Merlin Donald Origins of the Modern Mind Harvard University Press. 1991

pg. 148 Keywords: episodic memory - semantic memory - event perception - The difference between an ape and a human child apparently resides in how event perceptions are encoded and remembered. In the ape, where episodic memory is the dominant device, the event can only be remembered in a literal, situation-specific manner APE CULTURE - HOMINID CULTURE EPISODIC MEMORY It is difficult to account for the evolution of human cognition without considering the cultures of apes and hominids. As Lovejoy pointed out in his treatment of the bipedal locomotion of australopithecines, traditional one-dimensional explanations for the adoption of erect posture—that walking erect enables better hunting and self-defense in tall grasses, for instance—are inadequate; walking erect would have been a fatal adaptation without other major changes in the survival strategies of apes. These included changes in habitat, food sharing, rates of procreation, and child-rearing strategies. The result is a much richer, and more convincing, theory of how australopithecine cullture evolved. Cognitive skills are necessarily implemented in a culture, which we define as a collective system of knowledge and behavior. The culture typical of a given species reflects the cognitive capacities of the individuals making up the culture. It follows that, when considering the origins of a radical change in human cognitive skill, we must also look at the sequence of cultural changes. Cultural patterns constitute our most fundamental information on the cognition of animals, and ethological observation is a more accurate and more efficient method of mapping out basic cognitive features of a species than laboratory experimentation. Knowledge can be furthered by controlled experiment; but a rich ethological framework gives meaning and context to the experimental data. The same principle should apply to human cognition; there ought to be a cognitive ethology of human culture, a temporal framework for the emergence of mind. There are no existing classifications of ape or human culture that are based primarily on cognitive considerations. Ape cultures are usually described in terms of feeding habits or territory, as in "frugivorous" or arboreal" cultures. Human cultures are sometimes specified in terms of technology - for instance, Stone Age and Bronze Age cultures. Sometimes they are specified in terms of religion and myth: Campbell (1959, 1988) refers to the cultures of "animal powers' referring to the common creation myths of certain aboriginal cultures. Sometimes they are named after their manner of obtaining food: Hunter-gatherers, slash-and-burn agriculturalists, pastoralis[s, and so on. They have been labeled after their tools, trading practices, diet, level of social organization, and geographical setting. They have been labeled as repressive and post-repressive, warlike and peaceful, patrilineal and matrilineaL But they have not typically been classified in terms of their predominant cognitive features. In fact, the only such classification in wide use is along the dimension of literacy which is not particularly useful in characterizing premodern cultures, since they were all preliterate. Yet when we are trying to develop a strategy for bridging the gap between humans and the rest of the animal kingdom, cognition is the most important dimension along which cultures are distributed. A cognitive classification of culture could be built on a number of cognitive dimensions, but the most likely place to start would seem to be in the area of representational strategy. Modern humans have many apparently novel systems of representation in memory, and their development may be the central processes underlying our cultural evolution from the apes. If apes are taken as the starting point, how might their overriding representational strategy be described ? Despite their formidable skills, they lack language, and they also lack much of the nonverbal knowledge evident in humans who have been stripped of language. Their behavior, complex as it is, seems unreflective, concrete, and situation bound. Even their uses of signing and their social behavior are immediate, short-term responses to the environment. In fact, the word that seems best to epitomize the cognitive culture of apes (and probably of many other mammals as well, although this is tangential to the argument) is the term episodic. Their lives are lived entirely in the present, as a series of concrete episodes, and the highest element in their system of memory representation seems to be at the level of event representation. Where humans have abstract symbolic memory representations, apes are bound to the concrete situation or episode; and their social behavior reflects this situational limitation. Their culture might be therefore classified as an episodic culture. EPISODIC MEMORY: Although used here in a rather idiosyncratic sense, the term "episodic" is ultimately derived from the commonly used term for concrete or time-bound memory, which Tulving (1983) labeled as episodic memory. Episodic memory is, as the name implies, memory for specific episodes in life, that is, events with a specific time-space locus. Thus, we can remember the specifics of an experience: the place, the weather, the colors and smells, the voices of the past. Typical examples of episodic memories are found in the details of specific experiences: a death in the family, first love, and so on. Such memories are rich in specific perceptual content. By definition, episodes are bound in time and space to specific dates and places. The important feature of this type of memory is its concrete, perceptual nature and its retention of specific episodic details. PROCEDURAL MEMORY: The ancient foil to episodic memory is procedural memory. Procedural memory is quite different and structurally more archaic. For the most part, procedural memories can be regarded as the mnemonic component of learned action patterns. Simple organisms can learn patterns of action without any detailed episodic recall; procedural memory involves the storage of the algorithms, or schemas, that underlie action. Sherry and Schacter (1987) have observed that in terms of its storage strategy, procedural memory is the opposite of episodic memory. Whereas episodic memory preserves the specifics of events, procedural memory preserves the generalities of action, across events. Procedural memories must preserve general principles for action and ignore the specifics of each situation. For example, in learning to catch a ball one must learn the principle of tracking a moving object, no matter what the speed of the object, the starting point, or one's initial posture at the time it is thrown. It would be cumbersome to remember the exact speed, starting point and position of each successful practice catch; a new throw is unlikely to match any specific counterpart in practice. Thus, learning a procedure, even on this level, involves setting parameters and forming general rules . Detailed episodic recall would interfere with this process. Episodic and procedural memory involve different neural mechanisms, as can be shown in birds, who will lose their songs (a procedural memory system) if lesioned in one nucleus, and their ability to hide and relocate food (an episodic memory system) if lesioned in another. The same distinction exists in humans, as seen in amnesics. In the famous case of H.M., followed by Milner and her associates (1966, 1975) for over twenty years, the patient developed catastrophie anterograde amnesia following neurosurgery. He retained a capacity for acquiring new procedural memories, that is, he could still learn new motor skills. But his capacity for new episodic memories was destroyed; he could not record any new events in his life. For instance, he had to be reintroduced over and over to the doctors who were treating him after his surgery. And although he acquired new motor skills he could not recall ever having learned them. Thus, new procedural memories were acquired but the specific episodes during which they were acquired were not recalled. Both episodic and procedural memory systems seem to be present in a variety of animals, including mammals and birds. Sherry and Schacter (1987) reasoned that episodic memory evolved separately from procedural memory for the very good reason that their storage strategies are mutually incompatible. Whereas procedural memories generalize across situations and life events, episodic memory stores the specific details of situations and life events. Thus, one memory system stores the generalties and discards the specifics; the other system, the episodic, stores the speci6cs but does not generalize. Obviously the same neural mechanism would have difficulty doing both therefore, two separate mechanisms evolved for the two types of storage, and the distinction has endured across many species. Episodic memory is apparently more evolved in apes than it is in many other species, in the sense that apes are sensitive to subtleties of social and pragmatic situations that other animals cannot register. This reminds us that there are significant gradations in the computational power of the episodic memory system; the complexity of its contents may vary widely between species. This will become clearer in the following section, which deals with event perception. But even though it may vary tremendously in power, episodic memory differs fundamentally from procedural memory in that it involves a degree of conscious awareness; an animal lacking a capacity for episodic memory and restricted to the procedural level would be not much more than a stimulus-response organism, a high-level automaton of the sort favored by the early behaviorists. SEMANTIC MEMORY: Episodic memory also differs fundamentally from the dominant form of human memory, specifically semantic memory. The third category mentioned by Tulving in his original taxonomy was semantic memory, which is usually considered to be symbolic in nature and characteristic of humans. The kinds of facts usually tested on IQ tests or college entrance examinations involve semantic memory. The closest parallel to human semantic memory in animals might be found in the signing behavior of trained apes; but even apes who sign or use visual symbols do not appear to store up large numbers of facts and propositions about the world, the way humans do. Since humans and nonhuman mammals, including apes, differ so fundamentally in the types of memories they can retain, it is possible to use this fact to characterize their two types of society. Most animals, including humans, possess procedural memories, and therefore the term is not particularly useful in characterizing the dominant cognitive feature of mammalian culture. Episodic memory is probably unique to birds and mammals, forming the basis for Oakley's definition of rudimentary consciousness. Humans possess both procedural and episodic memory systems, but these have been superseded in us by semantic memory, which is by far the dominant form of memory in human culture, at least in terms of the hierarchy of control. In contrast, episodic memory is dominant in most mammals, including apes. Animals do not seem to possess the systems of representation that would allow them to have elaborate semantic networks. Their experience, in this light, is entirely episodic. The pinnacle of episodic culture, the culture of the great apes, marked the starting point of the human journey.The dependence of apes upon episodic memory throws light on their difficulty with sign language, even when trained by humans under extraordinarily favorable conditions. Signing has a procedural aspect, which is simply the motor "skill" of reproducing the movement that constitutes the sign. And it might be expected to have a semantic memory aspect, much like the human use of words. But the use of signing by apes is restricted to situations in which the eliciting stimulus, and the reward, are clearly specified and present, or at least very close to the ape at the time of signing (Terrace and Bever, 1980; Savage-Rumbaugh, 1980). Situational specificity is not typical of semantic memory. If apes possessed an abstract semantic representation to which the sign referred, this concrete, situational limitation would not apply.The reason apes use signs in such a concrete manner is that they are using episodic memory to remember how to use the sign; the best they can manage is a virtual "flashback" of previous performances. Thus, their understanding of the sign is largely perceptual and situation-specific. To say their understanding is perceptual in nature could be misunderstood to mean that they only perceive simple features of the environment. In fact, event perception is the most evolved form of cognition and the basic component of episodic memory. The episode is the "atom" of ape experience, and event perception is the building block of episodic culture. Event Perception in Apes Event perception is, broadly speaking, the ability to perceive complex, usually moving, clusters and patterns of stimuli as a unit. This property also characterizes object perception, but event perception resolves input into much more than a single object: motion and context are taken into account. A passing car constitutes a perceptual event, and so does a kick, or a threatening grimace, or the lifting of a spear, or a hand sign. Event perception is a subject of some importance in the fields of auditory perception, touch, and artificial intelligence research (McCabe and Balzano, 1986). Visual event perception, in particular, has been studied in recent years. Poizner, Klima, and Bellugi (1987) have tried to model the signs of American Sign Language for the Deaf (ASL) as complex visual events. They have attempted to specify the physical parameters that might be applied to decoding ASL signs, and it is clear that such visual events are not simple. But they are indeed elementary when we consider what mammals can perceive in the environment; they can correctly perceive not only individual patterns of motion but also social situations in which many different agents and objects are involved over a significant period of time. The perception of events is the ultimate objective of the perceptual process, at least in reasonably complex animals. Intelligence in animals might even be defined in terms of the complexity of events they can perceive. Animals that we call intelligent are those that respond to events of increasing complexity and abstraction. Apes can discriminate hand signs that are too complex or subtle for dogs; but dogs can read aspects of behavior that are missed completely by rats. Events can be arranged in a hierarchy of complexity; the simplest events are those that are closest to the level of object perception. A hand sign is an object in motion, and the perception of a hand sign or visual emblem as a unified event is well within the capacity of an ape. Yet, this level of event perception obviously does not suffice for language. Complex events are made up of smaller perceptual segments. Higher mammals, including apes, have no difficulty in perceiving complex life events, even when the perceptual segments change. A dog can assess, quickly and effortlessly, a visual array that includes a female in heat, another dominant male, and the presence of various humans on the scene. His behavior will be affected by a variety of contingencies and variables, each of which can be understood only with reference to the other elements present in the situation Substitute a different cast of characters, change a single element, including things as subtle as the weather conditions or the distances between the animals and people in the scene, and the dog's behavior will change. Perceiving and reacting to events of this complexity are the normal stock-in-trade of higher mammals. Apes are particularly good at visual-event perception. In the "insight" experiments cited earlier, Kohler's monkeys and apes were asked to assess complex visual events in the laboratory; they proved more than capable of assessing the relationships between the various elements present in the room. Their capacity for event perception "solved" the problem of how to reach the banana; they were able to break down the perceptual components in the situation and imagine another arrangement of the same components. This type of complex event perception, which bears close resemblance to perceptual decomposition in Corballis's sense, appears to be the highest achievement of the episodic mind. The difference between an ape and a human child apparently resides in how event perceptions are encoded and remembered. In the ape, where episodic memory is the dominant device, the event can only be remembered in a literal, situation-specific manner. Thus, the "meaning" of an ASL sign to an ape is simply the episodic representation of the events in which it has been rewarded. This is not qualitatively different from more conventional forms of operant conditioning; the only difference is in the degree of perceptual "intelligence" shown by the ape, that is, in the complexity of the events so encoded. The human child eventually breaks out of this episodic world and develops completely different semantic representations. Apes and other large brained mammals such as whales, elephants, dogs, and cats may differ greatly in which perceptual events they can encode, but they share the same cognitive limitation: their dominant form of memory is episodic, and their cultures, whatever their individual features, are therefore episodic. In this respect, these cultures are globally different from human culture, in which the dominant forms of memory are semantic. Although the mechanisms of event perception are not well understood, progress is being made in constructing models of simpler forms of perception, including elementary event perception. Perception, by its very nature, extracts generalities. Recent parallel distributed processing (PDP) models of early perceptual processing demonstrate this principle in a convincing manner. McClelland and Rumelhart's Interactive Processing model (1986) showed that, even on the most elementary level of processing, circuits tend to generalize their initially learned perceptual categories to new items on the basis of similarity. Perceptual learning, in other words, even in simulations, does indeed involve differentiation, as James Gibson wrote in 1950. But it involves more than differentiation of the object. At the level of object movement the perceiver effortlessly extracts what appear to be perceptual templates of action. Johansson's (1968) classic experiments with the perception of dance steps illustrates this point. He placed a few luminous strips of tape on dancers, either on major limbs or on joints, and then had them go through various routines in the dark. Videotapes of these luminous patterns were immediately understood by observers; in fact, for those who were familiar with the dancers, it was even possible to recognize individuals from their style of walking and moving, with nothing but faint luminous lines and points as cues to their body shapes and positions. The only way one could explain this capability would be to posit a perceptual process that could immediately and effortlessly perceive the resemblance between the overall patterns of movement of the luminous lines and the movement patterns stored during normal perceptual learning. This implies a level of perceptual processing that integrates entire patterns of action. This is a prerequisite for elementary event perception.

HOME      BOE     SAL     TEXTE