Diagram of a synapse onto a dendritic spine showing measures of synaptic morphology that have been used to study experience‐dependent structural alterations at synapses. Experience can also increase the number of synapses, presumably either by forming them outright or by selectively preserving some synapses from a population that is continuously being replaced.

Morphological plasticity of synapses dependent on behavioral experience (top) or longterm potentiation (LTP) in hippocampus (bottom). Top: densities of synapses and neuronal nuclei were determined in upper occipital cortex (layers I‐IV), and these values were used to derive an estimate of the number of synapses per neuron. Rats were reared from 23 to 55 days of age in a complex environment (EC), in pairs in social cages (SC), or in individual cages (IC). Enriched rearing conditions resulted in an increased number of synapses per neuron. Neuronal density was decreased in EC animals because of a greater volume of neuronal somata, neuronal processes, and glia. Bottom: brief bursts of high‐frequency stimulation (100 H_z for 1 s or 200 H_z for 0.5 s) produced LTP of the Schaffer collateral‐CA1 system in rat. In the CA1 zone, LTP increased the number of synapses on the dendritic shafts (A) and the number of sessile spine synapses (B). Density of the much more common spine synapse was not affected.

Three events in the development of specific connections in the vertebrate nervous system: cell death, axon collateral elimination, and synapse elimination. It is estimated that 50% of the nerve cells that are initially formed die before postnatal life. Process elimination by the surviving neurons involves both the removal of whole pathways and the elimination of functional synapses within remaining pathways. In this way, specific connections and networks are sculpted out of the initially formed nervous system.

Pattern of arborization of a single geniculocortical afferent in the visual cortex of a 17‐day‐old kitten (left) and an adult cat (right). Prior to the segregation of cortical inputs into alternating left‐eye and right‐eye columns, the aborization of individual afferents extends uniformly over a disk‐shaped area >2 mm in diameter. This area is destined to be segregated into at least 4 columns. In the adult, after column formation is complete, individual afferents do not aborize uniformly. The arborization is divided into a number of clumps separated by gaps whose dimensions correspond to the size of individual ocular dominance columns. The gaps are filled with afferents serving the other eye.

Top left: ocular‐dominance distribution of 223 cells recorded from striate cortex of normal adult cats. Cells in groups 1 and 7 could be driven by only one eye. In groups 2, 3, 5, and 6, cells were driven better by one eye than the other. In group 4, cells were driven about equally by each eye. Bottom: ocular‐dominance distribution of 34 cells recorded from the left and right striate cortex of a cat without visual experience in the right eye from 9 wk to 6 mo of age. Majority of cells responded only to visual stimulation from the experienced eye. Top right: ocular‐dominance distribution of 126 cells recorded from cats without visual experience in either eye from ∼1 wk to ∼3 mo of age. Thirty‐seven cells did not respond to visual stimulation. In contrast to the effects of unilateral eye closure, after bilateral eye closure both eyes could drive cortical cells.

Summary of cortical visual areas and their known connections. There are 2 major routes from striate cortex (V1): one follows a ventral route into the temporal lobe via area V4 and the other follows a dorsal route into the parietal lobe via MT. Filled arrowheads indicate “forward” projections; arrowheads indicate “backward” projections; lines with filled arrowheads at both ends indicate “intermediate” projections; d indicates that the projection is limited to the dorsal portion of the area; m indicates that it is limited to the medial portion. Other potential pathways into the parietal lobe include those carrying input from the peripheral visual field (dotted lines).

Recognition performance on the delayed nonmatching‐to‐sample test. Left: monkeys see an object and then 10 s later must choose the novel, unfamiliar object. Different objects are used on every trial. Right: performance by groups with bilateral lesions of posterior temporal cortex (TEO) or anterior temporal cortex (TE) and a group of unoperated controls (N). Numbers to the left of the curves are average numbers of trials needed after surgery to relearn the basic task, which involved remembering a single object for 10 s. First point on the curve is the average final score in this condition. Animals were subsequently tested in the same task with delay gradually increased from 10 s to 120 s.

A chess‐specific memory skill. Left: board position after the 25th move of Game 6 of the 1984 World Chess Championship in Moscow between A. Karpov (black) and G. Kasparov (white). Right: a random arrangement of the same 19 pieces. After briefly viewing the board of a real game, master players can reconstruct the board from memory much better than weaker players. With a randomly arranged board, experts and beginners perform the same.

Acquisition of a memory skill. In 20 mo involving ∼190 h of practice (1 h/day, 3–5 days/wk), a college student increased his digit span from 7 to 79 digits. Random digits were read to him at the rate of 1/s. If a sequence was recalled correctly, 1 digit was added to the next sequence.

Extended digit span. Five amnesic patients (including patient HM) and 20 control subjects were read a sequence of 5 digits. If the sequence was repeated back correctly, 1 digit was added to the next sequence; if not, the same sequence was given until it was repeated correctly. Amnesic patients had a normal digit span but required an abnormal number of trials to learn supraspan strings of digits. No amnesic patient recalled more than 12 digits within the testing limit of 25 trials.

Serial‐position curves for 6 amnesic patients and 6 control subjects. Subjects were read 10‐word lists and then asked to recall them in any order. Percentage of items recalled is plotted as a function of their position in the list. Amnesic patients show the normal recency effect but an impaired primacy effect.

Serial‐position curves for normal rats and rats with hippocampal lesions. Rats first visited 8 arms of a radial maze in a fixed sequence; 20 s (no delay) or 30 min (delay) later they were given a choice test involving arms 1 vs. 2, 4 vs. 5, or 7 vs. 8. They were rewarded for entering the arm that they had visited earlier. A: performance of normal animals showing both a primacy and recency effect in the no‐delay condition. B: performance of normal and operated animals in the no‐delay condition. C: performance of operated animals in the no‐delay and delay conditions.

Acquisition and retention of a mirror‐reading skill despite amnesia for the learning experience. A: patients prescribed bilateral (BIL) or right unilateral (RUL) electroconvulsive therapy and depressed patients (DEP) not receiving electroconvulsive therapy practiced mirror reading during 3 sessions (50 trials/session). For the patients receiving electroconvulsive therapy, 1 treatment intervened between the first 2 sessions, and an average of 7 treatments intervened between the second and third sessions. B: at the beginning of the third session, subjects were tested for their recollection of the previous testing sessions (9‐point interview) and for their memory of the words they had read (recognition test, chance = 50%).

Intact priming in amnesia. Amnesic patients and control subjects saw common words and then were asked to recall the words (free recall) or were cued with the first 3 letters of the words and asked to recall them (cued recall). Amnesic patients were impaired in these 2 conditions but performed normally when they were given the first 3 letters of words and instructed simply to form the first word that came to mind (completion). Base‐line guessing rates in the word completion condition were 9%.

Diminution of priming effects across modality. Amnesic patients (AMN), alcoholic control subjects (ALC), and medical inpatients (INPT) read or heard words and then were cued visually with the first 3 letters of these words and asked to form the first words that came to mind. Priming was equivalent across groups and was higher when the study words and the test cues were in the same modality (right). (Hatching shows base‐line word‐completion performance when the words were not presented for study.) Amnesic patients were markedly impaired at recalling words, and recall was unaffected by the modality of word presentation (left).

A tentative memory taxonomy. Declarative memory includes what can be declared or brought to mind, as a proposition or an image. It includes both episodic and semantic memory and the related terms, working and reference memory. Procedural memory includes motor skills, cognitive skills, and simple classical conditioning, as well as habituation, sensitization, various perceptual aftereffects, and other instances where the facility for engaging specific cognitive operations is improved by experience.

Failure of several types of cortical insult to produce major functional disruption. These results were difficult to reconcile with electrical field theory or any hypothesis of cerebral organization based on horizontal intracortical conduction. A: multiple subpial knife cuts through the depth of the gray matter in sensorimotor cortex of monkey. B: similar knife cuts in visual cortex of cat. C: lateral and dorsal X‐ray views of tantalum wire insertions in visual cortex of cat. D: lateral X‐ray and dorsal surface views of mica plate insertions in visual cortex of cat.

Ground plan of Lashley's Maze III, used extensively in his efforts to localize the engram. S, starting compartment; F, food compartment. Rats were given 5 trials per day until 10 consecutive errorless trials were recorded. Normal animals required 19 trials and 47 errors to learn the maze and 40 days later relearned it in 2 trials and 7 errors.

Relationship between extent of cerebral lesion and errors in learning or relearning Lashley's Maze III. A: original learning (n = 37). Ordinate indicates the percentage of cortex removed for each animal; abscissa indicates the number of errors made during learning. B: relearning of the same maze task in rats with lesions made ∼2 wk after original learning (n = 59). Ordinate indicates the percentage of cortex removed; abscissa indicates the number of errors made during the postoperative retention test.

Striking localization of function in the monkey. Monkeys with bilateral lesions of the middle third of the sulcus principalis (B) were unable to relearn the delayed‐alternation task, whereas lesions of the anterior (A) or posterior (C) thirds of the sulcus principalis had little effect. Three cross sections through different levels of the lesions are also shown for each brain. Blackened areas represent the extent of sulcus principalis damage; stippled areas indicate ineffective periarcuate and inferior parietal lesions. A.S., arcuate sulcus; C.S., central sulcus; I.P.S., inferior parietal sulcus; L.F., lateral fissure; L.S., lunate sulcus; P.S., sulcus principalis; S.T.S., superior temporal sulcus.

Marked localization of language functions in the peri‐sylvian region of the left hemisphere. Six language‐related functions were tested during neurosurgery for epilepsy in a 30‐yr‐old female, bilingual in English and Greek: naming of pictured objects in English (N) and Greek (G), reading of simple sentences (R), verbal memory across a short distraction‐filled interval (VI, VS, or VO), mimicry of single and sequential orofacial movements (M), and phoneme identification (P). Each site stimulated is represented by a rectangle; symbols within rectangles indicate that consistent errors of a particular type were made during electrical stimulation of that site. VI, memory with stimulation at the time of registration of information; VS, memory with stimulation during the distraction interval; VO, memory with stimulation at the time of retrieval; A, site of speech arrest; F, sites of evoked face movement and sensation. Frequently only one of the tested language functions was disrupted at any site. Naming in English and Greek could be separately disrupted.

Columnlike organization of interdigitating projections to frontal lobe in monkeys. Upper right: tritiated amino acids (H³–AA) were injected into sulcus principalis (PS) of the left hemisphere to label callosal projections to the PS in the right hemisphere. Horseradish peroxidase (HRP) pellets were implanted in the posterior bank of the intraparietal sulcus (IPS) in the right hemisphere to label projections to the right PS. In this way, convergent projections to PS, located in the rectangle, were labeled in the same animal. Left: composite of 2 coronal sections through the convergence zone in the sulcus principalis of the right hemisphere. Labeled termination zones for callosal fibers (2,4,5,8,10 and 11) are indicated by coarse stipples; termination zones for the parietofrontal projection (1,3,6,7 and 9) are indicated by fine stipples. Calibration bar, 0.5 mm. Although the organization of the 2 projections is irregular and sometimes overlapping, the 2 inputs are distributed in approximately alternating columns that extend across all 6 cortical layers.

Functional parcellation of cortex beyond the macrocolumn. Diagram proposes a modular organization of macaque striate cortex. Each ocular dominance column (L and R) contains cells sensitive to all orientations and is interpenetrated as well by cells that are sensitive to color but not orientation. Layer 4C, the site of termination of afferents from the lateral geniculate, is also insensitive to orientation.

Diagram of the proposed neural circuit mediating a short‐latency (8‐ms) acoustic startle response in the rat. Arrows indicate synapses at points along the circuit. Habituation of the reflex appears to depend on changes in synapses formed by projections from ventral cochlear nucleus (VCN), ventral nucleus of the lateral lemniscus (VLL), or both. DLL, dorsal nucleus of the lateral lemniscus; CNIC, central nucleus of the inferior colliculus; DCN, dorsal cochlear nucleus; LM, medial lemniscus; MLF, medial longitudinal fasciculus; RGI, nucleus reticularis gigantocellularis; SO, superior olive; VAS, ventral acoustic stria; RPC, nucleus reticularis pontis caudalis. Labels to right show millimeters posterior to lambda.

Diagram of pathways for visually conditioned heart‐rate change in the pigeon. Upper panel summarizes the visual pathways that transmit conditioned‐stimulus information (whole‐field illumination). Each of the 3 ascending pathways (1 to the thalamus and 2 to the tectum) is capable of transmitting effective conditioned‐stimulus information, but interruption of all 3 in combination prevents acquisition of the conditioned response. Brackets indicate equivalent mammalian structures. Middle panel describes the route by which the telencephalic targets of the conditioned‐stimulus pathway ultimately access the portion of the avian amygdaloid homologue that is considered the start of the descending pathway mediating expression of the conditioned response. Lower panel illustrates this descending (conditioned‐response) pathway. AMD, nucleus archistriatum mediale, pars dorsalis; HOM, tractus occipitomesencephalicus, pars hypothalami; HV, hyperstriatum ventrale; LGNe, dorsal lateral geniculate nucleus; MFB, medial forebrain bundle; NCM neostriatum caudale, pars mediale; NIM, neostriatum intermedium, pars mediale.

Essential involvement of deep cerebellar nuclei in the classically conditioned eye‐blink response. A: sites where unit recordings did (filled circles) or did not (open circles) respond in relation to the amplitude and time course of the learned response. Large numbers above each section represent millimeters anterior to lambda; small numbers above each section represent millimeters lateral to midline. Small numbers to the side represent millimeters below bone at lambda. B; sites where stimulation did (filled circles) or did not (open circles) elicit eye‐blink responses. C: lesion of the dentate and interpositus nuclei that abolished the learned eye‐blink response to a tone without affecting the unconditioned eye‐blink response to an airpuff. D: composite of 3 lesions, shown in black, that were not effective in abolishing the learned response. ANS, ansiform lobule (crus I and crus II); ANT, anterior lobule; FL, flocculus; D, dentate nucleus; DCN, dorsal cochlear nucleus; F, fastigial nucleus; I, interpositus nucleus; IC, inferior colliculus; IO inferior olive; lob a, lobulus A (nodulus); PF, paraflocculus; VN, vestibular nuclei; cd, dorsal crus; cv, ventral crus; gvii, genu of the facial nerve; icp, inferior cerebellar penduncle; vii, facial nucleus; VCN, ventral cochlear nucleus; viii n, cranial nerve VIII.

Loss of the nictitating membrane (NM) conditioned eyelid response (CR) in the rabbit after a lesion of the left dorsolateral part of the interpositus nucleus of the cerebellum. Performance of the left eye was good on the last day of paired conditioned stimulus‐unconditioned stimulus training (LP) before the lesion. L1‐L4, 4 days of training on the left eye after the lesion showing that the CR was abolished and not relearned; R1‐R4, 4 days of training on the right eye after the lesion showing initial savings (R1) and further learning; L5, final day of training on the left eye without improvement. Each data point shows the average score for 30 trials.

Lateral view of the monkey brain showing one route for the cortical processing of visual information from striate cortex (OC) to inferotemporal cortex (TE). Area TE, like other higher‐order cortical sensory areas, projects to amygdala (amyg) and to hippocampus (hippo) via hippocampal gyrus and perirhinal cortex. The fact that medial temporal lobe lesions, which include hippocampus and amygdala, cause amnesia but nevertheless spare many premorbid memories suggests that the spared memories are stored upstream from the lesion, i.e., in neocortex.

Diagram of the vocal control system in songbirds, parts of which show seasonal fluctuations in size corresponding to fluctuations in singing. Arrows indicate anterograde connections between nuclei. Neural signals for song originate in nucleus interfacialis (NIF) and end at the vocal organ, the syrinx, via nucleus hyperstriatum ventrale pars caudale (HVc), nucleus robustus archistriatalis (RA), dorsomedial nucleus of nucleus intercollicularis (DM), and nucleus hypoglossus pars tracheosyringealis (nXIIts), in that order. Other nuclei [magnocellular nucleus of the anterior neostriatum (MAN), area X (X), and nucleus uva (UVA)] are inactive during vocalization. Hatched areas indicate known projection zones of the telencephalic auditory area (field L).

Summary maps from Penfield's series of patients showing where in each hemisphere experiential responses were produced by electrical stimulation. Top: lateral view. Middle: dorsal view. Frontal and parietal opercula have been removed to expose the plenum temporale. Bottom: ventral views. Experiential responses were obtained during 40 operations on 40 different patients out of the entire series of 1,132 patients. Experiential responses were obtained only in the temporal lobes in 7.7% of the 520 patients in which the temporal lobe was stimulated.

Temporal lobe structures from which experiential responses were elicited by electrical stimulation with and without afterdischarge. Eighty‐eight experiential responses were obtained from 35 patients. Medial temporal lobe structures were involved, often together with neocortical structures, in 82 responses. Experiential phenomena usually required the participation of limbic structures.

A: representation of the course and primary branchings of the dorsal tegmental bundle (DTB), which originates in the catecholaminergic cell bodies of the locus coeruleus (LC) and widely innervates the forebrain. C, cingulum; CC, corpus callosum; F, fornix; FS, fornix superior; HC, hippocampus; HR, hippocampal rudiment; MFB, medial forebrain bundle; ON, olfactory nuclei; ST, stria terminalis. B: later it was appreciated that locus coeruleus neurons reach neocortex by a trajectory that passes through the frontal lobe and then caudally within neocortex through the full longitudinal extent of each hemisphere. M, medial axons innervating medial cortex; L, lateral axons innervating lateral cortex.

Top left: effect of posttraining epinephrine on retention. Rats were trained on a 1‐trial passive avoidance task with a weak footshock. A: the most effective dose (0.1 mg/kg, given subcutaneously) significantly facilitated retention performance. B: as the injection of the best dose was delayed after training, its effect on memory decreased. Top right: effect of epinephrine on memory is influenced by the motivational conditions of training. Rats were trained using either a weak or strong footshock (FS) and given epinephrine immediately after training. Same dose that facilitated retention in the low‐footshock condition was disruptive in the high‐footshock condition. Bottom: epinephrine facilitated retention of control animals (ST‐sham) but had no effect on animals with lesions of stria terminalis (ST‐lesioned). Black areas (left) show step‐through latency of untrained animals. Numbers in parentheses show the number of animals in each group.

Top left: pathological findings in a representative case of Wernicke‐Korsakoff syndrome. Diencephalic localization of the lesions is shown in 3 coronal sections (from left to right) at the levels of the mammillary bodies, midthalamus, and pulvinar. Black areas show the extent of lesion. Bottom left: performance of 4 control monkeys (N) and 4 monkeys with lesions of the mediodorsal thalamic nucleus (MD) on the delayed nonmatching‐to‐sample task. Control monkeys required a mean of 140 trials to learn the task initially with a delay of 8 s, and the MD group required 315 trials. First point on the curve is the average final score during learning. Testing was then carried out at longer delays up to 10 min. Right, A: coronal sections through the thalamus showing the smallest (black areas) and the largest (hatched areas) extent of damage in the monkeys with MD lesions. Right, B: extent of fornix damage (hatched area) in 1 operated control animal. The extent of damage to MD averaged 38%. Adjacent thalamic nuclei were not consistently damaged. Cd, caudate nucleus; CM and CM‐Pf, centromedian nucleus; Hb, habenula; LD, lateral dorsal nucleus; LP, lateral posterior nucleus; MD, mediodorsal nucleus; VLc and VLps, ventral lateral nucleus; VPLc and VPLo, ventral posterolateral nucleus; VPM, ventral posteromedial nucleus.

Performance on 4 different tasks sensitive to human amnesia by normal monkeys (N) and monkeys with bilateral medial temporal lesions (H‐A) that included hippocampus, amygdala, and overlying cortex.

Combined data comparing the performance of normal monkeys (N), monkeys with hippocampal lesions (H), and monkeys with conjoint hippocampal‐amygdala lesions (HA) on the delayednonmatching‐to‐sample task.

Different kinds of amnesic patients perform similarly on many tests of new learning ability. Here, 4 kinds of amnesic patients and matched control subjects (healthy subjects, alcoholics, or depressed psychiatric patients) were presented 10 pairs of unrelated words (e.g., army‐table) and then asked to recall the second word when given the first word of each pair. Recall was attempted after each of 3 successive presentations of the word pairs. Case NA became amnesic for verbal material in 1960 as the result of a penetrating injury from a miniature fencing foil that damaged the left dorsal thalamic region. Data are also shown for 8 patients with Korsakoff's syndrome, 8 psychiatric patients tested 1 h after the fourth or fifth treatment in a series of prescribed bilateral electroconvulsive therapy (ECT), and 4 patients who became amnesic as the result of an anoxic or ischemic episode.

A: pattern of anterograde amnesia (A.A.) and retrograde amnesia (R.A.) in a severe case of head trauma, as determined by 3 clinical interviews after the trauma. Retrograde amnesia was temporally limited, with recent memory more affected than remote memory. B: temporally limited retrograde amnesia after electroconvulsive treatment, as determined by formal testing methods in mice and humans. Memory is considered to change for a long time after initial learning such that some material becomes unavailable (forgetting), while what remains becomes more resistant to disruption (consolidation).

Close connection between the medial temporal region and putative memory storage loci in the rest of cortex. Diagrams summarize cortical afferent and efferent connections of the rhesus monkey parahippocampal gyrus on lateral (inverted) and medial views of the cerebral hemisphere. Projections from widespread cortical areas converge on area 28 (entorhinal cortex), and most of these pathways are reciprocated by efferent projections.

Extensive remote‐memory impairment in patients with Korsakoff's syndrome in Boston (left) and San Diego (right), as measured by the same test of famous faces. Impairment is more severe for faces that came into the news in the most recent decade or two, presumably because anterograde amnesia was either already present or was gradually developing during this period. Impairment for more remote periods probably reflects true retrograde amnesia, i.e., loss of material that had been successfully learned before the onset of the syndrome. The higher level of performance in the San Diego population may reflect the easier method of cueing used for this group.

Left: lateral view of the monkey brain. Drawing below photograph shows cytoarchitectonic areas according to von Bonin and Bailey 503. Vertical line through the frontal lobe shows the level of the coronal section to the right. as, Arcuate sulcus; cs, central sulcus; las, lateral sulcus; mos, medial orbital sulcus; ps, sulcus principalis. Right: coronal section illustrating the location of the lesions most commonly used to study frontal lobe function in the monkey. Numbers refer to cytoarchitectonic areas following nomenclature by Walker 505.

Task‐specific impairments in short‐term or working memory caused by unilateral electrical stimulation of frontal or temporal neocortex. Horizontal lines indicate periods of stimulation. Top: performance on delayed response, with a cue presentation of 2 s and a delay of 8 s. Stimulation with 2‐s trains of 5.5‐ to 6.0‐mA current was applied to the posterior segment of inferotemporal cortex (temporal) or to dorsolateral frontal cortex (frontal) so that the electrodes straddled the sulcus principalis. Bottom: performance on delayed matching to sample, with a sample presentation of 2 s and a delay of 3 s. Vertical lines below time scale indicate the boundaries between sample, delay, match, and intertrial interval (ITI) portions of testing. Stimulation with 2‐s trains of 3.5‐ to 5.5‐ mA current was applied to the middle third of the temporal cortex (temporal) or to dorsolateral frontal cortex (frontal) so that the electrodes straddled the sulcus principalis.

Diagrams based on surgeon's drawings at the time of operation showing (in black) the estimated lateral extent of cortical excision in 7 representative cases of unilateral frontal lobectomy carried out for relief of focal epilepsy. Dots indicate points where electrical stimulation of the exposed cortex interfered with speech.

Impairment after unilateral frontal lobe lesions on a delayed‐comparison task. Two stimuli were presented 60 s apart, and subjects were asked to judge whether the second stimulus was the same as or different from the first. This task requires subjects not only to remember recent events but to suppress their memory for older similar occurrences so that judgments can be made about the most recent ones. Frontal patients performed well at nonsense figures, the only stimuli that were unique and did not repeat during testing.