergic, axo-axonic synapses.As shown in the figure, activation of thesensory receptors in the tail activates,through sensory neuron 2, a facilitatinginterneuron that excites sensory neuron 1 in

Figure 18-6 - The synaptic and chemical eventsunderlying presynaptic facilitation involved inproducing sensitization. See text for details.

Figure 18-7 - Long-term storage of implicit memoryfor sensitization involves changes shown in Fig. 18-6plus changes in protein synthesis that result information of new synaptic connections.

the pathway leading the gill withdrawal. Itdoes this either at the cell body or at theterminals of the sensory neuron on themotoneuron or the interneuron. Theconsequence of the sensitization process isto increase the size of the EPSP in themotoneuron without increasing the responseof sensory neuron 1. This will cause agreater response in the motoneuron andtherefore a greater withdrawal of the gill.

How all this occurs is illustrated inFig. 18-6, which shows an axo-axonicsynapse as might occur between thefacilitating interneuron and sensory neuron1. Serotonin (5-hydroxytryptamine or 5HT)is released by the presynaptic axon onto thepostsynaptic axon where it binds to

receptors and activates a G protein that, inturn, activates adenylyl cyclase to producecAMP. This cAMP activates a cAMP-dependent protein kinase, PKA. Along withanother kinase, PKC, PKA phosphorylatesand closes K channels (hypopolarizing thecell), mobilizes vesicles for exocytosis and

opens Ca channels. The end result is thatactivation of this 5HT pathway by tailstimulation causes more transmittersubstance to be released by siphonstimulation, the resulting larger EPSP leadsto a larger response by the gill.

With only short-term tail stimulation,the sensitization will fairly quickly disappearwhen tail stimulation ceases. However, thesensitization can be made relativelypermanent by repeated tail stimulation. Thislong-term sensitization (and also long-termhabituation) occurs because there arestructural changes that occur in thepresynaptic terminals (sensory neuron 1, forexample). With sensitization, there is an upto 2-fold increase in the number of synapticterminals in both sensory and motoneurons.Alternatively, with habituation, there is aone-third reduction in the number of

Figure 18-8 - A. Experimental setup for demon-strating LTP in the hippocampus. The Schaffercollateral pathway is stimulated to cause a responsein pyramidal cells of CA1. B. Comparison of EPSPsize in early and late LTP with the early phaseevoked by a single train and the late phase by 4 trainsof pulses.

Figure 18-9 - During normal low-frequency trans-mission, glutamate interacts with NM DA and non-MNDA (AM PA) and metabotropic receptors.

synaptic terminals. Both of these changesrequire altered protein synthesis bymechanisms shown in Fig. 18-7.

Long-term PotentiationAs previously detailed, the

hippocampus is important in storage ofdeclarative memory. In 1973, a phenomenonwas described in the hippocampus that mayaccount for declarative memory. Since thenthe same phenomenon has been observed invarious other places known to be involved inmemory storage. This phenomenon is calledlong-term potentiation (LTP).

A high-frequency train of stimuliapplied to fibers afferent to the hippocampusincrease the amplitude of EPSPs in the targetneurons. The increase lasts for days orweeks and requires activation of severalafferent axons together. This property has

been termed cooperativity, and it resultsfrom the requirement of NMDA receptorsthat glutamate bind to them and the cell behypopolarized, the binding opens thechannel and the hypopolarization displacesMg++ that blocks the channel lumen. Alsorequired is that the pre- and postsynapticcells both be active at the same time. Thislatter property has been termed associativity.The astute student will see that this isprecisely the condition that Hebb’s law saysshould exist.

The experimental setup fordemonstrating LTP is shown in Fig. 18-8A.Recordings are made intracellularly fromCA1 neurons of the hippocampus whilestimulation is applied to the Schaffercollaterals of CA3 neurons. The amplitudesof the EPSPs in the CA1 neurons are shownin B. For a single stimulus, the amplitude ofthe EPSPs is plotted at 100%. When a trainof stimuli is applied instead, the amplitudeof the EPSPs augment to about 150%,whereas with 4 such trains the amplitudeincreases to 250%. Many people think thatlong-term potentiation is an example ofHebb’s rule at work and that it is thephysiological basis of memory.

During normal synaptic transmission(Fig. 18-9), glutamate binds to non-NMDAreceptors allowing cations to flow throughthe channels and the cell membrane to

Figure 18-10 - With high-frequency stimulation otherevents occur as described in the text.

Figure 18-11 - For LTP to last (Late LTP) the eventsof Fig. 18-10 must also lead to changes in proteinsynthesis and to formation of new synapticconnections.

hypopolarize. Glutamate also binds tometabotropic receptors, activating PLC, andto NMDA receptors. As you may alreadyknow, NMDA receptor channels can bindglutamate but no current will flow throughthe channels unless the Mg++ that binds tothe channel lumen is displaced. The latterevent can be effected by hypopolarizing thecell.

By contrast, during the early phase ofLTP, the high-frequency stimulation opensnon-NMDA glutamate channels leading tohypopolarization. This dislodges Mg++ fromthe NMDA glutamate channels, and Ca++

enters the cells. The calcium triggers theactivity of Ca-dependent kinases, PKC andCa-calmodulin, and tyrosine kinase. Ca-calmodulin kinase phosphorylates non-NMDA channels, increasing their sensitivityto glutamate and a messenger is sentretrogradely to the presynaptic terminal toincrease the release of transmitter substance.All of this is illustrated in Fig.18-10. Theseevents increase the transmitter released bypresynaptic terminals. With LTP, there is adecrease in transmission failure, i.e.,synapses are more reliable in excitingpostsynaptic cells. This is also shown in thefigure.

In the late phase of LTP (Fig. 18-11),calcium enters the cell and triggers Ca-calmodulin, which in turn activates adenylylcyclase and cAMP kinase. The latter

translates to the nucleus of the cell and startsprocesses that lead to protein synthesis andto structural changes, i.e., the formation ofnew synapses. Many scientists believe thatthis is the substrate for long-termmemory–the formation of new synapses.

There are still unanswered questionsabout the relationship of LTP to memory.First, memories last decades whereas LTPhas been observed only for days or weeks.How long LTP can be maintained is difficultto determine. Admittedly, LTP is thelongest lasting process known inneuroscience. Still memories may last muchlonger. LTP occurs in most or all of theplaces where memories are known to bestored. What is not known is whetherdisruption of LTP also interferes withmemory.

SummaryNon-declarative (implicit) memory

involves different brain regions: fearconditioning involves amygdala; operantconditioning involves striatum andcerebellum; and classical conditioning,sensitization and habituation involve sensoryand motor systems used in the responses.This kind of memory involves a number ofprocesses: habituation involves decrease insynaptic strength from decreased transmitterrelease; sensitization involves increase insynaptic strength due to presynapticfacilitation; and classical conditioninginvolves increase in synaptic strength due topresynaptic facilitation that is dependent onactivity in both pre- and postsynaptic cells.

Declarative (explicit) memory alsoinvolves a number of brain regions: there isno general store for explicit memories;because the subject of memories ismultimodal, storage of different aspectsoccurs in different locations; thehippocampal formation is important inprocessing information for storage asmemory; and memories are actually storedin association cortex. This kind of memoryprobably makes use of long-termpotentiation. The early phase of LTPinvolves glutamatergic transmission;postsynaptic processes that produceenhanced sensitivity or receptors toglutamate as well as enhanced release oftransmitter substance. In the late phase ofLTP, protein synthesis leads to changes incell structure and formation of newsynapses.

References

Dudai, Y (1989) The Neurobiology ofMemory: Concepts, Findings, Trends.Oxford: Oxford University Press.

Hebb, DO (1949) The Organization of

Behavior: A Neuropsychological Theory.New York: John Wiley

Hull, CL (1943) Principles of Behavior.New York: Appleton-Century-Crofts.

Kandel, ER, JH Schwartz and TM Jessell(2000) Principles of Neural Science. NewYork: McGraw-Hill.

Kandel, ER and JH Schwartz (1982)Molecular biology of learning: Modulationof transmitter release. Science 218:433-443

Kimble, GA (1961) Hilgard and Marquis’Conditioning and Learning. 2nd Edition. NewYork: Appleton-Century-Crofts.

Nicoll, RA, JA Kauer and RC Malenka(1988) The current excitement in long-termpotentiation. Neuron 1:97-103.

Sargent, SS and KR Stafford (1965) BasicTeachings of the Great Psychologists.Garden City, NY: Dolphin Books.