Piero Scaruffi

Memories and Beyond (These are excerpts from, or extensions to, the material published in my book "The Nature of Consciousness") Memories are Made of This The mind’s cognitive faculties depend to a great extent on memory. If we could not learn and remember at all, our cognitive life would be virtually non-existent. If we could not remember where we live, where our office is, how to lace our shoes, how to drive a car, how to speak, and so forth, we would be mere objects devoid of real life. In fact, we would probably last only a few minutes. The more complex the organism, the more essential memory is to its survival because so much is required to keep the organism alive. Even the worst cases of amnesia do not completely erase memory. A patient who suffers complete amnesia does not remember anything from some date on in the past, but still remembers a lot of vital facts about living in the world. Evolutionarily speaking, memory provided a considerable advantage to creatures capable of remembering where water was or where predators lived. The more refined one's memory, the easier to navigate the environment, to survive in it and to find food in it. That animal memory is not just like a computer memory is a fact due precisely to the tasks that were required by it, tasks that rarely require perfection but do require speed and capacity. Learning to bike and memorizing the emergency phone number are both important tasks, but they are usually achieved in rather different ways. Thus it is not even correct to speak of "memory" as if it were just one task. It is probably more appropriate to speak of "memories". Memory is more than storage. Memory is also recognition. We are capable of recognizing a tree as a tree even if we have never seen that specific tree before. No two trees are alike. And even a specific tree never appears the same to us, as the perspective, the wind, the lighting can all dramatically change its appearance. In order to recognize a tree as a tree, and as a specific tree, we use our "memory". Whenever we see something, we ransack our memory looking for "similar" information. Without memory we would not see trees, but only patches of brown and green. The process of thinking depends on the process of categorizing: the mind deals with concepts, and concepts exist because memory is capable of organizing experience into concepts. Our mind, ultimately, looks like a processor of concepts. The mind's functioning is driven by memory, which is capable of organizing knowledge into concepts. So much so that, inevitably, a theory of memory becomes a theory of concepts, and a theory of concepts becomes a theory of thought. Cognition revolves around memory. All cognitive faculties use memory and would not be possible without memory. They are, in fact, but side effects of the process of remembering. There is a fundamental unity of cognition, organized around the ability to categorize, to create concepts out of experience. Memory's task is easily summarized: to remember past experience. But, unlike the memory of a computer, which can remember exactly its past experience, human memory never remembers exactly. The most peculiar feature of our memory is, perhaps, the fact that it is so bad at remembering. Our memory does not only forget most of the things that happen, but, even when it remembers, it does a lousy job of remembering. Memory of something is almost always approximate. Many details are forgotten right away. If we want to remember a poem by heart, we have to repeat it to ourselves countless times. And sometimes memory is also very slow: sometimes it takes a long time to retrieve a detail of a scene, sometimes it will take days before the name of a person comes back to mind. Rather than accessing memories by calendar day or person's name, we seem to access them by associations, which is a much more complicated way to navigate in the past. It is hard to think of something without thinking also of something else. It is hard to focus on a concept and not think of related concepts. And the related concepts that come to mind when we focus on a concept are usually things we care about, not abstract ideas. If we focus on "tree", we may also remember a particular hike in the mountains or an event that occurred by a tree. We build categories, we relate categories among them, we associate specific episodes with categories. For an entity that is supposed to be just a storage, anomalies abound. For example, we cannot count very easily. Do you know how your home looks like? Of course. How many windows does it have? You have looked at your home thousands of times, but you cannot say for sure how many windows it has. If you see a flock of birds in the sky, you can tell the shape, the direction, the approximate speed... but not how many birds are in the flock, even if there are only six or seven. Another weird feature of our memory is that it is not very good at remembering the temporal order of events: we have trouble remembering if something occurred before or after something else. On the other hand, our memory is good at ordering objects in space and at counting events in time. Human memory is a bizarre device that differs in a fundamental way from the memory of machines: a camera or a computer can replicate a scene in every minute detail, whereas our memory was just not designed to do that. What was our memory designed to do? The Reconstructive Memory A startling feature of our memory is that it does not remember things the way we perceived them. Something happens between the time we see or hear a scene and the time that the scene gets stored in memory. I can tell you the plot of a novel even if I cannot tell you a single sentence that was in the novel. If I tell you the plot twice, I will use different words. It would be almost impossible to use the same words. Nobody can remember all the sentences of a book, but everybody can remember what the "story" is. Compare with a computer. A computer can memorize the book page by page, word by word. Our memory does not memorize that way. It is not capable of memorizing a book page by page, word by word. On the other hand, it is capable of so many other things that a computer is not capable of. For example, we can recognize a plot, told by somebody else, as the plot of the same novel. That person's version of the plot and our version of the plot probably do not share a single sentence. Nonetheless, we can recognize that they are the same story. No computer can do that (yet), no matter how big its memory is. Size is obviously not the solution. In the 1930s the British psychologist Frederic Bartlett developed one of the earliest models of memory. Bartlett studied how memory "reconstructs" the essence of a scene. We can easily relate the plot of a movie, and even discuss the main characters, analyze the cinematography, and so forth, but we cannot cite verbatim a single line of the movie. We stored enough information about the movie that we can tell what it was about and perform all sorts of reasoning about it, but we cannot simply quote what a character said at one point or another. What Bartlett discovered is that events are not stored faithfully in memory: they are somehow summarized into a different form, a "schema". Individuals do not passively record stories verbatim, but rather actively code them in terms of schemas, and then can recount the stories by retranslating the schemas into words. Each new memory is categorized in a schema which depends on the already existing schemas. In practice, only what is strictly necessary is added. When a memory must be retrieved, the corresponding schema provides the instructions to reconstruct it. That is the reason why recognizing an object is much easier in its typical context than in an unusual context. The advantage of the "reconstructive" memory is that it can fit a lot of information in a relatively narrow space. Any memory that tried to store all the scenes, text and sound of a movie would require an immense amount of space. But our memory stores only what is indispensable for reconstructing the plot and other essential features of the movie, thereby losing lots of details but at the same time saving a lot of space. Reconstructing And Making Sense The mechanism of "reconstructing" the memory of an event is quite complex. There is more involved than a simple "retrieval" of encoded information. In 1904 German biologist Richard Semon had already speculated that memory was as much about retrieval as about storage. He introduced the concept of "engram": the unit of memory, or, better, the pattern used to encode it (the "memory trace"). He then introduced another concept: the "ecphoric stimulus": the cue that helps retrieve a specific memory. He noticed that the likelihood of finding a memory depends also on the cue that is used to retrieve it (the pattern used to decode it). We are often forced to remember something simply because we encountered a word or saw something that "reminds" us of something else. It was merely a fleeing moment, but enough to bring back the memory of something or somebody. Semon realized the power of cues: a cue is only a fraction of the engram, but it is enough to retrieve the whole engram. The fact that memory is not a linear recording of sensory input reveals that something helps memory make sense of the past. When memory reconstructs an event, it must have a way to do so in a "meaningful" way. The memory of an event is not just a disordered set of memories more or less related to that event. It is one flowing sequence of memories that follow one from the other. Sometimes you cannot finish relating the plot of a novel because you "forgot" a key part of it: the truth is that you forgot all of it and you were reconstructing it, and, while reconstructing it, you realized that something must be missing. You cannot reconstruct the plot because you have an inner sense of what reality must be like. You may know how the novel ends and how it goes up to a point, and then you realize that something is necessary in order to join that point to the ending. Your reconstructive memory knows that something is missing in the reconstruction because the reconstruction does not yet "make sense". The fascination of movies or novels is that you have to put together reality until it makes sense again. You have to find the missing elements so that the story gets "explained". Our brain has a sense of what makes sense and what does not. Cognitive Maps USA gestalt psychologist Edward Tolman can be credited with coining (in 1932) the concept of a "cognitive map": a rat knows how to navigate a maze because it maintains a cognitive map that covers much more than the rat has ever experienced directly. Tolman proved that rats build a cognitive map of an environment even without any reward, simply because they "were there". A cognitive map is a mental representation of the world in which we live. Cognitive maps both represent and participate in the creation of our experience of the world. A cognitive map is created and continuously improved through the individual’s experience and by interaction with other cognitive maps. At the same time, the cognitive map "is" the world, insofar as the individual is concerned. The map is how the world appears to be to the individual. The individual only knows her or his map of the world. This map thus works as an anticipatory schema, that determines what we expect to see and, ultimate, what we indeed see. The Partitioning of Memory The works of George Miller (1956), Donald Broadbent (1957), Allen Newell (1958) and Noam Chomsky (1957), that all came out in the second half of the 1950s, established a new paradigm in Psychology, broadly referred to as "cognitivism", that ended the supremacy of behaviorism. Where behaviorism was only interested in the relationship between input (stimuli) and output (behavior), cognitivism focused on the processing that occurs between the input and the output Herbert Simon and Alan Newell argued that the human mind is a symbolic processor and Noam Chomsky advanced a theory of language based on modules inside the mind which are capable of symbolic processing. The standard model that became popular in the late 1950s, due particularly to the work of the British psychologist Donald Broadbent and the USA psychologist George Miller, was based on the existence of two types of memory: a "short-term memory", limited to few pieces of information, capable of retrieving them very quickly and subject to decaying also very quickly; and a "long-term memory", capable of large storage and much slower in both retrieving and decaying. Items of the short-term memory move to the long-term memory after they have been "rehearsed" long enough. The idea was already implicit in William James’ writings (he called them "primary" and "secondary" memory), and in Theodore Ribot’s 1882 experiments on amnesia (the loss of memory is inversely proportional to the time elapsed between the event and the injury), but Broadbent also hypothesized that short-term memory may just be a set of pointers to blocks of information located in long-term memory. Broadbent also stated the principle of "limited capacity" to explain how the brain can focus on one specific object out of the thousands perceived by the retina at the same time. The selective character of attention is due to the limited capacity of processing by the brain. In other words, the brain can only be conscious of so many events at the same time. What actually gets the attention is complicated to establish, because Broadbent found out that attention originates from a multitude of attentional functions in different subsystems of the brain. Broadbent's model of memory (also known as the "filter theory") reflected at least two well-known features of memory: information about stimuli is temporarily retained but it will fade unless attention is turned quickly to it; the unattended information is "filtered out" without being analyzed. He drew a distinction between a sensory store of virtually unlimited capacity and a "categorical" short-term store of limited capacity. The latter is the way that a limited-capacity system such as human memory can cope with the overwhelming amount of information available in the world. At the same time, George Miller’s experiments proved that our short-term memory can hold only up to seven "chunks" of information and therefore provided an order of magnitude for it. It wasn’t clear, though, the "size" of a chunk: is the entire car a chunk of information, or is each wheel a chunk, or...? In Broadbent’s model, a chunk is a pointer to something that already exists. Therefore a chunk can be even very "big", as long as it is already in memory. Its "size" is not important (in short-term memory, it is only a pointer). This is consistent with experiments in which short-term memory proves to be capable of holding familiar images, but not of images never seen before. The British psychologist Alan Baddeley showed that a unitary short-term memory does not account for memory disorders and replaced short-term memory with a "working memory" that has basically three components: a short-term memory for verbal information, a short-term memory for visual information, and a control system. In fact, neural regions in the prefrontal cortex (the newest part of the brain, from an evolutionary standpoint) can draw data from other regions of the brain and hold them for as long as needed. The prefrontal cortex is unique in having a huge number of connections with the sensory system and with lower brain centers. The prefrontal cortex could be the locus of a "working memory", in which decisions, planning and behavior take place. Types of Memory Experiments performed in the 1970s by the Canadian psychologist Endel Tulving and his associate Daniel Schacter proved that "intension" (such as concepts) and "extension" (such as episodes) are dealt with by two different types of memory. "Episodic" memory contains specific episodes of the history of the individual, while semantic memory contains general knowledge (both concepts and facts) applicable to different situations. Episodic memory, which receives and stores information about temporally-dated episodes and spatiotemporal relations among them, is a faithful record of a person's experience. "Semantic" memory, instead, is organized knowledge about the world. Tulving believes these memory systems are physically distinct because their behavior is significantly different. In episodic memory, for example, the recall of a piece of information depends on the conditions ("cues") under which that piece of information has been learned (an explicit or implicit reference to it). There are at least two more aspects of memory that fall neither into the intension or extension. Procedural memory allows us to learn new skills and acquire habits. William James had been particularly interested in this kind of memory, having realized how important "habits" are to determine our behavior. He reduced habits to a sequence of "reflexes", i.e. stimulus-response events. Basically, each stimulus-response pattern, once learned, becomes the building block for more complex patterns which are our "habits", each of which is in turn a building block to create more complex "habits". The French philosopher Henri Bergson explicitly separated the memory of habits from the memory of events. Implicit memory is "unconscious" memory, memory without awareness: unlike other types of memories, retrieval cues do not bring about a recollection of them. Implicit memories are weakly encoded memories which can nonetheless affect conscious thought and behavior. Implicit memories are not lost: they just cannot be retrieved. Amnesia is the standard condition of human memory: most of what happens is not recorded in a form that can be retrieved. In the first years, because of incompletely developed brain structures, most memories are lost or warped. Nonetheless, memories of childhood are preserved without awareness of remembering. Implicit memory is the one activated in "priming" events, or in the identification of words and objects. That makes a grand total of four different types of memory: procedural, semantic, episodic and implicit. Tulving also devised a scheme by which memory can associate a new perception or thought to an old memory: the remembering of events always depends on the interaction (or compatibility) between encoding and retrieval conditions. It indeed appears that the brain accomodates several different memory systems, each of them involving the cortex but each characterized by different "pathways" leading from the cortex to other areas of the brain. Studies on amnesia (particularly by Neal Cohen in 1980) show that there are at least two separate memory systems: "declarative" memory (the memory that one can consciously remember, which is forgotten in an amnesia) and "procedural" memory (the skills and procedures which are usually not forgotten, as people with amnesia can still perform most actions they have learned throughout their lives). It appears that the hippocampus is the key to declarative memory, or at least the key to linking together declarative memories. Procedural memory is, instead, realized by circuits that involve the motor areas of the cortex and two loops that spread through the striatum and the cerebellum: acquiring skills is, indeed, a complex phenomenon. "Emotional" memory, on the other hand, seems to depend on the working of the amygdala, i.e. on yet another separate memory system. These three memory systems are physically connected to the cortex along different pathways, which means that they can work in parallel. The rememberer Tulving summarized the relationship between remembered and rememberer in the "encoding specificity principle": remembering depends on the affinity between encoding and decoding. Memories are encoded in a way that depends on the circumstances when the event originally happened. The likelihood of recalling a memory (of decoding it) depends on recreating those circumstances, on reinstating the same psychological state. In other words, the way we feel about an event play an important role in the way that event can later be recalled. For example, the feeling that I feel when I read a sentence is going to be important for later recalling that sentence. That feeling has become part of the episode, as it is encoded in my memory. Tulving's episodic memory packages different aspects of an event to give it the "autobiographical" feeling that makes it more than just a retrieval of information, it makes it a memory of something that happened in our life. In other words, an essential part of an episodic memory is the "rememberer". The rememberer does more than retrieve information about a past event: the rememberer experiences that event again. In fact, the episodic memory is more about the feeling of being there than about the event in itself: the feeling of the event is generally recalled in more accurate terms than the details of the event. In fact, it is easier to remember something that happened a long time ago but had a strong emotional impact on us than something that happened just minutes ago. I do not remember what I had for lunch two days ago, but I do remember episodes of my childhood that happened several decades ago (if I focus, I can even feel what I felt then). Ultimately, episodic memory is about the rememberer, not the remembered. Categories Arguably the most important function of memory is categorization. The rings of a tree or the scratches on a stone can be said to "remember" the past, but human memory can do more: it is capable of using the literal past to build abstractions that are useful to predict the future. It is able to build generalizations. Actually, categorization is the main way that humans make sense of their world. For example, if we analyze the grammar of our language, the basic mechanisms of meaning-bearing are processes of categorization. One can even wonder whether all living beings, or at least many of them, need to do some level of categorization in order to deal with the world. The German linguist Eric Lenneberg argued that all animals organize the sensory world through a process of categorization. They exhibit propensities for responding to categories of stimuli, not to single specific stimuli. In humans this process of categorization becomes "naming", the ability to assign a name to a category. But even in humans the process of categorization is still a process whose function is to enable "similar" response to "different" stimuli. For example, we "sit" on a "chair", regardless of how different this chair is from other chairs on which we sat in the past. Traditionally, categories were conceived as being closed by clear boundaries and defined by common properties of their members. In the 1950s the USA psychologist Jerome Bruner was influential in conceiving categories as sets of features: a category is defined by the set of features that are individually necessary and jointly sufficient for an object to belong to it. In order words, one can write down the rules that specify what is necessary and sufficient for a member to belong to a category. This seems to be the case for nominal types (the one invented by us, such as "mother" or "triangle"), but not necessarily for natural types. As the great Austrian philosopher Ludwig Wittgenstein pointed out, a category like "sport" does not fit the classical idea (both cards and chess and football are sports, but they have very little in common). A dog that does not bark or a dog with three legs or a vegetarian dog would probably still be considered a dog, even if it violates the set of features we usually associate with the concept of a dog. What unites a category is "family resemblance", plus sets of positive and negative examples. Its boundaries are not important: they can be extended at any time. Bruner was among the first to realize that most cognitive processes are nothing but classification processes in disguise. Cognitive activity ("thinking") depends on placing an event or situation in the appropriate category. A category is basically a set of events that can be treated the same way by the cognitive organism. Bruner also realized that categories are not "discovered" but "invented". They do not exist in the environment: they are construed by the human mind. Thus the inferences that matter are really the one that helps create a new category based on some events, and the inference that helps classify an event relative to the existing categories. A concept is a network of such inferences that allow to infer an event's category based on some observed attributes of the event, and then to infer the unobserved attributes of that event. His "functionalist" definition of an object, for example, is the network of inferences about it that one is capable of employing after an act of categorization. There exist different kinds of concepts that employ and trigger different kinds of inference. A common kind of inference is the one that all members of a category share some common attributes, but this is only one of the possible inferences to define a category. Prototype Theory The traditional view that categories are defined by common properties of their members was quickly replaced by Rosch's theory of prototypes. After all, the best way to teach a concept is to show an example of it. The USA psychologist Eleanor Rosch noted that some members of a category seem to be better examples of the category than others. Not all members are alike, even if they all share the same features of the category. This means that the features by themselves are not enough to determine the category. It also implies that there must exist a "best" example of the category, what she called the "prototype" of the category. In the 1970s she founded her early theory on two basic principles of categorization: 1. the task of category systems is to provide maximum information with the least cognitive effort; and 2. the perceived world comes as structured information. In other words, we do categorization because it helps save a lot of space in memory and because the world lends itself to categorization. Concepts promote a cognitive economy by partitioning the world into classes, and therefore allow the mind to substantially reduce the amount of information to be remembered and processed. In Rosch's theory of prototypes, a concept is represented through a prototype that expresses its most significant properties. Membership of an individual in a category is then determined by the perceived distance of resemblance of the individual to the prototype of the category. Next, Rosch proposed that thought in general must be organized around a privileged level of categorization. In the 1950s the USA psychologist Roger Brown had noted that children tend to learn concepts at a level which is not the most general and not the most specific: say, "chair", rather than "furniture" or "armchair". And in the 1960s the USA anthropologist Brent Berlin (through his studies on colors and on plants and animals naming) had reached a similar conclusion that applies on categories used by adults. The point was that we can name objects in many different ways: a cat is also a feline, a mammal, an animal, and it is also a specific variety of cat. However, we normally call it "a cat". The level at which we "naturally" name objects is the level of what Brown termed "distinctive action". The actions we perform on flowers are pretty much all the same, and certainly different from the actions that we perform on a cat (e.g., one we smell and one we pat). But the actions we perform on two different varieties of cats or two different types of flowers are the same (we pat both the same way, we smell both the same way). Our basic actions tell us that a cat is a cat and a flower is a flower, but they cannot tell us that a rose is not a lily. "Cat" and "flower" represent a "natural" level of categorization. Berlin had found that people categorize plants at the same "basic level" anywhere in the world (which roughly corresponds to the genus in biology). It is a level at which only shape, substance and pattern of change are involved, while no technical details are required. Rosch extended his ideas to artifacts and she found that we also classify artifacts at a "basic level" where technical details are not essential. We first create categories of "chair" and "car", and only later we specialize and generalize those categories (to "armchair", "furniture", "sport car", etc). At the basic level we can form a mental image of the category. We can form a mental image of "chair", but not of "furniture". We can form a mental image of "car", but not of "vehicle". We have a motor program for interacting with "chair", but not with "furniture". We have a motor program for interacting with "car", but not with "vehicle". Categorization initially occurs based on our interaction with the object. Meaning is in the interaction between the body and the world. Rosch postulated a level of abstraction at which the most basic category cuts are made (i.e., where "cue validity" is maximized), which she called the "basic" level. Categories are not merely organized in a hierarchy, from more specific to more general. There is one level of the hierarchy that is somewhat privileged when it comes to perception of form, movement of body parts, organization of knowledge, etc. "Chair" and "car" are examples of basic categories. We can form a mental picture of them. We have a motor program for dealing with them. They are the first ones learned by children. The category of "furniture", for example, is different: I cannot visualize it, I do not have a motor program to deal with it, and it takes some time for a child to learn it. Generalization tends to proceed upwards from this level, and specialization proceeds downward from this level. Superordinate categories are more abstract and more comprehensive. Subordinate categories are less abstract and less comprehensive. The most fundamental perception and description of the world occurs at the level of basic (or natural) categories. Rosch also realized that categories occur in systems, not alone, and they depend on the existence of contrasting categories within the same system. Each contrasting category limits a category (e.g., if a category for birds did not exist, the category for mammals would probably be bigger). At the basic level, categories are maximally distinct, i.e. they maximize perceived similarity among category members and minimize perceived similarities across contrasting categories. Technically, one can use the notion of "cue validity": the conditional probability that an object falls in a particular category given a specific feature. Category cue validity is the sum of all the individual cue-validities of the features associated with a category. The highest cue validity occurs at the basic level. The lowest cue-validities are those for super-ordinate categories. Fuzzy Concepts Later, Rosch recognized that categories are not mutually exclusive (an object can belong to more than one category to different degrees), i.e. that they are fundamentally ambigous. This led to the use of fuzzy logic in studying categorization. For example, the USA linguist George Lakoff borrowed ideas from Wittgenstein's family-resemblance theory, Rosch's prototype theory and Lotfi Zadeh's theory of fuzzy quantities for his theory of "cognitive models". Lakoff started off by demolishing the traditional view of categories: that categories are defined by common features of their members; that thought is the disembodied manipulation of abstract symbols; that concepts are internal representations of external reality; that symbols have meaning by virtue of their correspondence to real objects. Lakoff showed that categories depend on two more factors: the bodily experience of the "categorizer" and the "imaginative processes" (metaphor, metonymy, mental imagery) of the categorizer. Lakoff's theory is based on the assumption of "embodiment of mind": there is no green in the world, but green has to do with the relationship between my body (my eye, my retina, my brain) and the world. Meaning cannot be in the world because things are not in the world: they are in the relationship between us and the world. His close associate, the USA philosopher Mark Johnson, had shown that experience is structured in a meaningful way prior to any concepts: some schemas are inherently meaningful to people by virtue of their bodily experience (e.g., the "container" schema, the "part-whole" schema, the "link" schema, the "center-periphery" schema). We "know" these schemas even before we acquire the related concepts because such "kinesthetic" schemas come with a basic logic that is used to directly "understand" them. Thus Lakoff argued that thought makes use of symbolic structures which are meaningful to begin with (they are directly understood in terms of our physical experience): "basic-level" concepts (which are meaningful because they reflect our sensorimotor life) and kinesthetic image schemas (which are meaningful because they reflect our spatial life). Other meaningful symbolic structures are built up from these elementary ones through imaginative processes such as metaphor. As a corollary, everything we use in language, even the smallest unit, has meaning. And it has meaning not because it refers to something, but because it is either related to our bodily experience or because it is built on top of other meaning-bearing elements. Thought is embodiment of concepts via direct and indirect experience. Concepts grow out of bodily experience and are understood in terms of it. The core of our conceptual system is directly grounded in bodily experience. This explains why Rosch’s basic level is what it is: the one that reflects our bodily nature. Meaning is based on experience. With Putnam, "meaning is not in the mind". But, at the same time, thought is imaginative: those concepts that are not directly grounded in bodily experience are created by imaginative processes such as metaphor. Knowledge is organized into categories by what Lakoff calls "idealized cognitive models". Each model employs four kinds of categorizing processes: "propositional" (which specifies elements, their properties and relations among them in a manner similar to frames); "image-schematic" (which specifies spatial images in a manner similar to Ronald Langacker’s image schemas); "metaphoric" (which maps a propositional or image-schematic model in one domain to a model in another domain); and "metonymic" (which maps an element of a model to another element of the same model). Some models are classical (in that they yield categories that have rigid boundaries and are defined by necessary and sufficient conditions), some models are scalar (they yield categories whose members have only degrees of membership). All models are embodied, i.e. they are linked with bodily experience. Models build what the French linguist Gilles Fauconnier calls "mental spaces", interconnected domains that consist of elements, roles, strategies and relations between them. Mental spaces allow for alternative views of the world. The mind needs to create multiple cognitive spaces in order to engage in creative thought. Lakoff argues that the conceptual system of a mind, far from being one gigantic theory of the world, is normally not consistent. We have available in our minds many different ways of making sense of situations. We constantly keep alternative conceptualizations of the world. The Origin of Categories The German philosopher Immanuel Kant of the 18th century held that experience is possible only if we have knowledge, and knowledge evolves from concepts. Some concepts must therefore be native. We must be born with an infrastructure that allows us to learn concepts and to build concepts on top of concepts. Chomsky proved something similar for language: that human brains are designed to acquire a language, that they contain a "universal grammar" ready to adopt the specific grammar of whatever language we are exposed to. We speak because our brain is meant to speak. Kant, in a sense, stated the same principle for thinking in general: we think in concepts because we are meant to think in concepts. Our mind creates categories because it is equipped with some native categories and a mechanism to build categories on top of existing categories. Just like Chomsky said that grammar is innate and universal, so one can claim that some concepts are innate and universal. Conceptual Holism Inspired by Willard Quine's holism, the USA psychologist Frank Keil argues that concepts are always related to other concepts. No concept can be understood in isolation from all other concepts. Concepts are not simple sets of features. Concepts embody "systematic sets of causal beliefs" about the world and contain implicit explanations about the world. Concepts are embedded in theories about the world, and they can only be understood in the context of such theories. In particular, natural kinds (such as "gold") are not defined by a set of features or by a prototype: they derive their concept from the causal structure that underlies them and explains their superficial features. They are defined by a "causal homeostatic system", which tends to stability over time in order to maximize categorizing. Nominal kinds (e.g., "odd numbers") and artifacts (e.g., "cars") are similarly defined by the theories they are embedded in, although such theories are qualitatively different. There is a continuum between pure nominal kinds and pure natural kinds with increasing "well-definedness" as we move towards natural kinds. What develops over time is the awareness of the network of causal relations and mechanisms that are responsible for the essential properties of a natural kind. The theory explaining a natural kind gets refined over the years. The Mind’s Growth The fundamental feature of the mind is that it is not always the same. Just like every other organ in the body, it undergoes growth. It is not only a matter of memory getting "bigger": the "quality" of the thought system changes in a significant way. What we are capable of doing with our minds changes dramatically during the growth of the mind from childhood to adulthood. It is more than just learning about the environment: the mind literally "grows" into something else, capable of new types of actions. The brain, as well as the rest of the body, undergoes a massive change in shape and volume. Somehow this also results in significant new skills. The balance between "nature" and "nurture" (between "nativism" and "constructivism") is the key to understanding the mind’s growth. Humans are born with "instincts", and then "experience" shapes the mind, i.e. nature and nurture coexist and interact. Constructivism In the 1930s the Swiss psychologist Jean Piaget introduced an important framework to study the growth of mind. The biological context of his ideas is that living beings are in constant interaction with their environment, and survival depends on maintaining a state of equilibrium between the organism and the environment. The organism has to regulate its own behavior in order to continuously adapt to the information flow from the environment. At the same time the behavior of the organism shapes the environment, and, of course, the aim of the organism is to shape the environment so as to maximize the chances of maintaining the vital equilibrium. Cognition, therefore, is but self-regulation. A dynamic exchange between organism and environment is also the basis of his theory of knowledge, which he labeled "genetic epistemology". The cognitive process (the self-regulation) consists in a loop of assimilation and accomodation. This process occurs in stages. The development of children's intellect proceeds from simple mental arrangements to progressively more complex ones not by gradual evolution but by sudden rearrangements of mental operations that produce qualitatively new forms of thought. Cognitive faculties are not fixed at birth but evolve during the lifetime of the individual. First a child lives a "literal" sensorymotor life, in which knowledge of the world is only due to her actions in it. Slowly, the mind creates "schemas" of behavior in the world. Autonomous, self-regulated functioning of "schemas" lead to "interiorized" action. The child begins to deal with internal symbols and introspection. Then the child learns to perform internal manipulations on symbols that represent real objects, i.e. internal action on top of external action. Finally, the mental life extends to abstract objects, besides real objects. This four-step transition leads from a stage in which the dominant factor is perception, which is irreversible, to a stage in which the dominant is thought, which is reversible. Language appears between the sensorymotor and the symbolic stages, and is but one of the elements of symbolic thought. On the contrary of Jerry Fodor's innate "language of thought", symbolic representation is constructed during child development. The mind’s growth is due to the need to maintain a balance between the mind and its knowledge of the world. Rationality is the overall way in which an organism adapts to its environment. Rational action occurs every time the organism needs to solve a problem, i.e. when the organism needs to reach a new form of balance with its environment. Once that balance has been achieved, the organism proceeds by instinct. Rationality will be needed only when the equilibrium is broken again. In conclusion, Piaget did not recognize a major role for any innate knowledge. He only accepted a set of sensory reflexes and three processes: assimilation, accomodation and equilibration. These processes are very general, not specific to any domain. The same processes are supposed to operate on development of language, reasoning, physics, etc. Piaget’s child is a purely sensorymotor device. Piaget's stance is almost behavioristic, except that he grants the child an inner growth. From the social to the personal By studying the behavior of chimpanzees, the Russian psychologist Lev Vygotsky reached the conclusion that thought and speech originate from different processes, and then evolve in parallel but independently of each other. The close correspondence between thought and speech is unique to adult humans. Children initially behave like chimpanzees: language and thought are unrelated (language is irrational, thought is nonverbal). They learn the names of objects only when told so. At some point the attitude changes: it is the child who becomes curious about the names of things. At that point the child's vocabulary increases dramatically, with much less coaching from adults. The child has learned that objects have names, or, equivalently, that one of the properties of an object is its name. At this point in the development of the child, thought and speech merge. Vygotsky redrew Piaget's theory of egocentric speech (the kind of speech that ignores the rest of the world) in pre-school children. One generally becomes aware of her/his actions when they are interrupted. Speech is an expression of the process of becoming aware of one's actions. The egocentric child is no exception: its egocentric speech is the sign of a process of becoming aware after something disrupted the action underway. In other words, the child is thinking aloud. A few years later this process has become silent: when the child needs to find a solution to a problem, the "thinking" is no longer aloud, has become an inner conversation. When egocentric speech disappears, it still exists, but it moved inside. The reason it is no longer "vocal" is because it does not serve a social function anymore (it does not need to be heard by others). According to Piaget, social speech follows egocentric speech, but Vygotsky believes that speech is originally social in nature, and egocentric speech is a specialization of it to the case when the child has to reflect. Egocentric speech is an evolution of social speech that eventually becomes silent thought. Cognitive faculties are internalized versions of social processes. The unit of verbal thought is word meaning, which also represents the fusion of thought and speech. Adults and children use the same word to refer to the same object, but Vygotsky believes that the meanings are different. Vygotsky believes that the meanings of words evolve during childhood. Word meanings are dynamic, not static, entities. Thought is therefore determined by language. And both are determined by society. Language provides a semiotic mediation of knowledge. Language guides the child's cognitive growth. Cognition thus develops in different ways depending on the cultural conditions. Vygotsky thinks that higher mental functions, too, have social origins. Language is a system of signs that the individual needs in order to interact with the environment and only afterwards it is interiorized and can be utilized to express thought. The meaning of a word is initially a purely emotional fact. Only with time it will acquire a precise reference to an object and then an abstract meaning. Child development is a sequence of stages that lead to the transformation of an interpersonal process into an intrapersonal process. Children think by memorizing, while adults memorize by thinking. In children something is memorized, in adults the individual memorizes something. In the former case a link is created because of the simultaneous occurrence of two stimuli. In the latter case the individual creates that link. Remembering is transformed into an external activity. Humans are then able to influence their relation with the environment and through that environment change their own behavior. The mastering of nature and the mastering of behavior are interdependent. Some nativism Piaget's model of cognitive development constitutes a powerful paradigm but does not explain everything. In particular, Piaget's theory is inadequate to explain how children learn language. Without any a-priori knowledge of language, it would be terribly difficult to learn the theory of language that every child eventually learns. The British psychologist Annette Karmiloff-Smith, a student of Piaget, has proposed a model of child development that bridges Fodor's nativism (built-in knowledge) and Piaget's constructivism (learning), i.e. innate capacities of the human mind and subsequent representational changes. Karmiloff-Smith envisions a mind that is both equipped with some innate capacities and that grows through a sequence of subsequent changes. Karmiloff-Smith's child is genetically pre-wired to absorb and organize information in an appropriate format. Each module develops independently, as proved by children who exhibit one mental disorder but are perfectly capable in other ways. Karmiloff-Smith's starting point is Fodor's model of the mind (that the mind is made of a number of independent, specialized modules), but, based on evidence of the brain's plasticity (the brain can restructure itself to adapt to an early damage), Karmiloff-Smith believes that modules are not static and that they "grow" during the child's development, and that new modules are created during child's development ("gradual modularization"). She points out that children display from the very beginning a whole array of cognitive skills, albeit still unrelated and specific (for example, identifying sounds, imitating other people's movements, recognizing the shapes of faces). Therefore, the child must be born with a set of pre-wired modules that account for these cognitive skills. Somehow, during development the modules start interacting and working together and adult life takes shape. Initially, children learn by instinct, or at least "implicitly". Then their thinking develops, and consists of redescribing the world from an implicit form to more and more explicit forms, to more and more verbal knowledge. Of course, the environment that drives the mind's growth also includes the other individuals. Education and playing are forms of influencing the evolution of the thought system of a child. Karmiloff-Smith notes a thread that is common to several spheres of cognition: the passage from procedural non-expert to the automatic (nonprocedural) expert also involves a parallel passage from implicit to explicit knowledge (from executing mechanically to understanding how it works). Child development is not only about learning new procedures, it is about building theories of why those procedures do what they do. This "representational redescription" occurs through three stages: first the child learns to become a master of some activity; then she analyzes introspectively what she has learned; and, finally, she reconciles her performance with her introspection. At this point the child has created a "theory" of why things work the way they work. Therefore Karmiloff-Smith admits cognitive progress like Piaget, but her "representational redescription" occurs when the child has reached a stable state (mastery), whereas in Piaget's model progress only occurs when the child is in a state of disequilibrium. This process of "representational redescription" involves re-coding information from one representational format (the procedural one) to another (a quasi-linguistic format). There are therefore different levels at which knowledge is encoded (in contrast with Fodor's language of thought). The same "redescription" process operates within each module, but not necessarily at the same pace. In each field, children acquire domain-specific principles that augment the general-purpose principles (such as representational redescription) that guide their cognitive life. Moreover, the cultural context determines which modules arise. Finally, mapping across domains is a fundamental achievement by the child's mind. Another USA developmental psychologist, Patricia Greenfield, modified this theory by showing that initially the child's mind has no modules, only a general-purpose learning system. Modules start developing later in the life of the child. Greenfield identified a common neurological layer in the early stages of development (up to two years old) that accounts for both linguistic skills and object manipulation skills. As the child's brain develops, those skills split.   The Evolution of Memories German anthropologist Michael Tomasello believes that humans are genetically equipped with the "ability to identify with conspecifics." This attitude is particularly visible in children, who spend most of their time imitating others. Natural selection rewarded the best "copycats". This ability is crucial to the development of social skills such as language: Children learn a language because their brains are predisposed to "identify with conspecifics" and because they are exposed to such conspecifics (they socialize with speaking humans). However, this skill is shared by other primates, who nonetheless never achieve the sophistication of human civilizations: they only learn by imitation, without any clue of why others do what they do. They are are mere copycats. The key factor in the development of human civilizations (which Tomasello thinks emerges in the ninth month of life) is the understanding of others as goal-directed agents. Tomasello thinks that recognizing the intentions of others is crucial to "learn" from previous generations. Humans do not just imitate other humans: humans also understand why other humans did what they did. Tomasello thinks that this is the secret of rapid learning and of transmission of learned knowledge from one generation to the next one. Over evolutionary and historical time, the "ratchet effect" due to this attitude to imitation generates the civilizations we are familiar with, civilizations that other primates cannot even dream of. Evolutionarily speaking, the progress made by the human species is impressive. Tomasello believes that the secret to such speedy evolution lies in the unique human attitude towards conspecifics. The Evolution of Memories Memory is not a storage, because it cannot recall events exactly the way they were. Memories change all the time, therefore memory is not a static system, it is a dynamic system. Memory is pivotal for the entire thought system of the individual. Therefore, memory is about thought, it is not limited to remembering. Memory stores and retrieves thoughts. Memory can be viewed as an evolving population of thoughts. Thoughts that survive and reproduce are variations of original thoughts, and somehow "contain" those original thoughts, but adapted to the new circumstances. Memories are descendants of thoughts that occurred in the past. Thoughts are continuously generated from previous ones, just like the immune system generates antibodies all the time and just like species are created from previous ones. Memory, far from being a static storage, is changing continuously. It is not a location, it is the collective process of thinking. Further Reading Baddeley, Alan: YOUR MEMORY (MacMillan, 1982) Baddeley, Alan: WORKING MEMORY (Clarendon Press, 1986) Baddeley, Alan: HUMAN MEMORY (Simon & Schuster, 1990) Barsalou, Lawrence: COGNITIVE PSYCHOLOGY (Lawrence Erlbaum, 1992) Bartlett, Frederic Charles: REMEMBERING (1932) (Cambridge Univ Press, 1967) Broadbent, Donald: PERCEPTION AND COMMUNICATION (Pergamon, 1958) Broadbent, Donald: DECISION AND STRESS (Academic Press, 1971) Bruner Jerome: A STUDY OF THINKING (Wiley, 1956) Campbell, Joseph: PRIMITIVE MYTHOLOGY: THE MASKS OF GODS (Viking, 1959) Collins, Alan: THEORIES OF MEMORY (Lawrence Erlbaum, 1993) Crowder, Robert: PRINCIPLES OF LEARNING AND MEMORY (Erlbaum, 1976) Eichenbaum, Howard: COGNITIVE NEUROSCIENCE OF MEMORY (Oxford Univ Press, 2002) Estes, William: CLASSIFICATION AND COGNITION (Oxford University Press, 1994) Fauconnier, Gilles: MENTAL SPACES (MIT Press, 1994) Greene, Robert: HUMAN MEMORY (Lawrence Erlbaum, 1992) Greenfield, Patricia: LANGUAGE, TOOLS, AND BRAIN (1991) James, William: PRINCIPLES OF PSYCHOLOGY (1890) Karmiloff-Smith Annette: BEYOND MODULARITY (MIT Press, 1992) Keil Frank: CONCEPTS, KINDS AND COGNITIVE DEVELOPMENT (Cambridge University Press, 1989) Lakoff, George: WOMEN, FIRE AND DANGEROUS THINGS (Univ of Chicago Press, 1987) Lenneberg, Eric: BIOLOGICAL FOUNDATIONS OF LANGUAGE (Wiley, 1967) Piaget, Jean: EQUILIBRATION OF COGNITIVE STRUCTURES (University of Chicago Press, 1985) Reiser, Morton: MEMORY IN MIND AND BRAIN (Basic, 1990) Roediger, Henry: VARIETIES OF MEMORY AND CONSCIOUSNESS (Lawrence Erlbaum, 1989) Rosch, Eleanor: COGNITION AND CATEGORIZATION (Erlbaum, 1978) Schacter, Daniel & Tulving Endel: MEMORY SYSTEMS (MIT Press, 1994) Schacter, Daniel: SEARCHING FOR MEMORY (Basic Books, 1996) Schanks, David: HUMAN MEMORY: A READER (Oxford Univ Press, 1997) Semon, Richard: DIE MNEME (1904) Tolman, Edward: PURPOSIVE BEHAVIOR IN ANIMALS AND MEN (1932) Tulving, Endel: ORGANIZATION OF MEMORY (Academic Press, 1972) Tulving, Endel: ELEMENTS OF EPISODIC MEMORY (Oxford Univ Press, 1983) Tulving, Endel: OXFORD HANDBOOK OF MEMORY (Oxford Univ Press, 2000) Wittgenstein, Ludwig: PHILOSOPHICAL INVESTIGATIONS (Macmillan, 1953)