This book presents a theory of behaviour based on the premise that nervous systems have evolved to enable animals to engage in a variety of useful activities. Information about the outside world is essential for most of these activities, but the theory that sensory input must shape its own information processing system is rejected. After the behaviourists banished the immaterial self from psychology they replaced it by stimulus input, and for many years behaviour was attributed entirely to sensory input. Only recently has the notion started to develop that it is the response mechanism of the brain that determines what stimuli are required to perform an action. In the model presented here, an executive system located in the frontal region of the brain employs the extensive reciprocal connections of the sensory cortex to select the input needed to guide the motor system. The consequences of having an autonomous response planner, instead of one subservient to outside stimuli, are far reaching. Neural representations of broad categories that can coexist with multiple distinct subclasses, and the related phenomenon of stimulus equivalence, become easier to understand, for example. It may also enable us to understand why we usually think that we make our own decisions. The book also has suggestions about the way serial order is learned and the role of the frontal regions of the brain in reinforcement, expectancy and response planning.
PSYCOLOQUY CALL FOR BOOK REVIEWERS
Below is the Precis of "The Autonomous Brain" by Peter M Milner (694 lines). This book has been selected for multiple review in Psycoloquy. If you wish to submit a formal book review please write to email@example.com indicating what expertise you would bring to bear on reviewing the book if you were selected to review it.
(If you have never reviewed for PSYCOLOQUY or Behavioral & Brain Sciences before, it would be helpful if you could also append a copy of your CV to your inquiry.) If you are selected as one of the reviewers and do not have a copy of the book, you will be sent a copy of the book directly by the publisher (please let us know if you have a copy already). Reviews may also be submitted without invitation, but all reviews will be refereed. The author will reply to all accepted reviews.
FULL PSYCOLOQUY BOOK REVIEW INSTRUCTIONS AT:
Psycoloquy reviews are of the book not the Precis. Length should be about 200 lines [c. 1800 words], with a short abstract (about 50 words), an indexable title, and reviewer's full name and institutional address, email and Home Page URL. All references that are electronically accessible should also have URLs.
AUTHOR'S RATIONALE FOR SOLICITING MULTIPLE REVIEW:
"I hope this book will attract the attention of psychologists, philosophers and others interested in the mind, to some advances in neuroscience that are relevant to cognition. In particular, I think the idea that the neural processes involved in motivation and intention play an essential role in the learning of percepts should be more widely recognised. I acknowledge that no single person can hope to be well informed about the entire fields of neuroscience and cognition and it is my hope that multiple reviews of the work by experts may expose the inevitable inaccuracies and gaps, providing a partial substitute for multiple authorship. Having a number of reviews in one readily accessible place would provide a valuable resource for the reader."
1. After a very brief review of some previous attempts to relate behaviour to brain function, the introductory chapter asks why the chicken crossed the road, or to be more general, what spurs an animal to action? The behaviourist's answer that a stimulus has elicited a habit is incomplete; activity is often initiated with no apparent external stimulus. Once an action is embarked upon, however, environmental stimuli are usually needed to guide it. We might say that the animal pays attention to the stimuli needed for effective action.
2. On the other hand, irrelevant stimuli must be prevented from disturbing ongoing activity, except in emergencies. This is illustrated by the behaviour of a hypothetical bug that approaches the smell of food but is repelled by light. If its behaviour were continuously under the influence of both smell and light, the bug would never reach either the food or a safe hiding place.
3. In animals that do not learn, a reinforcer (i.e., a stimulus that satisfies a need) also guides the response. In animals that learn this is not necessarily the case. Stimuli that are not satisfying in themselves may guide the response also. Different tasks require different guiding stimuli, which must therefore be selected for emphasis by the animal's response plan, or intention. To achieve this the executive system of the brain must exert a profound influence on the sensory systems, presumably via the strong reciprocal connections within these systems (Jones, 1974; Van Essen & Maunsel, 1983).
4. Learning is extremely important for human behaviour but the behaviour of many highly successful invertebrates exhibits hardly any long term change with experience. It seems likely that when learning occurs, it involves modifications of a highly organised core of innate behaviour. This is the case even at the highest levels of the evolutionary scale. A learning model should therefore have as its foundation a collection of stereotyped survival skills. During learning, some of these behaviours come under the control of stimuli that frequently accompany them. Thus, a dog learns to approach its food dish using vision, when initially it was attracted only by the smell of the food it contained. Just as animals become more interested in food when they are hungry, they also become more interested in stimuli associated with food. Attention originally directed to an innate reinforcer may be redirected, by learning, to an accompanying object.
5. The implications of such redirection for the neural representation of stimuli are interesting. Some motivational activity, at least, must be "hard wired" to sensory paths that convey signals from a satisfier, so that neonates may survive until they have learned about the world. Stimuli from objects closely related to a reinforcer, e.g., a food dish, acquire motivational properties, allowing them to attract attention. This link between a stimulus and a motivational state must be maintained, even though the pattern of receptor excitation may vary greatly from moment to moment as either the object or the animal changes position.
6.These and other issues raised in the introductory chapter are discussed at greater length in the subsequent chapters.
7. For the purpose of determining the neural basis of behaviour it is important to have in mind a model that effectively summarises the behaviour to be explained. The assumption is made in this monograph that learning is a mechanism that has evolved to allow automaton-like organisms to adapt rapidly to environmental change. It follows, therefore, that we should try to understand the basic automaton before speculating about how learning might modify it.
8. Approach and avoidance are very basic forms of behaviour. It is relatively easy to imitate the behaviour mechanically (e.g., Grey Walter, 1953) or conceptually (Braitenberg, 1984; Milner, 1970). Task switching is also a characteristic attribute of most animals; each task requires its own specific release and guidance stimuli. If unrelated stimuli reached the motor system, most performances would be seriously crippled. Furthermore, it is usually desirable for a task to continue until it is completed, despite extraneous stimuli that might attempt to release other responses.
9. The hypothetical automaton is therefore assumed to contain a bundle of useful action sequences, or intentions, antagonistically interconnected so that an activated plan effectively suppresses the onset of activity in any of the others. The crucial question is how an intention appropriate to the situation is selected in the first place. The neural circuits responsible for making the choice must have evolved, like all the other statistically improbable body parts, in the service of species survival. It is postulated that signals indicative of the current situation reach the appropriate intention via innate paths. The plan that receives the strongest input during a period of low inhibition (at the completion of a previous action, for example) is the one that gains control of the motor system.
10. One important source of input is the state of need of the organism, another is from reinforcing stimuli in the environment. The strongest combination of need and reinforcing stimulus activates a response plan innately programmed to alleviate the need. This activity suppresses other components of the goal selecting mechanism.
11. The next step is to translate the active intention into effective motor activity. In animals this usually involves activation of approach or retreat behaviour relative to a target, which is determined, of course, by the need. Thus the need selects a response plan and the plan then filters the sensory input to the locomotor system, allowing entry only to signals from needed targets.
12. After a successful approach there may occur consummatory behaviour like eating, courtship or fighting. During this phase of the task the attention is switched by the response plan to pertinent targets such as a particular smell or taste.
13. Learning: In a suitable environment an automaton like the one described could survive without the ability to learn, but nearly all animals can learn something. Most neural transmission is chemical and chemical reactions have lasting after effects that can form the basis of memory.
14. Anything new in an animal's environment may signify danger, but novelty can be detected only if the animal remembers its environment well enough to detect a change. The usual response to a novel stimulus is cautious investigation. If the change is benign, a new image of the environment is eventually established and behaviour returns to normal.
15. If a reinforcer is uncovered during an investigation, the novel stimulus acquires some of its reinforcing properties. If the investigation reveals nothing of interest, then the novel stimulus ceases to be interesting. One way of accounting for the behaviour is to postulate that the animal is motivated by the memory of the outcome of previous responses to the situation. For example, if during exploration of a maze a left turn is found to lead to food, the next time the animal is in the same situation an intention to turn left activates an association with food. The association carries some of the response releasing property of food, releasing the left turn that is being planned. If the animal plans to make a previously punished response, the inhibition associated with the response aborts the plan.
16. This is an expectancy model of learning based on Tolman's (1932) theory, but with a more explicit account of how an expectation influences behaviour. A further complication is that planning a response has implications for the type of stimulus to which attention is directed.
17. All neural theories of behaviour assume that external events are represented in the brain, though there is no consensus as to how big a role experience plays in establishing the representations. Hebb (1949), in accord with the strong empiricist bias of the time, postulated that the connections of the "association" cortex in the new born infant are random, and that representations, as he called cell assemblies, are acquired by experience. During the last half century evidence has accumulated indicating that the cortex has much more innate structure than previously suspected.
18. The Pavlovian model of learning treats recognition as a passive process. Synaptic changes connect an input pattern to a response, so that in future the stimulus pattern is recognised as requiring that response. There is no way for the Pavlovian memory trace to generate an activity in the absence of input. Other theories postulate that traces can become active independently of the sensory input they represent. Hebb, for example, assumed that neurons fired by a stimulus acquire connections with each other, to form loops around which impulses "reverberate". These reverberations persist in the absence of external input, giving rise to a memory. Implementing this reverberatory model would be very difficult, however. I would go as far as to say impossible.
19. Positive feedback is notoriously difficult to control, even in simple circuits with constant, linear connections. The connections between neurons are not simple and their effectiveness is highly dependent on the level of activity. Furthermore, cortical neurons are subject to continuous random input from innumerable sources. In the visual system (and to a lesser extent in other sensory systems) receptors are not stimulus specific. Most of the visual receptors stimulated by a house are also stimulated by horses, or mountains. On the other hand, the retinal image of an object, displaced or changed in size, activates entirely different populations of receptors. Training a reverberatory loop to respond to specific combinations of an almost infinite number of different receptors, if it could be done at all, would take far longer than it takes a baby to demonstrate recognition of objects.
20. Another major objection to the reverberating representation is that it tends to go its own way, ignoring changes in the input. When the theory of electrical transmission at synapses was abandoned in favour of a chemical theory, Hebb's reverberatory explanation of memory images was rendered unnecessary. Chemical activity produces long lasting effects. An alternative to reverberation (Milner, 1957) is that input to neurons of the sensory cortex leaves a persistent chemical trace that reduces their threshold to background activity. Thus for some time after stimulation (either by sensory input or by association) neurons remain sensitised to non-specific input and are repeatedly fired by it. Later, I suggest that the background activation may originate in the response planning mechanism.
21. The difficulties of explaining stimulus representation are aggravated in the visual system by the precarious relationship between retinal stimulation and stimulus object mentioned earlier. The problem is much less severe in the olfactory system. A given type of odour molecule always binds to the same receptors, whereas an object like a ball may stimulate a pattern of receptors anywhere on the retina. To avoid the complication of stimulus equivalence, the theory of neural representation developed in this chapter is based on the olfactory system.
22. Olfaction is closely related to some basic motivations, so it is easier to put forward a convincing argument for its innate organisation. An animal that is deprived of food, or water, searches for what it needs, often using its nose. Although it is possible, for example, that an insect lays its eggs on the leaves of a specific plant because its smell is familiar from its first days of feeding, a more likely explanation is that the species evolved a sensitivity to the smell of leaves that ensure a high survival rate of the grubs. In any case, the model is based on the assumption that smells of food, potential mates and predators are innately linked to corresponding motivational activity.
23. Odorous stimulation is signalled to subcortical nuclei and the olfactory cortex via the lateral olfactory tract. Although feedback and changes in sensitivity introduce variability, to a first approximation, the sensory profile of neurons in the primary cortex is genetically determined. These neurons are also innately connected to motivational systems, so that when the animal seeks food, or a mate, motivational activity sensitises the corresponding sensory neurons, enabling them to gain control of the motor system.
24. For most of this century psychologists have tried to understand the sensory systems without reference to their function, or how they may be affected by other brain activity. The modulation of sensory pathways by need illustrates that the engram of food, for example, is determined at least in part by motivational input to neurons of the olfactory cortex. Food is a primary reinforcer, so the initial engram is innate, but after an animal that can learn has received food from a caretaker on a number of occasions, stimuli from the caretaker acquire the meaning of food. If cortical neurons are being stimulated by sensory input at the same time as they receive input from executive activity (which includes motivational and intentional activity), their connections with both these inputs are potentiated, thus they behave in many ways like the sensory neurons that are innately fired by the motivation.
25. In this way, a group of cortical neurons that receive input from an initially meaningless stimulus can acquire associations with a background input to form an engram, whose meaning is related to the nature of the background input.
26. Association of Ideas: according to the model just described, stimuli are associated with each other via common input from the executive system. The smell of food is represented by neurons innately connected to hunger and the plans to eat. Objects and sounds that accompany eating also acquire connections from and to the same executive sources. Thus they can facilitate plans to eat and by so doing arouse the food engram with which the plans to eat are also associated.
27. Most neural theories of association blithely assume that there are ineffective connections between all the engrams representing concepts and that when two engrams fire at the same time the connections between them are potentiated. This may work in network models with a dozen or so concepts, but in the human brain with perhaps hundreds of thousands of concepts, total linkage is quite impossible. Associations between engrams must therefore take place through an intermediary; the reciprocal attention component of the executive system seems to be a likely candidate.
28. Stimulus equivalence refers to the process mentioned previously whereby the sensory systems categorise stimuli and the animal responds in equivalent ways to stimuli of the same category. For example, an indefinitely large number of shapes are recognised as dogs. A number of attempts were made to explain this phenomenon, none of them very convincing (Hebb, 1949; Khler & Wallach, 1944; Lashley, 1942).
29. Hubel & Wiesel (1959) discovered neurons in the visual cortex which respond to a line as it is moved within a large area of the visual field, and speculated that these neurons are driven by many neurons, each responding to the line at a different location. In theory, this process of convergence can be cascaded, until neurons are preferentially fired by quite complicated shapes anywhere within a large field. We can, however, distinguish objects as well as categorise them. This means that the recognition process takes into account both similarities and differences between percepts.
30. Differences between overlapping percepts tend to be ignored unless they have consequences. Visually, squirrels and skunks may at first be assigned to the same category but attempts to treat them in the same way may lead to different associations of some visual features. No neuron represents a single object. All sensory neurons are excited by a range of stimuli and a single visual stimulus excites many neurons. The stimulus is defined by the total combination of excited neurons. Changing the position of an object changes the pattern of neural activity, but in general the new group of neurons overlaps the old one to a large extent.
31. Some of the neurons fired by an object acquire reciprocal associations with ongoing activity in the motivation and response planning systems. Consequently, the same stimulus may acquire more than one representation. A newspaper may be read or used as a fly swat. Depending on an animal's intention it may perceive two different members of a class as the same or different. (A car may be seen as a Volks or a yellow vehicle, depending on what I am looking for.) In other words, an engram does not depend on the stimulus alone. Motivation or intention are just as important.
32. Associations established between intention neurons and engram neurons explain why, when the intention to repeat an action arises, sensory neurons which had been active on previous occasions are facilitated or fired. Thus the stimuli are recalled and attention is directed to them. The engram itself is of no use in guiding a response, convergence has robbed it of localising data. The presence of a ball may be signalled, but not its position. This problem is resolved, however, by the reciprocal sensory pathways that project the engram activity back to earlier levels of the sensory system, where the neurons still carry positional information. Another possibility, suggested by Singer (1994) and his colleagues, is that engram neurons fire synchronously with the neurons at earlier levels, which can then be identified by fast acting coincidence detectors. These investigators also speculate that synchronous firing explains the binding problem in perception; the question of how the cortical areas specialised for colour, movement, distance etc, are bound together to produce an integrated percept.
33. Most responses involve a sequence of muscular contractions, in which the timing and order of the movements are crucial. The question of how learned sequences are stored has interested psychologists for many years. In an important paper, Lashley (1951) pointed out that a chain of associations was not an adequate explanation, partly because repeated occurrences of a response are not always followed by the same response, and also, in different actions, a response may be followed by any number of different responses. He suggested that the elements of an action were first selected and then arranged in the correct order, in some way that he did not make clear.
34. In chapter 5 Lashley's outline is fleshed out using, as he did, verbal behaviour as an example of sequence learning. Infants recognise words before they are able to pronounce them. In fact, most domestic animals understand the meaning of a small number of words. Recognition implies the ability to learn order. A cat is not a tack or an act. The obvious conclusion is that the neural activity produced by c + a is not the same as that produced by a + c. An after effect of the neural activity produced by the first sound modifies the neural activity produced by the second and subsequent sounds. All the neurons fired by the sequence of sounds contribute to the final engram, which is not, in itself, sequential. It acquires meaning in the same way as an olfactory or visual engram. Although it is influenced by the order of the sounds, it would be difficult to recover that order from the engram. The order is required, of course, for pronunciation.
35. Before learning to pronounce a word it is necessary to know how to reproduce a heard sound. This ability is probably learned in the course of babbling, when sounds are heard coincidently with the motor activity required to produce them. At this stage hearing a simple sound may induce mimicry.
36. When we hear a short sequence of sounds (verbal or otherwise) it can be recalled as a sort of echo for a few seconds. This is a stored sequence, but it is ephemeral and erased by any subsequent sound sequence. It is, however, an important step in the learning of a word. One mechanism that I believe could account for the echo is that a sound produces a state of excitation in corresponding cortical neurons, but immediately inhibits their output. The inhibition decays faster than the excitation. Depending on the intensity of background activation, the neurons fire again after a short delay and again turn themselves off. Neurons excited by a second sound and immediately inhibited recover after an equal delay, and so on, resulting in a repetition of the neural activity produced by the sound sequence. The sequence is stored on a sort of virtual time base.
37. This is not a permanent solution, but it retains the elements until the whole sequence has been completed and an engram of the word has been established. Once that has occurred the engram may be used to capture a sample of the neurons representing the elements of the sequence, preventing them from acquiring other associations. The difficulty mentioned earlier, with repeated occurrences of identical elements within the same sequence, may be overcome by having some high threshold neurons. These are not fired by a single input from the sound but are fired by a second input, whilst those fired by its first occurrence are still inhibited. The high and low threshold neurons therefore acquire associations with different adjacent sounds, although they represent the same sound.
38. The neurons eventually captured to form the motor sequence for the word may be selected from the original echo neurons by connections they have from neurons of the engram once it has been established. Other motor sequences, such as unlocking a door, or tying a knot, may also be learned in a similar manner.
39. The first half of the book is concerned mainly with neural models inferred from behaviour. The second half brings us closer to the physiological substrate.
40. In the early 1950s several patients who had temporal lobe surgery for the relief of epilepsy became amnesic after the operation, as was evident from pre- and postoperative psychological testing (Penfield & Milner, 1958; Scoville & Milner, 1957). In two of these cases considerable atrophy of the unoperated temporal lobe was subsequently discovered, while a third case, HM, had a bilateral operation. The loss appeared to be correlated with the extent of damage to the medial temporal lobe, including the hippocampus and amygdala.
41. The amnesia was of a type called anterograde, first reported by Korsakoff (1889) in a chronic alcoholic patient. In such cases, the memory of preoperative events is not seriously impaired but the patient forgets recent events very quickly, in severe cases within minutes. Motor skill learning is not affected. Attempts to replicate the memory loss in monkeys and other animals were not very successful at first, but eventually tests involving short-term memory, such as delayed matching to sample, were found that were impaired by the lesion.
42. Many theories have been proposed to account for the difference between learning that is impaired by the hippocampal area lesion, and learning that is spared (Hirsh, 1974; Mishkin, 1982; Teyler & DiScenna, 1986; Warrington & Weiskrantz, 1970). The theory favoured in Chapter 7 is based on the idea that different types of synaptic change account for the selective effect of anterograde amnesia (Milner, 1961). It is proposed that the immediate effect of a stimulus on the cortex is to produce a very short (seconds) change in synaptic effectiveness. A small but very long lasting change remains after the immediate effect has decayed.
43. It is also postulated that the stimulus induces synaptic changes in neurons of the hippocampal area, which may last for hours or days (a phenomenon later demonstrated, and called long-term potentiation). These changes store the memory of the event until, if it is not repeated and is of no great importance, the potentiation fades away. If the event occurs or is recalled repeatedly, the small residual after effects in the cortex accumulate until they are strong enough to carry the memory. When that has occurred, the hippocampal area may be removed without loss of the memory. Lesion experiments consistent with this interpretation are cited.
44. Chapters 8 and 9 are concerned with the basal ganglia and frontal cortex, which has been identified with the executive mechanism of the brain. Lesions of the frontal cortex usually result in hyperactivity for a time. The effect is even stronger for lesions that include the dorsal striatum, which receives much of its input from the cortex. Lesions of the ventral striatum, which receives strong input from the hippocampus, cause persistent searching behaviour. These old observations suggest that the striatum has a generally inhibitory effect on response generation. Damage to a dopamine pathway from the substantia nigra to the striatum produces symptoms of Parkinson's disease, difficulty in initiating movement, from which it may be inferred that dopamine has a generally inhibitory effect on the striatum. Dopamine agonists or enhancers potentiate responding.
45. Stimulation of the medial forebrain bundle, through which the dopamine path passes, has been found to be rewarding. This suggests that rewards involve dopaminergic inhibition of the striatum, releasing motor output from inhibition. Research during the last 20 years has shown that this account is oversimplified. The striatum contains a number of systems, two of which oppose one another. One releases the motor system, the other inhibits it. Dopamine facilitates the first and inhibits the second; both actions potentiate responding. If a cortical input to the striatum is predominantly to the first path it will act like a reward, if to the second it will have an aversive effect, stopping a response.
46. The notion that the basal ganglia incorporate a motivational system which can translate cortical signals into either rewards ("voluntary" response releasers) or punishments ("voluntary" response inhibitors) is applied to a number of typical learning situations in this chapter. For example, it is suggested that novel situations have little effect on striatal activity, so that after an initial period of alarm has abated, there is no striatal inhibition of investigatory behaviour. As synaptic connections between neurons fired by the novel situation are potentiated by use, striatal activity increases and exploration is suppressed.
47. Cortical activity that is accompanied by dopamine release in the striatum acquires connections with the response facilitatory path, and may acquire inhibitory connections with the inhibitory path. Thus it may act as a reward even when no dopamine is being released by primary reinforcement. Aversive input from the thalamus presumably is connected to the response inhibitory striatal path. Any cortical activity during an aversive episode therefore acquires connections with the activated response inhibitory path.
48. The response release and response inhibitory paths are mutually inhibitory, with overshoot during recovery from inhibition. Thus escape from punishment results in a burst of firing in the reward path. Withdrawal of a reward, or of an expected reward, is punishing.
49. The final chapter reviews and attempts to reinforce the main themes of the book. The excessive devotion of traditional psychologists to empiricism is denounced. In particular, the notion that the neural representation of a percept must be created by experience in a randomly connected mass of cortex is shown to be untenable. An alternative model is presented which takes into account the role of accompanying intentional and motivational neural activity, and is more in keeping with anatomical and physiological data. The importance for effective responding of the motivational links to the sensory systems, via attentional paths, is reiterated.
50. The chapter, and the book, end with some problems and speculations concerning time and memory, such as how we know how long ago some remembered event took place.
Braitenberg, V. 1984. Vehicles. Experiments in Synthetic Psychology. Cambridge, MA: MIT Press.
Hebb, D.O. 1949. The Organisation of Behaviour. New York: Wiley, pp. 1-335.
Hirsh, R. 1974. The hippocampus and contextual retrieval of information from memory: A theory. Behavioural Biology, 12, 421-444.
Jones, E.G. 1974. The anatomy of the extrageniculostriate visual mechanisms. In: F.O. Schmitt & F.G. Worden (eds.), The Neurosciences: Third study program. Cambridge, Mass.: MIT Press, pp. 215-227.
Khler, W. & Wallach, H. 1944. Figural after-effects: an investigation of visual processes. Proceedings of the American Philosophical Society, 88, 269-357.
Korsakoff, S.S. 1889. Etude medico-psychologique sur une forme des maladies de la memoire. Revue de Philosophie, 28, 501-530.
Lashley, K.S. 1942. The problem of cerebral organisation in vision. In: H. Klver (ed.), Visual Mechanisms. Biological Symposia, vol. 7. Lancaster, Pa.: Cattell Press. pp. 301-322.
Lashley, K.S. 1951. The problem of serial order in behaviour. In: L.A. Jeffress (ed.), Cerebral mechanisms in behaviour. New York: John Wiley, pp. 112-136.
Milner, P.M. 1957. The cell assembly: Mk. II. Psychological Review, 64, pp. 242-252.
Milner, P.M. 1961. The application of physiology to learning theory. In: R.A. Patton (ed.), Current Trends in Psychological Theory. Pittsburgh, PA: University of Pittsburgh Press, pp. 111-133.
Milner, P.M. 1970. Physiological psychology. New York: Holt, Rinehart and Winston, Inc., pp. 1-531.
Mishkin, M. 1982. A memory system in the monkey. Philosophical Transactions of the Royal Society of London B, 298, pp. 85-95.
Penfield, W. & Milner, B. 1958. Memory deficit produced by bilateral lesions in the hippocampal zone. AMA Archives of Neurology and Psychiatry, 79, pp. 475-497.
Scoville, W.B. & Milner, B. 1957. Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery and Psychiatry, 20, pp. 11-21.
Singer, W. 1994. Coherence as an organising principle of cortical functions. International Review of Neurobiology, 37, pp. 153-183.
Teyler, T.J. & DiScenna, P. 1986. The hippocampal memory indexing theory. Behavioural Neuroscience, 100, pp. 147-154.
Tolman, E.C. 1932. Purposive Behaviour in Animals and Men. New York: Century.
Van Essen, D.C. & Maunsell, J.H.R. 1983. Hierarchical organisation and functional streams in the visual cortex. Trends in Neurosciences, 6, pp. 370-375.
Walter, W.G. 1953. The living brain. New York: W.W. Norton & Co. Inc., pp. 1-311.
Warrington, E.K. & Weiskrantz, L. 1970. Amnesic syndrome: consolidation or retrieval? Nature, 228, pp. 628-630.